benchmark.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# RGBD Benchmark\n",
    "## Preparation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Envs_state_manager initialized, sys.modules and sys.path saved\n"
     ]
    }
   ],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "# must run this cell before any other cells in this notebook!\n",
    "import sys\n",
    "from bench_utils import *\n",
    "\n",
    "results = {\"nyudv2\": {}, \"sunrgbd\": {}}\n",
    "envs_helper = (\n",
    "    Envs_state_manager()\n",
    ")  # make sure those models are not disturbed by other models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Calculating the FLOPs and parameters"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'utils.utils', 'repositories.ACNet', 'repositories', 'utils', 'repositories.ACNet.ACNet_models_V1'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/ACNet\")\n",
    "from repositories.ACNet import ACNet_models_V1\n",
    "\n",
    "model = ACNet_models_V1.ACNet(num_class=40, pretrained=False).eval()\n",
    "results[\"nyudv2\"][\"ACNet\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = ACNet_models_V1.ACNet(num_class=37, pretrained=False).eval()\n",
    "results[\"sunrgbd\"][\"ACNet\"] = cal_flops_params(\n",
    "    model, shape=(544, 736), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")  # Since ACNet only accepts resolution of multiple of 32(2^5), we use 544x736\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/AsymFormer/src/convnext.py:159: UserWarning: Overwriting convnext_tiny in registry with src.convnext.convnext_tiny. This is because the name being registered conflicts with an existing name. Please check if this is not expected.\n",
      "  def convnext_tiny(pretrained=False,in_22k=False, **kwargs):\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/AsymFormer/src/convnext.py:168: UserWarning: Overwriting convnext_small in registry with src.convnext.convnext_small. This is because the name being registered conflicts with an existing name. Please check if this is not expected.\n",
      "  def convnext_small(pretrained=False,in_22k=False, **kwargs):\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/AsymFormer/src/convnext.py:177: UserWarning: Overwriting convnext_base in registry with src.convnext.convnext_base. This is because the name being registered conflicts with an existing name. Please check if this is not expected.\n",
      "  def convnext_base(pretrained=False, in_22k=False, **kwargs):\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/AsymFormer/src/convnext.py:186: UserWarning: Overwriting convnext_large in registry with src.convnext.convnext_large. This is because the name being registered conflicts with an existing name. Please check if this is not expected.\n",
      "  def convnext_large(pretrained=False, in_22k=False, **kwargs):\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/AsymFormer/src/convnext.py:195: UserWarning: Overwriting convnext_xlarge in registry with src.convnext.convnext_xlarge. This is because the name being registered conflicts with an existing name. Please check if this is not expected.\n",
      "  def convnext_xlarge(pretrained=False, in_22k=False, **kwargs):\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'timm.models.registry', 'src.mix_transformer', 'repositories.AsymFormer.src.AsymFormer', 'thop.profile', 'thop.__version__', 'repositories.AsymFormer', 'src.convnext', 'repositories', 'thop', 'repositories.AsymFormer.src', 'thop.vision.calc_func', 'src', 'thop.rnn_hooks', 'thop.utils', 'src.MLPDecoder', 'timm.models.layers', 'thop.vision', 'thop.vision.basic_hooks', 'distutils.version'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/AsymFormer\")\n",
    "from repositories.AsymFormer.src.AsymFormer import B0_T\n",
    "\n",
    "model = B0_T(num_classes=40).eval()\n",
    "results[\"nyudv2\"][\"AsymFormer\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = B0_T(num_classes=37).eval()\n",
    "results[\"sunrgbd\"][\"AsymFormer\"] = cal_flops_params(\n",
    "    model, shape=(544, 736), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")  # Since AsymFormer only accepts resolution of multiple of 32(2^5), we use 544x736\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/calflops/pytorch_ops.py:364: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.\n",
      "  return oldFunc(*args, **kwds)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'cv2.gapi.oak', 'cv2.misc.version', 'cv2.gapi.imgproc', 'cv2.gapi.imgproc.fluid', 'cv2.ipp', 'repositories.CEN.semantic_segmentation', 'cv2.segmentation', 'cv2.load_config_py3', 'cv2.version', 'cv2.gapi.core.fluid', 'cv2.gapi.onnx.ep', 'cv2.typing', 'utils', 'utils.meter', 'cv2.cuda', 'cv2.gapi.video', 'cv2.qt', 'cv2.ogl', 'cv2.gapi.ie.detail', 'cv2.gapi.render.ocv', 'cv2.gapi.ov', 'cv2.gapi.ot.cpu', 'repositories.CEN.semantic_segmentation.models.modules', 'cv2.barcode', 'repositories', 'cv2.gapi.streaming', 'repositories.CEN.semantic_segmentation.models', 'cv2.gapi.own', 'utils.helpers', 'cv2.gapi.render', 'cv2.parallel', 'cv2.utils', 'cv2.aruco', 'cv2.gapi.wip.gst', 'cv2.gapi.wip', 'cv2.Error', 'cv2.samples', 'repositories.CEN.semantic_segmentation.models.model', 'cv2.gapi.ie', 'cv2.ocl', 'repositories.CEN', 'cv2.gapi.core', 'cv2.gapi.onnx', 'cv2.gapi', 'cv2.dnn', 'cv2.misc', 'cv2.utils.fs', 'cv2.gapi.wip.onevpl', 'cv2.mat_wrapper', 'cv2.fisheye', 'cv2.gapi.own.detail', 'cv2.utils.nested', 'cv2.ml', 'cv2.gapi.core.cpu', 'cv2.gapi.wip.draw', 'cv2.detail', 'cv2.data', 'cv2.gapi.core.ocl', 'cv2.flann', 'cv2.videoio_registry', 'cv2.gapi.ot', 'cv2'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"./repositories/CEN/semantic_segmentation\")\n",
    "from repositories.CEN.semantic_segmentation.models.model import refinenet\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    refinenet(num_layers=101, num_classes=40, num_parallel=2, bn_threshold=2e-2).eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"CEN_101\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    refinenet(num_layers=152, num_classes=40, num_parallel=2, bn_threshold=2e-2).eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"CEN_152\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    refinenet(num_layers=101, num_classes=37, num_parallel=2, bn_threshold=2e-2).eval()\n",
    ")\n",
    "results[\"sunrgbd\"][\"CEN_101\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    refinenet(num_layers=152, num_classes=37, num_parallel=2, bn_threshold=2e-2).eval()\n",
    ")\n",
    "results[\"sunrgbd\"][\"CEN_152\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'mmcv.cnn.bricks.depthwise_separable_conv_module', 'fvcore.nn.print_model_statistics', 'yapf_third_party._ylib2to3.pgen2.pgen', 'mmengine.optim.optimizer.base', 'mmcv.visualization', 'cv2.cuda', 'mmengine.optim.scheduler', 'cv2.qt', 'mmengine._strategy.deepspeed', 'mmengine.optim.scheduler.momentum_scheduler', 'mmengine.runner', 'iopath.common', 'repositories', 'mmcv', 'mmengine.registry.build_functions', 'cv2.gapi.wip.gst', 'mmengine.fileio.backends', 'cv2.samples', 'importlib_metadata.compat', 'rich.table', 'mmcv.cnn.bricks.norm', 'rich.box', 'mmengine.optim.optimizer.builder', 'yapf_third_party._ylib2to3.patcomp', 'iopath.common.download', 'mmcv.cnn', 'fvcore.nn.smooth_l1_loss', 'mmcv.visualization.image', 'attr._config', 'mmengine._strategy.fsdp', 'rich', 'mmengine.utils.misc', 'mmcv.cnn.bricks.drop', 'yapf_third_party._ylib2to3.pgen2.tokenize', 'transformers.pytorch_utils', 'torch.distributed.optim.functional_adamax', 'cv2.gapi.core.fluid', 'mmengine._strategy', 'mmengine.hooks.runtime_info_hook', 'semseg.models.modules.ppm', 'mmengine.utils.path', 'zipp', 'rich._export_format', 'rich._palettes', 'iopath.version', 'mmcv.transforms.base', 'cv2.gapi.wip', 'torch.distributed.optim.zero_redundancy_optimizer', 'cv2.misc', 'rich.live', 'semseg.datasets.urbanlf', 'semseg.models.base', 'rich.protocol', 'mmengine.optim.optimizer.amp_optimizer_wrapper', 'mmcv.transforms.builder', 'mmengine.fileio.io', 'attr.converters', 'yapf_third_party._ylib2to3.fixer_util', 'semseg.models.modules.mspa', 'mmengine.fileio.backends.registry_utils', 'mmengine.registry.default_scope', 'einops.einops', 'semseg.models.layers.initialize', 'repositories.CMNeXt.semseg.models', 'yapf.pyparser.pyparser_utils', 'mmcv.cnn.rfsearch', 'cv2.parallel', 'mmengine.utils.dl_utils.setup_env', 'mmengine.visualization.vis_backend', 'iopath.tabular', 'mmengine.model.wrappers.distributed', 'yapf_third_party._ylib2to3.pgen2.parse', 'mmengine.model.wrappers.seperate_distributed', 'cv2.utils.fs', 'semseg.datasets.mcubes', 'mmengine.registry', 'cv2.mat_wrapper', 'semseg.models.backbones.cmx', 'transformers.optimization', 'mmengine.runner.base_loop', 'yapf.pytree.comment_splicer', 'torch.distributed.optim.functional_rprop', 'mmengine.utils.dl_utils.parrots_wrapper', 'mmengine.dist', 'yapf.pytree.pytree_utils', 'portalocker.exceptions', 'addict.addict', 'rich.file_proxy', 'mmengine.structures.base_data_element', 'termcolor', 'cv2.gapi.imgproc', 'tabulate', 'rich._loop', 'mmengine.runner.loops', 'rich.filesize', 'yapf.pytree', 'rich._fileno', 'yapf.yapflib.format_token', 'rich.pretty', 'portalocker.constants', 'importlib_metadata._meta', 'yapf.yapflib.identify_container', 'rich.containers', 'mmengine.device', 'yapf.yapflib.file_resources', 'cv2.gapi.core.cpu', 'mmengine.model.weight_init', 'mmengine.device.utils', 'mmengine.optim.optimizer.zero_optimizer', 'cv2.gapi.core.ocl', 'mmengine.runner.priority', 'mmcv.image.photometric', 'mmengine.fileio', 'mmengine.hooks.naive_visualization_hook', 'semseg.models.modules.ffm', 'cv2.gapi.oak', 'einops._backends', 'yapf.yapflib.line_joiner', 'portalocker', 'rich.errors', 'rich.markup', 'mmengine.structures.instance_data', 'importlib_metadata.compat.py39', 'yapf.yapflib', 'mmengine.utils.version_utils', 'rich._wrap', 'yapf.pytree.blank_line_calculator', 'mmengine.evaluator', 'cv2.gapi.streaming', 'iopath.common.non_blocking_io', 'mmcv.transforms.formatting', 'mmengine.hooks.early_stopping_hook', 'repositories.CMNeXt.semseg.models.cmnext', 'mmcv.cnn.bricks.generalized_attention', 'fvcore.nn.precise_bn', 'mmcv.arraymisc', 'mmengine.dataset.utils', 'rich.cells', 'yapf_third_party._ylib2to3', 'semseg.models.heads.fpn', 'fvcore.nn.parameter_count', 'mmengine.fileio.handlers.pickle_handler', 'mmcv.video', 'mmengine.model.wrappers.utils', 'mmengine.visualization', 'mmcv.cnn.bricks.upsample', 'mmengine.structures', 'torch.distributed.optim.functional_adadelta', 'torch.distributed.optim.named_optimizer', 'rich.region', 'torch.distributed.algorithms.model_averaging', 'mmengine.model.averaged_model', 'rich.measure', 'cv2.segmentation', 'fvcore.nn.flop_count', 'rich._emoji_replace', 'semseg', 'mmcv.cnn.resnet', 'rich.highlighter', 'yapf.pytree.subtype_assigner', 'yapf.pyparser.pyparser', 'mmcv.transforms.loading', 'yapf_third_party._ylib2to3.pgen2.driver', 'mmcv.cnn.bricks.conv_ws', 'importlib_metadata._functools', 'mmengine.hooks.empty_cache_hook', 'tifffile._imagecodecs', 'mmengine.hooks.sampler_seed_hook', 'mmengine.runner.checkpoint', 'addict', 'repositories.CMNeXt.semseg', 'cv2.flann', 'iopath', 'cv2.gapi.ot', 'mmengine.utils.dl_utils.time_counter', 'rich.default_styles', 'mmcv.cnn.utils.flops_counter', 'rich.themes', 'torch.distributed.algorithms.model_averaging.utils', 'fvcore.nn.giou_loss', 'importlib_metadata._text', 'mmcv.cnn.bricks.swish', 'mmengine.fileio.handlers.yaml_handler', 'tifffile.tifffile', 'mmcv.transforms.wrappers', 'iopath.common.event_logger', 'mmengine.registry.registry', 'attr.validators', 'cv2.gapi.own', 'mmengine.evaluator.metric', 'mmengine.model.base_model.data_preprocessor', 'mmengine.registry.root', 'semseg.models.modules.psa', 'attr', 'mmengine.model.efficient_conv_bn_eval', 'semseg.models.layers', 'mmcv.visualization.optflow', 'cv2.gapi.core', 'mmengine.dataset.sampler', 'cv2.fisheye', 'mmengine.optim', 'iopath.tabular.tabular_io', 'rich.spinner', 'mmengine.fileio.backends.local_backend', 'rich.pager', 'mmengine.fileio.handlers.base', 'semseg.datasets.kitti360', 'mmengine.logging.logger', 'cv2.videoio_registry', 'mmengine.dist.utils', 'mmengine', 'mmengine.hooks.profiler_hook', 'mmengine.config.lazy', 'cv2.version', 'semseg.models.layers.common', 'mmengine.fileio.backends.http_backend', 'semseg.models.cmnext', 'mmengine.model.utils', 'mmcv.image.misc', 'repositories.CMNeXt', 'repositories.CMNeXt.semseg.models.cmx', 'yapf_third_party._ylib2to3.pytree', 'mmcv.version', 'rich.live_render', 'cv2.utils', 'mmengine.utils.timer', 'semseg.models.heads.lawin', 'mmengine.model.base_module', 'transformers.trainer_pt_utils', 'mmcv.cnn.bricks.conv2d_adaptive_padding', 'yapf.pyparser', 'cv2.gapi', 'mmengine.fileio.handlers', 'attr._cmp', 'rich.repr', 'mmengine._strategy.base', 'mmengine.logging.message_hub', 'semseg.models.heads.hem', 'rich.segment', 'cv2.ipp', 'mmengine.utils.dl_utils.collect_env', 'mmengine.model', 'cv2.load_config_py3', 'attr.setters', 'mmengine.model.test_time_aug', 'rich.scope', 'rich._ratio', 'semseg.models.backbones.cmnext', 'fvcore.nn', 'rich.console', 'mmengine.hooks.checkpoint_hook', 'fvcore.nn.jit_analysis', 'mmcv.cnn.bricks.scale', 'mmcv.cnn.rfsearch.utils', 'yapf.pytree.split_penalty', 'torch.distributed.optim.functional_sgd', 'cv2.barcode', 'mmcv.cnn.bricks.non_local', 'mmcv.cnn.bricks.padding', 'mmengine.hooks.ema_hook', 'rich.terminal_theme', 'mmengine.fileio.file_client', 'cv2.Error', 'mmcv.image.io', 'cv2.gapi.ie', 'mmcv.video.optflow', 'attr._funcs', 'rich.color_triplet', 'mmengine.utils.package_utils', 'rich.screen', 'mmengine.model.wrappers', 'portalocker.portalocker', 'rich._extension', 'importlib_metadata._compat', 'portalocker.utils', 'einops.packing', 'mmcv.video.processing', 'semseg.models.heads.fapn', 'mmengine.fileio.handlers.registry_utils', 'mmengine.version', 'mmengine.optim.scheduler.lr_scheduler', 'mmengine.optim.optimizer.optimizer_wrapper_dict', 'fvcore.common.checkpoint', 'cv2.gapi.ie.detail', 'zipp.glob', 'rich.ansi', 'yapf_third_party._ylib2to3.pygram', 'rich.constrain', 'mmengine.runner._flexible_runner', 'mmcv.image', 'fvcore.nn.weight_init', 'mmengine.evaluator.utils', 'fvcore', 'mmengine.dataset', 'cv2.gapi.onnx', 'yapf.pytree.continuation_splicer', 'mmengine.hooks.logger_hook', 'rich.emoji', 'yapf', 'mmengine._strategy.utils', 'cv2.data', 'mmengine.config', 'mmengine.utils.dl_utils', 'mmcv.cnn.bricks.hswish', 'mmengine.fileio.backends.memcached_backend', 'cv2.gapi.onnx.ep', 'cv2.typing', 'yapf.yapflib.format_decision_state', 'mmengine.utils.dl_utils.torch_ops', 'mmengine.runner.log_processor', 'yapf.yapflib.subtypes', 'semseg.models', 'yapf.yapflib.yapf_api', 'mmengine._strategy.colossalai', 'rich.palette', 'mmengine.utils.dl_utils.hub', 'mmengine.utils', 'mmengine.model.wrappers.fully_sharded_distributed', 'semseg.models.backbones', 'mmcv.cnn.bricks.hsigmoid', 'cv2.ocl', 'mmengine.utils.dl_utils.misc', 'torch.distributed.optim.post_localSGD_optimizer', 'torch.distributed.optim.functional_adagrad', 'mmengine.fileio.backends.base', 'mmcv.image.geometric', 'semseg.augmentations_mm', 'timm.models.layers', 'einops.parsing', 'mmcv.cnn.vgg', 'semseg.models.heads.segformer', 'cv2.detail', 'mmengine.visualization.visualizer', 'mmcv.cnn.bricks.conv', 'termcolor.termcolor', 'mmengine.structures.label_data', 'rich._emoji_codes', 'cv2.misc.version', 'mmengine.utils.dl_utils.trace', 'rich.style', 'mmengine.optim.optimizer.apex_optimizer_wrapper', 'rich.padding', 'mmengine.fileio.handlers.json_handler', 'rich.jupyter', 'mmengine.optim.optimizer', 'torch.distributed.optim.functional_adam', 'mmengine.dist.dist', 'torch.distributed.optim', 'rich.progress', 'mmcv.cnn.rfsearch.operator', 'rich.panel', 'yapf.yapflib.errors', 'yapf_third_party._ylib2to3.pgen2.literals', 'mmengine.logging', 'cv2.dnn', 'yapf.pyparser.split_penalty_visitor', 'mmengine.runner.runner', 'attr._make', 'mmcv.transforms.utils', 'yapf.pytree.pytree_visitor', 'mmengine.utils.progressbar', 'mmcv.cnn.bricks.plugin', 'mmcv.transforms', 'attr.exceptions', 'mmengine.dataset.base_dataset', 'attr._version_info', 'mmengine.logging.history_buffer', 'semseg.models.cmx', 'yapf.yapflib.reformatter', 'torch.distributed.optim.apply_optimizer_in_backward', 'cv2.gapi.ov', 'torch.distributed.algorithms.model_averaging.averagers', 'semseg.models.heads.fcn', 'rich.styled', 'mmengine._strategy.distributed', 'semseg.datasets.mfnet', 'semseg.datasets.deliver', 'mmcv.video.io', 'semseg.utils.utils', 'zipp.compat.py310', 'mmengine.model.base_model.base_model', 'yapf_third_party._ylib2to3.pgen2.token', 'mmengine.registry.utils', 'semseg.models.modules', 'torch.distributed.optim.utils', 'mmengine.config.utils', 'importlib_metadata._collections', 'mmengine.utils.manager', 'iopath.common.file_io', 'tabulate.version', 'semseg.models.heads.condnet', 'mmengine.fileio.parse', 'mmcv.image.colorspace', 'mmengine.structures.pixel_data', 'cv2.ogl', 'yapf.yapflib.style', 'mmengine.utils.progressbar_rich', 'rich._spinners', 'termcolor._types', 'cv2.gapi.render.ocv', 'semseg.models.heads.upernet', 'mmcv.cnn.bricks', 'mmcv.cnn.utils', 'yapf_third_party._ylib2to3.pgen2.grammar', 'rich._null_file', 'mmengine.dataset.dataset_wrapper', 'rich.progress_bar', 'mmengine.optim.scheduler.param_scheduler', 'mmcv.visualization.color', 'mmengine.runner.activation_checkpointing', 'cv2.utils.nested', 'mmengine.hooks.hook', 'mmengine.hooks.test_time_aug_hook', 'mmcv.cnn.bricks.wrappers', 'cv2', 'mmengine.hooks', 'transformers.trainer_utils', 'yapf_third_party', 'mmengine.config.config', 'mmcv.cnn.rfsearch.search', 'semseg.datasets', 'yapf.pytree.pytree_unwrapper', 'mmengine.runner.amp', 'fvcore.common', 'torch.distributed.optim.functional_rmsprop', 'semseg.utils', 'mmengine.optim.optimizer.optimizer_wrapper', 'mmcv.cnn.bricks.context_block', 'mmcv.cnn.bricks.conv_module', 'mmengine.fileio.backends.lmdb_backend', 'yapf.yapflib.split_penalty', 'cv2.gapi.render', 'rich.text', 'cv2.aruco', 'portalocker.__about__', 'rich._log_render', 'mmengine.model.base_model', 'zipp.compat', 'mmengine.visualization.utils', 'cv2.gapi.wip.onevpl', 'mmengine.hooks.iter_timer_hook', 'semseg.models.heads', 'importlib_metadata._adapters', 'cv2.gapi.own.detail', 'cv2.gapi.wip.draw', 'cv2.ml', 'mmengine.hooks.param_scheduler_hook', 'importlib_metadata._itertools', 'mmcv.cnn.utils.fuse_conv_bn', 'torch.distributed.optim.optimizer', 'mmcv.transforms.processing', 'rich.abc', 'einops', 'transformers.integrations', 'attr._next_gen', 'cv2.gapi.imgproc.fluid', 'mmengine.runner.utils', 'attr._compat', 'mmengine.hooks.sync_buffer_hook', 'transformers.integrations.deepspeed', 'importlib_metadata', 'cv2.gapi.video', 'rich.theme', 'yapf.yapflib.logical_line', 'fvcore.nn.jit_handles', 'yapf.yapflib.object_state', 'mmcv.cnn.bricks.activation', 'mmengine.optim.optimizer.default_constructor', 'cv2.gapi.ot.cpu', 'torch.distributed.optim.functional_adamw', 'mmcv.arraymisc.quantization', 'semseg.models.heads.sfnet', 'yapf_third_party._ylib2to3.pgen2', 'attr.filters', 'fvcore.nn.focal_loss', 'rich.control', 'fvcore.nn.activation_count', 'mmcv.cnn.alexnet', 'rich.align', 'semseg.datasets.nyu', 'mmengine.evaluator.evaluator', 'mmengine.fileio.backends.petrel_backend', 'tifffile', 'rich.color', 'mmengine._strategy.single_device', 'rich._pick', 'rich._cell_widths'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/CMNeXt\")\n",
    "from repositories.CMNeXt.semseg.models.cmnext import CMNeXt\n",
    "\n",
    "model = decorate_model_for_zip_input(CMNeXt(\"CMNeXt-B4\", 40, [\"img\", \"depth\"]))\n",
    "results[\"nyudv2\"][\"CMNeXt-B4\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(CMNeXt(\"CMNeXt-B4\", 37, [\"img\", \"depth\"]))\n",
    "results[\"sunrgbd\"][\"CMNeXt-B4\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'mmcv.cnn.bricks.depthwise_separable_conv_module', 'fvcore.nn.print_model_statistics', 'yapf_third_party._ylib2to3.pgen2.pgen', 'mmengine.optim.optimizer.base', 'mmcv.visualization', 'cv2.cuda', 'mmengine.optim.scheduler', 'cv2.qt', 'mmengine._strategy.deepspeed', 'mmengine.optim.scheduler.momentum_scheduler', 'mmengine.runner', 'iopath.common', 'repositories', 'mmcv', 'mmengine.registry.build_functions', 'cv2.gapi.wip.gst', 'mmengine.fileio.backends', 'cv2.samples', 'importlib_metadata.compat', 'rich.table', 'mmcv.cnn.bricks.norm', 'rich.box', 'mmengine.optim.optimizer.builder', 'yapf_third_party._ylib2to3.patcomp', 'iopath.common.download', 'mmcv.cnn', 'fvcore.nn.smooth_l1_loss', 'mmcv.visualization.image', 'attr._config', 'mmengine._strategy.fsdp', 'rich', 'mmengine.utils.misc', 'mmcv.cnn.bricks.drop', 'yapf_third_party._ylib2to3.pgen2.tokenize', 'torch.distributed.optim.functional_adamax', 'cv2.gapi.core.fluid', 'mmengine._strategy', 'mmengine.hooks.runtime_info_hook', 'semseg.models.modules.ppm', 'mmengine.utils.path', 'zipp', 'rich._export_format', 'rich._palettes', 'iopath.version', 'mmcv.transforms.base', 'cv2.gapi.wip', 'torch.distributed.optim.zero_redundancy_optimizer', 'cv2.misc', 'rich.live', 'semseg.datasets.urbanlf', 'semseg.models.base', 'rich.protocol', 'mmengine.optim.optimizer.amp_optimizer_wrapper', 'mmcv.transforms.builder', 'mmengine.fileio.io', 'attr.converters', 'yapf_third_party._ylib2to3.fixer_util', 'semseg.models.modules.mspa', 'mmengine.fileio.backends.registry_utils', 'mmengine.registry.default_scope', 'einops.einops', 'semseg.models.layers.initialize', 'repositories.CMNeXt.semseg.models', 'yapf.pyparser.pyparser_utils', 'mmcv.cnn.rfsearch', 'cv2.parallel', 'mmengine.utils.dl_utils.setup_env', 'mmengine.visualization.vis_backend', 'iopath.tabular', 'mmengine.model.wrappers.distributed', 'yapf_third_party._ylib2to3.pgen2.parse', 'mmengine.model.wrappers.seperate_distributed', 'cv2.utils.fs', 'semseg.datasets.mcubes', 'mmengine.registry', 'cv2.mat_wrapper', 'semseg.models.backbones.cmx', 'mmengine.runner.base_loop', 'yapf.pytree.comment_splicer', 'torch.distributed.optim.functional_rprop', 'mmengine.utils.dl_utils.parrots_wrapper', 'mmengine.dist', 'yapf.pytree.pytree_utils', 'portalocker.exceptions', 'addict.addict', 'rich.file_proxy', 'mmengine.structures.base_data_element', 'termcolor', 'cv2.gapi.imgproc', 'rich._loop', 'tabulate', 'mmengine.runner.loops', 'rich.filesize', 'yapf.pytree', 'rich._fileno', 'yapf.yapflib.format_token', 'rich.pretty', 'portalocker.constants', 'importlib_metadata._meta', 'yapf.yapflib.identify_container', 'rich.containers', 'mmengine.device', 'yapf.yapflib.file_resources', 'cv2.gapi.core.cpu', 'mmengine.model.weight_init', 'mmengine.device.utils', 'mmengine.optim.optimizer.zero_optimizer', 'cv2.gapi.core.ocl', 'mmengine.runner.priority', 'mmcv.image.photometric', 'mmengine.fileio', 'mmengine.hooks.naive_visualization_hook', 'semseg.models.modules.ffm', 'cv2.gapi.oak', 'einops._backends', 'yapf.yapflib.line_joiner', 'portalocker', 'rich.errors', 'rich.markup', 'mmengine.structures.instance_data', 'importlib_metadata.compat.py39', 'yapf.yapflib', 'mmengine.utils.version_utils', 'rich._wrap', 'yapf.pytree.blank_line_calculator', 'mmengine.evaluator', 'cv2.gapi.streaming', 'iopath.common.non_blocking_io', 'mmcv.transforms.formatting', 'mmengine.hooks.early_stopping_hook', 'repositories.CMNeXt.semseg.models.cmnext', 'mmcv.cnn.bricks.generalized_attention', 'fvcore.nn.precise_bn', 'mmcv.arraymisc', 'mmengine.dataset.utils', 'rich.cells', 'yapf_third_party._ylib2to3', 'semseg.models.heads.fpn', 'fvcore.nn.parameter_count', 'mmengine.fileio.handlers.pickle_handler', 'mmcv.video', 'mmengine.model.wrappers.utils', 'mmengine.visualization', 'mmcv.cnn.bricks.upsample', 'mmengine.structures', 'torch.distributed.optim.functional_adadelta', 'torch.distributed.optim.named_optimizer', 'rich.region', 'torch.distributed.algorithms.model_averaging', 'mmengine.model.averaged_model', 'rich.measure', 'cv2.segmentation', 'fvcore.nn.flop_count', 'rich._emoji_replace', 'semseg', 'mmcv.cnn.resnet', 'rich.highlighter', 'yapf.pytree.subtype_assigner', 'yapf.pyparser.pyparser', 'mmcv.transforms.loading', 'yapf_third_party._ylib2to3.pgen2.driver', 'mmcv.cnn.bricks.conv_ws', 'importlib_metadata._functools', 'mmengine.hooks.empty_cache_hook', 'tifffile._imagecodecs', 'mmengine.hooks.sampler_seed_hook', 'mmengine.runner.checkpoint', 'addict', 'repositories.CMNeXt.semseg', 'cv2.flann', 'iopath', 'cv2.gapi.ot', 'mmengine.utils.dl_utils.time_counter', 'rich.default_styles', 'mmcv.cnn.utils.flops_counter', 'rich.themes', 'torch.distributed.algorithms.model_averaging.utils', 'fvcore.nn.giou_loss', 'importlib_metadata._text', 'mmcv.cnn.bricks.swish', 'mmengine.fileio.handlers.yaml_handler', 'tifffile.tifffile', 'mmcv.transforms.wrappers', 'iopath.common.event_logger', 'mmengine.registry.registry', 'attr.validators', 'cv2.gapi.own', 'mmengine.evaluator.metric', 'mmengine.model.base_model.data_preprocessor', 'mmengine.registry.root', 'semseg.models.modules.psa', 'attr', 'mmengine.model.efficient_conv_bn_eval', 'semseg.models.layers', 'mmcv.visualization.optflow', 'cv2.gapi.core', 'mmengine.dataset.sampler', 'cv2.fisheye', 'mmengine.optim', 'iopath.tabular.tabular_io', 'rich.spinner', 'mmengine.fileio.backends.local_backend', 'rich.pager', 'mmengine.fileio.handlers.base', 'semseg.datasets.kitti360', 'mmengine.logging.logger', 'cv2.videoio_registry', 'mmengine.dist.utils', 'mmengine', 'mmengine.hooks.profiler_hook', 'mmengine.config.lazy', 'cv2.version', 'semseg.models.layers.common', 'mmengine.fileio.backends.http_backend', 'semseg.models.cmnext', 'mmengine.model.utils', 'mmcv.image.misc', 'repositories.CMNeXt', 'repositories.CMNeXt.semseg.models.cmx', 'yapf_third_party._ylib2to3.pytree', 'mmcv.version', 'rich.live_render', 'cv2.utils', 'mmengine.utils.timer', 'semseg.models.heads.lawin', 'mmengine.model.base_module', 'mmcv.cnn.bricks.conv2d_adaptive_padding', 'yapf.pyparser', 'cv2.gapi', 'mmengine.fileio.handlers', 'attr._cmp', 'rich.repr', 'mmengine._strategy.base', 'mmengine.logging.message_hub', 'semseg.models.heads.hem', 'rich.segment', 'cv2.ipp', 'mmengine.utils.dl_utils.collect_env', 'mmengine.model', 'cv2.load_config_py3', 'attr.setters', 'mmengine.model.test_time_aug', 'rich.scope', 'rich._ratio', 'semseg.models.backbones.cmnext', 'fvcore.nn', 'rich.console', 'mmengine.hooks.checkpoint_hook', 'fvcore.nn.jit_analysis', 'mmcv.cnn.bricks.scale', 'mmcv.cnn.rfsearch.utils', 'yapf.pytree.split_penalty', 'torch.distributed.optim.functional_sgd', 'cv2.barcode', 'mmcv.cnn.bricks.non_local', 'mmcv.cnn.bricks.padding', 'mmengine.hooks.ema_hook', 'rich.terminal_theme', 'mmengine.fileio.file_client', 'cv2.Error', 'mmcv.image.io', 'cv2.gapi.ie', 'mmcv.video.optflow', 'attr._funcs', 'rich.color_triplet', 'mmengine.utils.package_utils', 'rich.screen', 'mmengine.model.wrappers', 'portalocker.portalocker', 'rich._extension', 'importlib_metadata._compat', 'portalocker.utils', 'einops.packing', 'mmcv.video.processing', 'semseg.models.heads.fapn', 'mmengine.fileio.handlers.registry_utils', 'mmengine.version', 'mmengine.optim.scheduler.lr_scheduler', 'mmengine.optim.optimizer.optimizer_wrapper_dict', 'fvcore.common.checkpoint', 'cv2.gapi.ie.detail', 'zipp.glob', 'rich.ansi', 'yapf_third_party._ylib2to3.pygram', 'rich.constrain', 'mmengine.runner._flexible_runner', 'mmcv.image', 'fvcore.nn.weight_init', 'mmengine.evaluator.utils', 'fvcore', 'mmengine.dataset', 'cv2.gapi.onnx', 'yapf.pytree.continuation_splicer', 'mmengine.hooks.logger_hook', 'rich.emoji', 'yapf', 'mmengine._strategy.utils', 'cv2.data', 'mmengine.config', 'mmengine.utils.dl_utils', 'mmcv.cnn.bricks.hswish', 'mmengine.fileio.backends.memcached_backend', 'cv2.gapi.onnx.ep', 'cv2.typing', 'yapf.yapflib.format_decision_state', 'mmengine.utils.dl_utils.torch_ops', 'mmengine.runner.log_processor', 'yapf.yapflib.subtypes', 'semseg.models', 'yapf.yapflib.yapf_api', 'mmengine._strategy.colossalai', 'rich.palette', 'mmengine.utils.dl_utils.hub', 'mmengine.utils', 'mmengine.model.wrappers.fully_sharded_distributed', 'semseg.models.backbones', 'mmcv.cnn.bricks.hsigmoid', 'cv2.ocl', 'mmengine.utils.dl_utils.misc', 'torch.distributed.optim.post_localSGD_optimizer', 'torch.distributed.optim.functional_adagrad', 'mmengine.fileio.backends.base', 'mmcv.image.geometric', 'semseg.augmentations_mm', 'timm.models.layers', 'einops.parsing', 'mmcv.cnn.vgg', 'semseg.models.heads.segformer', 'cv2.detail', 'mmengine.visualization.visualizer', 'mmcv.cnn.bricks.conv', 'termcolor.termcolor', 'mmengine.structures.label_data', 'rich._emoji_codes', 'cv2.misc.version', 'mmengine.utils.dl_utils.trace', 'rich.style', 'mmengine.optim.optimizer.apex_optimizer_wrapper', 'rich.padding', 'mmengine.fileio.handlers.json_handler', 'rich.jupyter', 'mmengine.optim.optimizer', 'torch.distributed.optim.functional_adam', 'mmengine.dist.dist', 'torch.distributed.optim', 'rich.progress', 'mmcv.cnn.rfsearch.operator', 'rich.panel', 'yapf.yapflib.errors', 'yapf_third_party._ylib2to3.pgen2.literals', 'mmengine.logging', 'cv2.dnn', 'yapf.pyparser.split_penalty_visitor', 'mmengine.runner.runner', 'attr._make', 'mmcv.transforms.utils', 'yapf.pytree.pytree_visitor', 'mmengine.utils.progressbar', 'mmcv.cnn.bricks.plugin', 'mmcv.transforms', 'attr.exceptions', 'mmengine.dataset.base_dataset', 'attr._version_info', 'mmengine.logging.history_buffer', 'semseg.models.cmx', 'yapf.yapflib.reformatter', 'torch.distributed.optim.apply_optimizer_in_backward', 'cv2.gapi.ov', 'torch.distributed.algorithms.model_averaging.averagers', 'semseg.models.heads.fcn', 'rich.styled', 'mmengine._strategy.distributed', 'semseg.datasets.mfnet', 'semseg.datasets.deliver', 'mmcv.video.io', 'semseg.utils.utils', 'zipp.compat.py310', 'mmengine.model.base_model.base_model', 'yapf_third_party._ylib2to3.pgen2.token', 'mmengine.registry.utils', 'semseg.models.modules', 'torch.distributed.optim.utils', 'mmengine.config.utils', 'importlib_metadata._collections', 'mmengine.utils.manager', 'iopath.common.file_io', 'tabulate.version', 'semseg.models.heads.condnet', 'mmengine.fileio.parse', 'mmcv.image.colorspace', 'mmengine.structures.pixel_data', 'cv2.ogl', 'yapf.yapflib.style', 'mmengine.utils.progressbar_rich', 'rich._spinners', 'termcolor._types', 'cv2.gapi.render.ocv', 'semseg.models.heads.upernet', 'mmcv.cnn.bricks', 'mmcv.cnn.utils', 'yapf_third_party._ylib2to3.pgen2.grammar', 'rich._null_file', 'mmengine.dataset.dataset_wrapper', 'rich.progress_bar', 'mmengine.optim.scheduler.param_scheduler', 'mmcv.visualization.color', 'mmengine.runner.activation_checkpointing', 'cv2.utils.nested', 'mmengine.hooks.hook', 'mmengine.hooks.test_time_aug_hook', 'mmcv.cnn.bricks.wrappers', 'cv2', 'mmengine.hooks', 'yapf_third_party', 'mmengine.config.config', 'mmcv.cnn.rfsearch.search', 'semseg.datasets', 'yapf.pytree.pytree_unwrapper', 'mmengine.runner.amp', 'fvcore.common', 'torch.distributed.optim.functional_rmsprop', 'semseg.utils', 'mmengine.optim.optimizer.optimizer_wrapper', 'mmcv.cnn.bricks.context_block', 'mmcv.cnn.bricks.conv_module', 'mmengine.fileio.backends.lmdb_backend', 'yapf.yapflib.split_penalty', 'cv2.gapi.render', 'rich.text', 'cv2.aruco', 'portalocker.__about__', 'rich._log_render', 'mmengine.model.base_model', 'zipp.compat', 'mmengine.visualization.utils', 'cv2.gapi.wip.onevpl', 'mmengine.hooks.iter_timer_hook', 'semseg.models.heads', 'importlib_metadata._adapters', 'cv2.gapi.own.detail', 'cv2.gapi.wip.draw', 'cv2.ml', 'mmengine.hooks.param_scheduler_hook', 'importlib_metadata._itertools', 'mmcv.cnn.utils.fuse_conv_bn', 'torch.distributed.optim.optimizer', 'mmcv.transforms.processing', 'rich.abc', 'einops', 'attr._next_gen', 'cv2.gapi.imgproc.fluid', 'mmengine.runner.utils', 'attr._compat', 'mmengine.hooks.sync_buffer_hook', 'importlib_metadata', 'cv2.gapi.video', 'rich.theme', 'yapf.yapflib.logical_line', 'fvcore.nn.jit_handles', 'yapf.yapflib.object_state', 'mmcv.cnn.bricks.activation', 'mmengine.optim.optimizer.default_constructor', 'cv2.gapi.ot.cpu', 'torch.distributed.optim.functional_adamw', 'mmcv.arraymisc.quantization', 'semseg.models.heads.sfnet', 'yapf_third_party._ylib2to3.pgen2', 'attr.filters', 'fvcore.nn.focal_loss', 'rich.control', 'fvcore.nn.activation_count', 'mmcv.cnn.alexnet', 'rich.align', 'semseg.datasets.nyu', 'mmengine.evaluator.evaluator', 'mmengine.fileio.backends.petrel_backend', 'tifffile', 'rich.color', 'mmengine._strategy.single_device', 'rich._pick', 'rich._cell_widths'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/CMNeXt\")\n",
    "from repositories.CMNeXt.semseg.models.cmx import CMX\n",
    "\n",
    "model = decorate_model_for_zip_input(CMX(\"CMX-B2\", 40, [\"img\", \"depth\"]))\n",
    "results[\"nyudv2\"][\"CMX-B2\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(CMX(\"CMX-B4\", 40, [\"img\", \"depth\"]))\n",
    "results[\"nyudv2\"][\"CMX-B4\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(CMX(\"CMX-B5\", 40, [\"img\", \"depth\"]))\n",
    "results[\"nyudv2\"][\"CMX-B5\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(CMX(\"CMX-B2\", 37, [\"img\", \"depth\"]))\n",
    "results[\"sunrgbd\"][\"CMX-B2\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(CMX(\"CMX-B4\", 37, [\"img\", \"depth\"]))\n",
    "results[\"sunrgbd\"][\"CMX-B4\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(CMX(\"CMX-B5\", 37, [\"img\", \"depth\"]))\n",
    "results[\"sunrgbd\"][\"CMX-B5\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Warning] `validation-batch-size` not given, using default: 24.\n",
      "[Warning] Dataset 'nyuv2' does not have a 'valid' split, using 'test' split instead.\n",
      "Running with args:\n",
      " {\n",
      "    \"action\": null,\n",
      "    \"activation\": \"relu\",\n",
      "    \"aug_scale_max\": 1.4,\n",
      "    \"aug_scale_min\": 1.0,\n",
      "    \"batch_size\": 8,\n",
      "    \"cache_dataset\": false,\n",
      "    \"checkpointing_best_only\": false,\n",
      "    \"checkpointing_metrics\": null,\n",
      "    \"checkpointing_skip\": 0.0,\n",
      "    \"context_module\": \"ppm\",\n",
      "    \"dataset\": \"nyuv2\",\n",
      "    \"dataset_path\": null,\n",
      "    \"debug\": false,\n",
      "    \"decoder_normalization\": \"batchnorm\",\n",
      "    \"depth_encoder_backbone\": \"resnet34\",\n",
      "    \"depth_encoder_backbone_block\": null,\n",
      "    \"depth_encoder_backbone_pretrained_weights_filepath\": null,\n",
      "    \"depth_encoder_backbone_resnet_block\": \"nonbottleneck1d\",\n",
      "    \"device\": \"cuda\",\n",
      "    \"disable_progress_bars\": false,\n",
      "    \"dropout_p\": 0.1,\n",
      "    \"enable_panoptic\": false,\n",
      "    \"encoder_decoder_fusion\": null,\n",
      "    \"encoder_decoder_skip_downsamplings\": [\n",
      "        4,\n",
      "        8,\n",
      "        16\n",
      "    ],\n",
      "    \"encoder_fusion\": \"se-add-uni-rgb\",\n",
      "    \"encoder_normalization\": \"batchnorm\",\n",
      "    \"he_init\": [\n",
      "        \"encoder-fusion\"\n",
      "    ],\n",
      "    \"hostname\": \"fk4t54pg31ls1-0\",\n",
      "    \"hypersim_subsample\": 1,\n",
      "    \"hypersim_use_old_depth_stats\": false,\n",
      "    \"input_height\": 480,\n",
      "    \"input_modalities\": [\n",
      "        \"rgb\",\n",
      "        \"depth\"\n",
      "    ],\n",
      "    \"input_width\": 640,\n",
      "    \"instance_center_encoding\": \"sigmoid\",\n",
      "    \"instance_center_heatmap_apply_foreground_mask\": false,\n",
      "    \"instance_center_heatmap_nms_kernel_size\": 17,\n",
      "    \"instance_center_heatmap_threshold\": 0.1,\n",
      "    \"instance_center_heatmap_top_k\": 64,\n",
      "    \"instance_center_loss\": \"mse\",\n",
      "    \"instance_center_sigma\": 8,\n",
      "    \"instance_decoder\": \"emsanet\",\n",
      "    \"instance_decoder_block\": \"nonbottleneck1d\",\n",
      "    \"instance_decoder_block_dropout_p\": 0.2,\n",
      "    \"instance_decoder_downsamplings\": [\n",
      "        16,\n",
      "        8,\n",
      "        4\n",
      "    ],\n",
      "    \"instance_decoder_dropout_p\": 0.1,\n",
      "    \"instance_decoder_n_blocks\": 3,\n",
      "    \"instance_decoder_n_channels\": [\n",
      "        512,\n",
      "        256,\n",
      "        128\n",
      "    ],\n",
      "    \"instance_decoder_upsampling\": \"learned-3x3-zeropad\",\n",
      "    \"instance_encoder_decoder_fusion\": \"add-rgb\",\n",
      "    \"instance_no_multiscale_supervision\": false,\n",
      "    \"instance_offset_distance_threshold\": null,\n",
      "    \"instance_offset_encoding\": \"tanh\",\n",
      "    \"instance_weighting\": [\n",
      "        2,\n",
      "        1\n",
      "    ],\n",
      "    \"is_resumed_training\": false,\n",
      "    \"learning_rate\": 0.01,\n",
      "    \"learning_rate_scheduler\": \"onecycle\",\n",
      "    \"momentum\": 0.9,\n",
      "    \"n_epochs\": 500,\n",
      "    \"n_workers\": 8,\n",
      "    \"no_pretrained_backbone\": false,\n",
      "    \"no_zero_init_decoder_residuals\": false,\n",
      "    \"normal_decoder\": \"emsanet\",\n",
      "    \"normal_decoder_block\": \"nonbottleneck1d\",\n",
      "    \"normal_decoder_block_dropout_p\": 0.2,\n",
      "    \"normal_decoder_downsamplings\": [\n",
      "        16,\n",
      "        8,\n",
      "        4\n",
      "    ],\n",
      "    \"normal_decoder_dropout_p\": 0.1,\n",
      "    \"normal_decoder_n_blocks\": 3,\n",
      "    \"normal_decoder_n_channels\": [\n",
      "        512,\n",
      "        256,\n",
      "        128\n",
      "    ],\n",
      "    \"normal_decoder_upsampling\": \"learned-3x3-zeropad\",\n",
      "    \"normal_encoder_decoder_fusion\": \"add-rgb\",\n",
      "    \"normal_loss\": \"l1\",\n",
      "    \"normal_no_multiscale_supervision\": false,\n",
      "    \"normalization\": null,\n",
      "    \"notes\": \"\",\n",
      "    \"optimizer\": \"sgd\",\n",
      "    \"orientation_kappa\": 1.0,\n",
      "    \"overfit_n_batches\": -1,\n",
      "    \"raw_depth\": false,\n",
      "    \"results_basepath\": \"./results\",\n",
      "    \"resume_ckpt_filepath\": null,\n",
      "    \"resume_ckpt_interval\": 20,\n",
      "    \"resume_path\": null,\n",
      "    \"rgb_encoder_backbone\": \"resnet34\",\n",
      "    \"rgb_encoder_backbone_block\": null,\n",
      "    \"rgb_encoder_backbone_pretrained_weights_filepath\": null,\n",
      "    \"rgb_encoder_backbone_resnet_block\": \"nonbottleneck1d\",\n",
      "    \"rgbd_encoder_backbone\": \"resnet34\",\n",
      "    \"rgbd_encoder_backbone_pretrained_weights_filepath\": null,\n",
      "    \"rgbd_encoder_backbone_resnet_block\": \"nonbottleneck1d\",\n",
      "    \"scannet_semantic_n_classes\": 40,\n",
      "    \"scannet_subsample\": 50,\n",
      "    \"scene_loss_label_smoothing\": 0.1,\n",
      "    \"semantic_class_weighting\": \"median-frequency\",\n",
      "    \"semantic_class_weighting_logarithmic_c\": 1.02,\n",
      "    \"semantic_decoder\": \"emsanet\",\n",
      "    \"semantic_decoder_block\": \"nonbottleneck1d\",\n",
      "    \"semantic_decoder_block_dropout_p\": 0.2,\n",
      "    \"semantic_decoder_downsamplings\": [\n",
      "        16,\n",
      "        8,\n",
      "        4\n",
      "    ],\n",
      "    \"semantic_decoder_dropout_p\": 0.1,\n",
      "    \"semantic_decoder_n_blocks\": 3,\n",
      "    \"semantic_decoder_n_channels\": [\n",
      "        512,\n",
      "        256,\n",
      "        128\n",
      "    ],\n",
      "    \"semantic_decoder_upsampling\": \"learned-3x3-zeropad\",\n",
      "    \"semantic_encoder_decoder_fusion\": \"add-rgb\",\n",
      "    \"semantic_loss_label_smoothing\": 0.0,\n",
      "    \"semantic_no_multiscale_supervision\": false,\n",
      "    \"skip_sanity_check\": false,\n",
      "    \"split\": \"train\",\n",
      "    \"subset_deterministic\": false,\n",
      "    \"subset_train\": 1.0,\n",
      "    \"sunrgbd_depth_do_not_force_mm\": false,\n",
      "    \"sunrgbd_instances_version\": \"panopticndt\",\n",
      "    \"tasks\": [\n",
      "        \"semantic\"\n",
      "    ],\n",
      "    \"tasks_weighting\": [\n",
      "        1\n",
      "    ],\n",
      "    \"upsampling_context_module\": \"bilinear\",\n",
      "    \"upsampling_decoder\": null,\n",
      "    \"upsampling_prediction\": \"learned-3x3-zeropad\",\n",
      "    \"use_original_scene_labels\": false,\n",
      "    \"validation_batch_size\": 24,\n",
      "    \"validation_force_interval\": 20,\n",
      "    \"validation_full_resolution\": false,\n",
      "    \"validation_input_height\": 480,\n",
      "    \"validation_input_width\": 640,\n",
      "    \"validation_only\": false,\n",
      "    \"validation_scannet_benchmark_mode\": false,\n",
      "    \"validation_scannet_subsample\": 100,\n",
      "    \"validation_skip\": 0.0,\n",
      "    \"validation_split\": \"test\",\n",
      "    \"visualization_output_path\": null,\n",
      "    \"visualize_validation\": false,\n",
      "    \"wandb_mode\": \"online\",\n",
      "    \"wandb_name\": null,\n",
      "    \"wandb_project\": \"EMSANet\",\n",
      "    \"weight_decay\": 0.0001,\n",
      "    \"weights_filepath\": null\n",
      "}\n",
      "Loaded NYUv2 dataset without files\n",
      "Loaded NYUv2 dataset without files\n",
      "[Warning] `validation-batch-size` not given, using default: 24.\n",
      "[Warning] Dataset 'nyuv2' does not have a 'valid' split, using 'test' split instead.\n",
      "Running with args:\n",
      " {\n",
      "    \"action\": null,\n",
      "    \"activation\": \"relu\",\n",
      "    \"aug_scale_max\": 1.4,\n",
      "    \"aug_scale_min\": 1.0,\n",
      "    \"batch_size\": 8,\n",
      "    \"cache_dataset\": false,\n",
      "    \"checkpointing_best_only\": false,\n",
      "    \"checkpointing_metrics\": null,\n",
      "    \"checkpointing_skip\": 0.0,\n",
      "    \"context_module\": \"ppm\",\n",
      "    \"dataset\": \"nyuv2\",\n",
      "    \"dataset_path\": null,\n",
      "    \"debug\": false,\n",
      "    \"decoder_normalization\": \"batchnorm\",\n",
      "    \"depth_encoder_backbone\": \"resnet34\",\n",
      "    \"depth_encoder_backbone_block\": null,\n",
      "    \"depth_encoder_backbone_pretrained_weights_filepath\": null,\n",
      "    \"depth_encoder_backbone_resnet_block\": \"nonbottleneck1d\",\n",
      "    \"device\": \"cuda\",\n",
      "    \"disable_progress_bars\": false,\n",
      "    \"dropout_p\": 0.1,\n",
      "    \"enable_panoptic\": false,\n",
      "    \"encoder_decoder_fusion\": null,\n",
      "    \"encoder_decoder_skip_downsamplings\": [\n",
      "        4,\n",
      "        8,\n",
      "        16\n",
      "    ],\n",
      "    \"encoder_fusion\": \"se-add-uni-rgb\",\n",
      "    \"encoder_normalization\": \"batchnorm\",\n",
      "    \"he_init\": [\n",
      "        \"encoder-fusion\"\n",
      "    ],\n",
      "    \"hostname\": \"fk4t54pg31ls1-0\",\n",
      "    \"hypersim_subsample\": 1,\n",
      "    \"hypersim_use_old_depth_stats\": false,\n",
      "    \"input_height\": 480,\n",
      "    \"input_modalities\": [\n",
      "        \"rgb\",\n",
      "        \"depth\"\n",
      "    ],\n",
      "    \"input_width\": 640,\n",
      "    \"instance_center_encoding\": \"sigmoid\",\n",
      "    \"instance_center_heatmap_apply_foreground_mask\": false,\n",
      "    \"instance_center_heatmap_nms_kernel_size\": 17,\n",
      "    \"instance_center_heatmap_threshold\": 0.1,\n",
      "    \"instance_center_heatmap_top_k\": 64,\n",
      "    \"instance_center_loss\": \"mse\",\n",
      "    \"instance_center_sigma\": 8,\n",
      "    \"instance_decoder\": \"emsanet\",\n",
      "    \"instance_decoder_block\": \"nonbottleneck1d\",\n",
      "    \"instance_decoder_block_dropout_p\": 0.2,\n",
      "    \"instance_decoder_downsamplings\": [\n",
      "        16,\n",
      "        8,\n",
      "        4\n",
      "    ],\n",
      "    \"instance_decoder_dropout_p\": 0.1,\n",
      "    \"instance_decoder_n_blocks\": 3,\n",
      "    \"instance_decoder_n_channels\": [\n",
      "        512,\n",
      "        256,\n",
      "        128\n",
      "    ],\n",
      "    \"instance_decoder_upsampling\": \"learned-3x3-zeropad\",\n",
      "    \"instance_encoder_decoder_fusion\": \"add-rgb\",\n",
      "    \"instance_no_multiscale_supervision\": false,\n",
      "    \"instance_offset_distance_threshold\": null,\n",
      "    \"instance_offset_encoding\": \"tanh\",\n",
      "    \"instance_weighting\": [\n",
      "        2,\n",
      "        1\n",
      "    ],\n",
      "    \"is_resumed_training\": false,\n",
      "    \"learning_rate\": 0.01,\n",
      "    \"learning_rate_scheduler\": \"onecycle\",\n",
      "    \"momentum\": 0.9,\n",
      "    \"n_epochs\": 500,\n",
      "    \"n_workers\": 8,\n",
      "    \"no_pretrained_backbone\": false,\n",
      "    \"no_zero_init_decoder_residuals\": false,\n",
      "    \"normal_decoder\": \"emsanet\",\n",
      "    \"normal_decoder_block\": \"nonbottleneck1d\",\n",
      "    \"normal_decoder_block_dropout_p\": 0.2,\n",
      "    \"normal_decoder_downsamplings\": [\n",
      "        16,\n",
      "        8,\n",
      "        4\n",
      "    ],\n",
      "    \"normal_decoder_dropout_p\": 0.1,\n",
      "    \"normal_decoder_n_blocks\": 3,\n",
      "    \"normal_decoder_n_channels\": [\n",
      "        512,\n",
      "        256,\n",
      "        128\n",
      "    ],\n",
      "    \"normal_decoder_upsampling\": \"learned-3x3-zeropad\",\n",
      "    \"normal_encoder_decoder_fusion\": \"add-rgb\",\n",
      "    \"normal_loss\": \"l1\",\n",
      "    \"normal_no_multiscale_supervision\": false,\n",
      "    \"normalization\": null,\n",
      "    \"notes\": \"\",\n",
      "    \"optimizer\": \"sgd\",\n",
      "    \"orientation_kappa\": 1.0,\n",
      "    \"overfit_n_batches\": -1,\n",
      "    \"raw_depth\": false,\n",
      "    \"results_basepath\": \"./results\",\n",
      "    \"resume_ckpt_filepath\": null,\n",
      "    \"resume_ckpt_interval\": 20,\n",
      "    \"resume_path\": null,\n",
      "    \"rgb_encoder_backbone\": \"resnet34\",\n",
      "    \"rgb_encoder_backbone_block\": null,\n",
      "    \"rgb_encoder_backbone_pretrained_weights_filepath\": null,\n",
      "    \"rgb_encoder_backbone_resnet_block\": \"nonbottleneck1d\",\n",
      "    \"rgbd_encoder_backbone\": \"resnet34\",\n",
      "    \"rgbd_encoder_backbone_pretrained_weights_filepath\": null,\n",
      "    \"rgbd_encoder_backbone_resnet_block\": \"nonbottleneck1d\",\n",
      "    \"scannet_semantic_n_classes\": 40,\n",
      "    \"scannet_subsample\": 50,\n",
      "    \"scene_loss_label_smoothing\": 0.1,\n",
      "    \"semantic_class_weighting\": \"median-frequency\",\n",
      "    \"semantic_class_weighting_logarithmic_c\": 1.02,\n",
      "    \"semantic_decoder\": \"emsanet\",\n",
      "    \"semantic_decoder_block\": \"nonbottleneck1d\",\n",
      "    \"semantic_decoder_block_dropout_p\": 0.2,\n",
      "    \"semantic_decoder_downsamplings\": [\n",
      "        16,\n",
      "        8,\n",
      "        4\n",
      "    ],\n",
      "    \"semantic_decoder_dropout_p\": 0.1,\n",
      "    \"semantic_decoder_n_blocks\": 3,\n",
      "    \"semantic_decoder_n_channels\": [\n",
      "        512,\n",
      "        256,\n",
      "        128\n",
      "    ],\n",
      "    \"semantic_decoder_upsampling\": \"learned-3x3-zeropad\",\n",
      "    \"semantic_encoder_decoder_fusion\": \"add-rgb\",\n",
      "    \"semantic_loss_label_smoothing\": 0.0,\n",
      "    \"semantic_no_multiscale_supervision\": false,\n",
      "    \"skip_sanity_check\": false,\n",
      "    \"split\": \"train\",\n",
      "    \"subset_deterministic\": false,\n",
      "    \"subset_train\": 1.0,\n",
      "    \"sunrgbd_depth_do_not_force_mm\": false,\n",
      "    \"sunrgbd_instances_version\": \"panopticndt\",\n",
      "    \"tasks\": [\n",
      "        \"semantic\"\n",
      "    ],\n",
      "    \"tasks_weighting\": [\n",
      "        1\n",
      "    ],\n",
      "    \"upsampling_context_module\": \"bilinear\",\n",
      "    \"upsampling_decoder\": null,\n",
      "    \"upsampling_prediction\": \"learned-3x3-zeropad\",\n",
      "    \"use_original_scene_labels\": false,\n",
      "    \"validation_batch_size\": 24,\n",
      "    \"validation_force_interval\": 20,\n",
      "    \"validation_full_resolution\": false,\n",
      "    \"validation_input_height\": 480,\n",
      "    \"validation_input_width\": 640,\n",
      "    \"validation_only\": false,\n",
      "    \"validation_scannet_benchmark_mode\": false,\n",
      "    \"validation_scannet_subsample\": 100,\n",
      "    \"validation_skip\": 0.0,\n",
      "    \"validation_split\": \"test\",\n",
      "    \"visualization_output_path\": null,\n",
      "    \"visualize_validation\": false,\n",
      "    \"wandb_mode\": \"online\",\n",
      "    \"wandb_name\": null,\n",
      "    \"wandb_project\": \"EMSANet\",\n",
      "    \"weight_decay\": 0.0001,\n",
      "    \"weights_filepath\": null\n",
      "}\n",
      "Loaded SUNRGBD dataset without files\n",
      "Loaded SUNRGBD dataset without files\n",
      "Deleted module: {'scipy.fft._fftlog', 'torchmetrics.functional.classification.hinge', 'scipy.optimize._trustregion_krylov', 'torchmetrics.retrieval.fall_out', 'nicr_mt_scene_analysis.data.preprocessing.rgb', 'torchmetrics.functional.classification.cohen_kappa', 'scipy.stats.morestats', 'scipy.integrate._quadrature', 'cv2.cuda', 'nicr_mt_scene_analysis.data', 'scipy.stats._relative_risk', 'cv2.qt', 'scipy.optimize._lsq.dogbox', 'nicr_mt_scene_analysis.model.decoder.scene', 'scipy.ndimage._filters', 'scipy.spatial._plotutils', 'scipy.spatial.ckdtree', '_csparsetools', 'scipy.optimize._slsqp', 'scipy._lib.array_api_compat.array_api_compat.numpy._aliases', 'nicr_mt_scene_analysis.model.backbone.base', 'torchmetrics.text.wip', 'scipy.sparse.linalg._matfuncs', 'numpy.typing', 'scipy.linalg.flinalg', 'repositories', 'scipy.stats.contingency', 'scipy.optimize._nonlin', 'torchmetrics.functional.classification', 'torchmetrics.utilities.exceptions', 'scipy.stats._stats', 'cv2.gapi.wip.gst', 'torchmetrics.text.bleu', 'scipy.stats._mstats_extras', 'cv2.samples', 'scipy.spatial._geometric_slerp', 'nicr_mt_scene_analysis', 'torchmetrics.functional.classification.hamming', 'scipy.optimize._linprog_highs', 'scipy.optimize._highs', 'torchmetrics.functional.regression.spearman', 'scipy.sparse.sparsetools', 'scipy.optimize._shgo_lib._complex', 'scipy.sparse.csgraph._traversal', 'scipy.linalg._decomp_lu', 'torchmetrics.functional.text.ter', 'scipy.stats.distributions', 'scipy.integrate.odepack', 'scipy.optimize.lbfgsb', 'torchmetrics.functional.retrieval.ndcg', 'scipy.optimize._shgo', 'scipy.sparse.linalg.interface', 'scipy', 'scipy.special._ellip_harm_2', 'torchmetrics.text', 'torchmetrics.functional.regression.pearson', 'torchmetrics.functional.regression.kl_divergence', 'torchmetrics.classification.exact_match', 'scipy._lib._uarray._uarray', 'scipy.optimize._highs._highs_wrapper', 'scipy.optimize.linesearch', 'nicr_mt_scene_analysis.model.decoder.mlp_base', 'repositories.EMSANet.emsanet.args', 'transformers.pytorch_utils', 'packaging.tags', 'cv2.gapi.core.fluid', 'nicr_mt_scene_analysis.data.preprocessing.utils', 'scipy.integrate._bvp', 'scipy.sparse.construct', 'torchmetrics.functional.text.bert', 'torchmetrics.utilities.compute', 'torchmetrics.functional.classification.calibration_error', 'scipy.special._orthogonal', 'nicr_scene_analysis_datasets.datasets.scenenetrgbd', 'scipy.stats._binned_statistic', 'cityscapesscripts.helpers.labels', 'torchmetrics.regression.r2', 'nicr_mt_scene_analysis.model.decoder.instance', 'nicr_mt_scene_analysis.model.block', 'scipy.optimize._tnc', 'scipy.sparse.linalg._onenormest', 'scipy.sparse._sputils', 'torchmetrics.classification.stat_scores', 'torchmetrics.functional.image.uqi', 'nicr_mt_scene_analysis.model.postprocessing.normal', 'scipy.optimize._direct_py', 'scipy.stats._hypotests', 'scipy.optimize._direct', 'nicr_mt_scene_analysis.model.backbone', 'torchmetrics.regression', 'scipy.sparse._construct', 'nicr_mt_scene_analysis.model.postprocessing.base', 'scipy.interpolate._rbf', 'cv2.gapi.wip', 'scipy.stats._stats_mstats_common', 'numpy.testing._private', 'scipy.stats._odds_ratio', 'torchmetrics.text.infolm', 'scipy.linalg.blas', 'nicr_mt_scene_analysis.multi_task', 'scipy.fft._basic', 'cv2.misc', 'scipy.stats.mstats', 'torchmetrics.regression.wmape', 'scipy.optimize._linprog_simplex', 'transformers.generation.stopping_criteria', 'scipy.stats._biasedurn', 'torchmetrics.regression.explained_variance', 'scipy.optimize._numdiff', 'torchmetrics.wrappers', 'scipy.sparse.linalg._eigen._svds', 'scipy.sparse._base', 'scipy.stats._distr_params', 'scipy.optimize._optimize', 'scipy.stats._mannwhitneyu', 'scipy.sparse.dia', 'scipy.stats._unuran', 'scipy.integrate._ivp.bdf', 'scipy.sparse.linalg._isolve', 'scipy.sparse.dok', 'nicr_mt_scene_analysis.model.decoder.dense_base', 'transformers.generation.beam_constraints', '_moduleTNC', 'scipy.special.specfun', 'nicr_scene_analysis_datasets.version', 'torchmetrics.retrieval.base', 'scipy._lib.array_api_compat.array_api_compat._internal', 'scipy.linalg._flinalg', 'scipy.sparse._lil', 'nicr_scene_analysis_datasets.datasets.cityscapes.cityscapes', 'packaging._musllinux', 'nicr_scene_analysis_datasets.datasets.hypersim.hypersim', 'scipy.sparse.linalg.isolve', 'pkg_resources._vendor.more_itertools.more', 'torchmetrics.utilities.checks', 'torchmetrics.functional.pairwise.helpers', 'nicr_mt_scene_analysis.utils._normal', 'torchmetrics.text.mer', 'scipy.optimize._remove_redundancy', 'cv2.parallel', 'nicr_mt_scene_analysis.model.backbone.resnet', 'scipy._lib.array_api_compat.array_api_compat.numpy.linalg', 'nicr_scene_analysis_datasets.datasets.coco', 'torchmetrics.audio.pit', 'torchmetrics.text.ter', 'scipy.spatial.qhull', 'transformers.cache_utils', 'torchmetrics.classification.recall_at_fixed_precision', 'torchmetrics.classification.hinge', 'nicr_mt_scene_analysis.data.preprocessing.panoptic', 'transformers.quantizers.quantizer_quanto', 'transformers.integrations.peft', 'cv2.utils.fs', 'scipy.stats._boost.beta_ufunc', 'cv2.mat_wrapper', 'torchmetrics.wrappers.tracker', 'scipy._lib._bunch', 'scipy.sparse.linalg._isolve.utils', 'pkg_resources._vendor.packaging._elffile', 'torchmetrics.functional.retrieval.recall', 'scipy.sparse.linalg.eigen', 'nicr_mt_scene_analysis.model.initialization', 'nicr_mt_scene_analysis.metric.pq', 'pkg_resources._vendor.more_itertools.recipes', 'repositories.EMSANet.emsanet', 'scipy.optimize._minpack_py', 'scipy.sparse.csgraph._validation', 'torchmetrics.audio.snr', 'nicr_mt_scene_analysis.task_helper.panoptic', 'torchmetrics', 'scipy.stats._sobol', 'scipy.integrate._ivp.radau', 'numpy.testing._private.extbuild', 'termcolor', 'scipy.stats._variation', 'cv2.gapi.imgproc', 'nicr_mt_scene_analysis.data.preprocessing.crop', 'scipy.integrate._vode', 'scipy.linalg.misc', 'torchmetrics.functional.classification.exact_match', 'scipy.interpolate._ppoly', 'scipy.linalg._decomp', 'scipy.optimize._cobyla', 'torchmetrics.functional.pairwise.cosine', 'nicr_mt_scene_analysis.loss.mse', 'scipy.stats._boost.skewnorm_ufunc', 'scipy.constants._constants', 'transformers.generation.utils', 'torchmetrics.functional.text', 'torchmetrics.classification.hamming', 'nicr_mt_scene_analysis.visualization.generic', 'scipy.optimize._root_scalar', 'scipy.sparse._bsr', 'scipy.interpolate', 'repositories.EMSANet.emsanet.optimizer', 'scipy.integrate._quadpack_py', 'scipy.optimize._lsq.givens_elimination', 'scipy.stats.biasedurn', 'scipy.special._ellip_harm', 'nicr_mt_scene_analysis.model.encoder_fusion', 'torchmetrics.functional.classification.average_precision', 'scipy.optimize.nonlin', 'nicr_scene_analysis_datasets.datasets.sunrgbd.dataset', 'torchmetrics.functional.audio.snr', 'scipy.stats._boost.nct_ufunc', 'torchmetrics.text.sacre_bleu', 'pkg_resources.extern.jaraco', 'torchmetrics.wrappers.multioutput', 'transformers.safetensors_conversion', 'scipy.optimize._hessian_update_strategy', 'scipy.spatial._spherical_voronoi', 'torchmetrics.functional.classification.precision_recall_curve', 'scipy.fft._pocketfft.helper', 'cv2.gapi.core.cpu', 'scipy.sparse._matrix_io', 'nicr_scene_analysis_datasets.datasets.cityscapes.dataset', 'nicr_scene_analysis_datasets.dataset_base._rgbd_dataset', 'scipy.linalg.special_matrices', 'scipy.stats.mstats_extras', 'scipy.optimize._lsq.least_squares', 'scipy.linalg.decomp_schur', 'scipy.sparse.linalg._eigen', 'scipy.interpolate._fitpack', 'cv2.gapi.core.ocl', 'nicr_mt_scene_analysis.data.preprocessing', 'torchmetrics.functional.classification.confusion_matrix', 'torchmetrics.functional.text.chrf', 'scipy._lib._ccallback_c', 'scipy.linalg._decomp_lu_cython', 'pkg_resources._vendor.packaging._musllinux', 'torchmetrics.functional.retrieval.precision', 'nicr_mt_scene_analysis.model.context_module', 'nicr_mt_scene_analysis.model.normalization', 'scipy.sparse._dok', 'scipy.stats.ncx2_ufunc', 'scipy.linalg._decomp_ldl', 'cv2.gapi.oak', 'scipy.optimize._differentialevolution', 'scipy.interpolate._fitpack2', 'scipy.stats.skewnorm_ufunc', 'scipy.stats._multicomp', 'torchmetrics.classification.calibration_error', 'torchmetrics.classification.specificity', 'scipy.stats._resampling', 'scipy.interpolate.rbf', 'scipy.optimize._highs._highs_constants', 'scipy.stats._covariance', 'torchmetrics.functional.audio.sdr', 'scipy.stats._discrete_distns', 'torchmetrics.regression.kl_divergence', 'nicr_mt_scene_analysis.utils._misc', 'scipy._distributor_init', 'scipy.stats.sampling', 'torchmetrics.classification', 'cv2.gapi.streaming', 'scipy.sparse.linalg.dsolve', 'scipy.optimize._linprog_util', 'scipy.interpolate._pade', 'torchmetrics.image.sam', 'torchmetrics.functional.text.mer', 'nicr_scene_analysis_datasets.dataset_base._rgb_dataset', 'torchmetrics.text.wil', 'pkg_resources.extern.packaging.version', 'torchmetrics.functional.classification.f_beta', 'regex._regex', 'scipy.stats.nct_ufunc', 'scipy.optimize._lsq.common', 'transformers.generation', 'transformers.utils.quantization_config', 'scipy._lib.deprecation', 'torchmetrics.functional.audio', 'torchmetrics.metric', 'scipy.linalg._matfuncs', 'scipy._lib._threadsafety', 'scipy.sparse.compressed', 'scipy.special._basic', 'nicr_scene_analysis_datasets.datasets.coco.coco', 'scipy.stats._morestats', 'scipy.sparse.extract', 'scipy.linalg._interpolative_backend', 'numpy.testing.overrides', 'scipy.optimize.cobyla', 'scipy.sparse.linalg.matfuncs', 'scipy.sparse.csc', 'scipy.optimize._minimize', 'cityscapesscripts.helpers', 'torchmetrics.functional.pairwise.euclidean', 'scipy.sparse.linalg._isolve.lsqr', 'nicr_scene_analysis_datasets.datasets.nyuv2', 'torchmetrics.functional.image.gradients', 'scipy.__config__', 'torchmetrics.functional.classification.roc', 'scipy.linalg._decomp_polar', 'scipy.optimize._trustregion_constr.projections', 'scipy.integrate.dop', 'cv2.segmentation', 'nicr_scene_analysis_datasets.utils._colormaps', 'torchmetrics.functional.image', 'torchmetrics.image', 'torchmetrics.retrieval.r_precision', 'nicr_mt_scene_analysis.model.upsampling', 'scipy.sparse._extract', 'torchmetrics.regression.spearman', 'scipy.spatial', 'torchmetrics.functional.regression.cosine_similarity', 'scipy.interpolate._bsplines', 'scipy.stats._binomtest', 'scipy.optimize._lsq.trf_linear', 'scipy._lib.messagestream', 'scipy.integrate._ode', 'scipy.spatial.transform._rotation', 'torchmetrics.regression.symmetric_mape', 'scipy.ndimage._ni_label', 'torchmetrics.audio.sdr', 'scipy.optimize._trustregion_constr', 'scipy.stats._page_trend_test', 'scipy.optimize._lsq', 'torchmetrics.regression.cosine_similarity', 'nicr_mt_scene_analysis.utils._printing', 'scipy.stats._warnings_errors', 'torchmetrics.classification.f_beta', 'torchmetrics.collections', 'scipy.version', 'torchmetrics.functional.classification.auc', 'nicr_mt_scene_analysis.model.decoder.normal', 'torchmetrics.functional.text.squad', 'scipy.optimize.minpack2', 'scipy.interpolate._interpolate', 'scipy.sparse.sputils', 'scipy.sparse.lil', 'scipy.spatial.transform', 'torchmetrics.functional.retrieval.r_precision', 'scipy.linalg._procrustes', 'scipy.special.basic', 'torchmetrics.functional.retrieval.fall_out', 'torchmetrics.functional.image.helper', 'nicr_scene_analysis_datasets.datasets.nyuv2.dataset', 'scipy.special.spfun_stats', 'scipy.optimize._trustregion_constr.minimize_trustregion_constr', 'nicr_mt_scene_analysis.model.encoder', 'scipy.linalg._flapack', 'scipy.fft._pocketfft.pypocketfft', 'scipy.ndimage.filters', 'scipy.stats._boost', 'scipy.stats._levy_stable', 'cv2.flann', 'pkg_resources.extern.jaraco.context', 'cv2.gapi.ot', 'scipy.spatial._distance_pybind', 'scipy._lib._uarray._backend', 'scipy.interpolate.fitpack', 'torchmetrics.functional.audio.pit', 'nicr_mt_scene_analysis.visualization._pil', 'scipy.stats._crosstab', 'scipy._lib.array_api_compat.array_api_compat', 'scipy.stats.mstats_basic', 'scipy.fft._helper', 'nicr_scene_analysis_datasets.datasets', 'scipy.sparse.linalg._eigen.arpack._arpack', 'scipy.stats.stats', 'torchmetrics.functional.classification.stat_scores', 'scipy.stats._unuran.unuran_wrapper', 'scipy.linalg.interpolative', 'scipy.optimize._lbfgsb_py', 'nicr_mt_scene_analysis.data.preprocessing.orientation', 'scipy.linalg.basic', 'scipy.fft._realtransforms_backend', 'nicr_mt_scene_analysis.task_helper.semantic', 'scipy.linalg._decomp_cossin', 'nicr_mt_scene_analysis.data.preprocessing.normalize', 'scipy.fft._pocketfft', 'scipy.spatial._ckdtree', 'cv2.gapi.own', 'transformers.generation.configuration_utils', 'scipy._lib.array_api_compat.array_api_compat.common._linalg', 'torchmetrics.functional.classification.jaccard', 'scipy.stats._entropy', 'scipy.sparse._csc', 'nicr_mt_scene_analysis.utils._torch', 'scipy.stats._boost.invgauss_ufunc', 'scipy.stats', 'scipy.interpolate.interpnd', 'torchmetrics.retrieval', 'cv2.gapi.core', 'torchmetrics.text.squad', 'scipy.sparse.linalg._isolve.lsmr', 'cv2.fisheye', 'scipy.stats.beta_ufunc', 'repositories.EMSANet.emsanet.data', 'scipy.stats._rvs_sampling', 'scipy.optimize._bglu_dense', 'scipy.stats._rcont', 'scipy.optimize._trustregion_constr.report', 'nicr_mt_scene_analysis.model.postprocessing', 'cv2.videoio_registry', 'scipy.sparse.csgraph._laplacian', 'scipy.interpolate._rgi', 'torchmetrics.utilities.data', 'scipy.stats._boost.hypergeom_ufunc', 'nicr_mt_scene_analysis.utils.panoptic_merge', 'torchmetrics.image.d_lambda', 'scipy.fft._backend', 'scipy.constants', 'torchmetrics.functional.text.sacre_bleu', 'torchmetrics.classification.cohen_kappa', 'cv2.version', 'scipy.stats._sampling', 'scipy.special.add_newdocs', 'nicr_mt_scene_analysis.metric.rmse', 'scipy.optimize._root', 'torchmetrics.functional.classification.recall_at_fixed_precision', 'scipy.spatial._voronoi', 'scipy.stats._boost.ncx2_ufunc', 'torchmetrics.aggregation', 'scipy.integrate.vode', 'pkg_resources._vendor.packaging', 'scipy.stats._distn_infrastructure', 'nicr_mt_scene_analysis.data.preprocessing.semantic', 'torchmetrics.regression.concordance', 'torchmetrics.utilities.enums', 'torchmetrics.functional.text.rouge', 'scipy._lib.decorator', 'scipy.optimize._trlib', 'cv2.utils', 'nicr_scene_analysis_datasets.datasets.scannet.scannet', 'scipy.spatial.transform._rotation_spline', 'scipy.stats._boost.nbinom_ufunc', 'nicr_scene_analysis_datasets.datasets.sunrgbd.sunrgbd', 'scipy.linalg._decomp_schur', 'scipy.stats._common', 'transformers.quantizers.quantizer_awq', 'pkg_resources._vendor.more_itertools', 'torchmetrics.functional.pairwise', 'cv2.gapi', 'pkg_resources._vendor', 'scipy.optimize._trustregion_dogleg', 'scipy.linalg._decomp_update', 'transformers.activations', 'cityscapesscripts', 'torchmetrics.functional.classification.precision_recall', 'scipy.optimize._linprog_rs', 'scipy.optimize._milp', 'scipy.interpolate.interpolate', 'transformers.quantizers.quantizer_bnb_8bit', 'scipy._lib._util', 'scipy.linalg._decomp_qr', 'transformers.quantizers', 'plistlib', 'torchmetrics.functional.text.perplexity', 'pkg_resources._vendor.platformdirs.unix', 'scipy.sparse._csparsetools', 'cv2.ipp', 'cv2.load_config_py3', 'scipy.optimize._dcsrch', 'scipy.linalg.decomp_qr', 'scipy._lib.array_api_compat.array_api_compat.common', 'scipy.integrate._quad_vec', 'torchmetrics.regression.mape', 'scipy.stats._qmvnt', 'repositories.EMSANet.emsanet.model', 'scipy.sparse.linalg._interface', 'scipy.integrate._odepack', 'torchmetrics.regression.pearson', 'torchmetrics.functional.classification.matthews_corrcoef', 'torchmetrics.classification.precision_recall', 'scipy.optimize._trustregion_exact', 'transformers.quantizers.quantizer_aqlm', 'cv2.barcode', 'scipy.sparse.bsr', 'scipy.interpolate._ndgriddata', 'pkg_resources.extern.jaraco.text', 'nicr_mt_scene_analysis.data.preprocessing.torch', 'numpy._typing._add_docstring', 'cv2.Error', 'torchmetrics.functional.regression.wmape', 'pkg_resources._vendor.packaging._manylinux', 'scipy.sparse.linalg._dsolve.linsolve', 'cv2.gapi.ie', 'scipy.stats._boost.binom_ufunc', 'scipy.stats._levy_stable.levyst', 'scipy.sparse.linalg._isolve.minres', 'torchmetrics.functional.regression.mse', 'scipy.integrate._ivp.common', 'scipy.linalg._decomp_qz', 'scipy.ndimage._nd_image', 'torchmetrics.functional.text.infolm', 'scipy.sparse.csgraph._flow', 'scipy.special', 'scipy.sparse.base', 'torchmetrics.functional.text.cer', 'torchmetrics.wrappers.minmax', 'scipy.optimize._lsq.trf', 'scipy.interpolate.polyint', 'repositories.EMSANet.emsanet.decoder', 'nicr_mt_scene_analysis.model.activation', 'scipy.optimize._lsap', 'scipy.interpolate._fitpack_impl', 'scipy.sparse.linalg._special_sparse_arrays', 'nicr_mt_scene_analysis.model.context_module.appm', 'scipy.integrate._ivp.base', 'scipy.sparse.linalg._dsolve', 'scipy.optimize._trlib._trlib', 'nicr_mt_scene_analysis.metric', 'scipy.special._logsumexp', 'nicr_mt_scene_analysis.task_helper.normal', 'torchmetrics.functional.text.helper_embedding_metric', 'torchmetrics.functional.image.psnr', 'torchmetrics.functional.regression.symmetric_mape', 'scipy._lib._array_api', 'scipy.stats._bws_test', 'scipy.optimize._trustregion_constr.equality_constrained_sqp', 'scipy.sparse._csr', 'scipy.stats._ksstats', 'cv2.gapi.ie.detail', 'scipy.fft', 'scipy.ndimage', 'torchmetrics.text.eed', 'nicr_scene_analysis_datasets.dataset_base._base_dataset', 'torchmetrics.text.perplexity', 'scipy.sparse._data', 'scipy.sparse._coo', 'repositories.EMSANet', 'torchmetrics.functional.text.eed', 'nicr_scene_analysis_datasets.datasets.scannet.dataset', 'nicr_scene_analysis_datasets.datasets.scannet.scannet200_constants', 'scipy.spatial.kdtree', 'nicr_mt_scene_analysis.visualization.normal', 'scipy.sparse.csr', 'torchmetrics.image.ergas', 'scipy.special.cython_special', 'nicr_mt_scene_analysis.metric.miou', 'scipy.stats._sensitivity_analysis', 'torchmetrics.functional.classification.specificity', 'torchmetrics.classification.auc', 'scipy.sparse.csgraph._reordering', 'scipy._lib.array_api_compat.array_api_compat.common._helpers', 'torchmetrics.functional.classification.auroc', 'scipy.sparse._matrix', 'nicr_mt_scene_analysis.loss.vonmises', 'pkg_resources._vendor.platformdirs.version', 'torchmetrics.retrieval.precision_recall_curve', 'nicr_mt_scene_analysis.model.backbone.swin', 'scipy.linalg._expm_frechet', 'transformers.modeling_utils', 'cv2.gapi.onnx', 'nicr_scene_analysis_datasets.dataset_base._concat_dataset', 'scipy.sparse.linalg._eigen.arpack', 'nicr_scene_analysis_datasets.datasets.scannet', 'scipy.optimize._constraints', 'transformers.quantizers.quantizers_utils', 'scipy.stats._mvn', 'torchmetrics.utilities', 'torchmetrics.classification.confusion_matrix', 'scipy.stats._axis_nan_policy', 'scipy.integrate._ivp.lsoda', 'scipy.optimize._shgo_lib', 'scipy._lib._docscrape', 'scipy.optimize._nnls', 'pkg_resources.extern.packaging', 'nicr_mt_scene_analysis.model.postprocessing.instance', 'scipy.linalg.decomp_cholesky', 'nicr_scene_analysis_datasets.utils.img', 'cv2.data', 'scipy.stats.binom_ufunc', 'torchmetrics.functional.regression', 'nicr_scene_analysis_datasets.utils', 'torchmetrics.classification.auroc', 'scipy.sparse.linalg._isolve._gcrotmk', 'scipy.optimize._differentiable_functions', 'scipy.linalg._fblas', 'scipy.linalg._misc', 'torchmetrics.classification.roc', 'nicr_mt_scene_analysis.model.postprocessing.semantic', 'torchmetrics.functional.image.d_lambda', 'cv2.gapi.onnx.ep', 'nicr_mt_scene_analysis.model.decoder.semantic', 'scipy.spatial._distance_wrap', 'scipy.stats._survival', 'cv2.typing', 'scipy.spatial._hausdorff', 'nicr_mt_scene_analysis.task_helper.scene', 'nicr_mt_scene_analysis.model.postprocessing.panoptic', 'scipy.stats.invgauss_ufunc', 'scipy.sparse.linalg._norm', 'scipy.ndimage._interpolation', 'scipy.ndimage._ni_docstrings', 'nicr_mt_scene_analysis.utils._orientation', 'scipy.interpolate.ndgriddata', 'scipy.special._spherical_bessel', 'scipy.linalg._interpolative', 'scipy.sparse._compressed', 'torchmetrics.functional.retrieval.hit_rate', 'nicr_mt_scene_analysis.types', 'scipy.integrate._ivp.dop853_coefficients', 'scipy.sparse.data', 'scipy.special._sf_error', 'scipy.special._ufuncs', 'scipy.optimize.minpack', 'torchmetrics.functional.text.wil', 'transformers.quantizers.base', 'scipy.special._lambertw', 'scipy.ndimage._fourier', 'scipy.interpolate._rbfinterp', 'scipy.special.sf_error', 'torchmetrics.__about__', 'scipy.optimize._basinhopping', 'torchmetrics.retrieval.recall', 'cv2.ocl', 'scipy.optimize._trustregion_ncg', 'transformers.generation.logits_process', 'scipy.linalg.cython_blas', 'scipy._lib._finite_differences', 'scipy.fft._pocketfft.basic', 'scipy.spatial._procrustes', 'pkg_resources.extern.packaging.markers', 'scipy.optimize._qap', 'torchmetrics.classification.jaccard', 'scipy.ndimage.morphology', 'torchmetrics.functional.classification.ranking', 'scipy.stats._qmc_cy', 'scipy.ndimage._morphology', 'nicr_mt_scene_analysis.utils', 'scipy.fft._pocketfft.realtransforms', 'nicr_scene_analysis_datasets.datasets.scenenetrgbd.scenenetrgbd', 'cv2.detail', 'scipy.optimize._cobyla_py', 'termcolor.termcolor', 'nicr_scene_analysis_datasets.utils.io', 'scipy.sparse.linalg._isolve.tfqmr', 'scipy.optimize._highs.cython.src._highs_wrapper', 'scipy.linalg', 'scipy.optimize._shgo_lib._vertex', 'repositories.EMSANet.emsanet.lr_scheduler', 'transformers.generation.candidate_generator', 'torchmetrics.classification.average_precision', 'nicr_mt_scene_analysis.model.postprocessing.scene', 'nicr_scene_analysis_datasets.datasets.nyuv2.nyuv2', 'torchmetrics.functional.pairwise.manhattan', 'torchmetrics.functional.image.ergas', 'cv2.misc.version', 'scipy.sparse._dia', 'torchmetrics.text.wer', 'nicr_mt_scene_analysis.loss.ce', 'torchmetrics.functional.classification.dice', 'nicr_scene_analysis_datasets.dataset_base._class_weighting', 'nicr_mt_scene_analysis.model.backbone.swin_multimodal', 'scipy.spatial.distance', 'torchmetrics.audio', 'scipy.interpolate._bspl', 'scipy.sparse.linalg._isolve.iterative', 'pkg_resources.extern.more_itertools', 'scipy._lib.doccer', 'torchmetrics.classification.ranking', 'packaging._elffile', 'torchmetrics.retrieval.reciprocal_rank', 'nicr_mt_scene_analysis.metric.mae', 'pkg_resources._vendor.jaraco', 'nicr_mt_scene_analysis.model.decoder.base', 'scipy.special.orthogonal', 'torchmetrics.functional.classification.accuracy', 'numpy.testing', 'torchmetrics.regression.tweedie_deviance', 'torchmetrics.regression.log_mse', 'pkg_resources._vendor.platformdirs.api', 'regex', 'scipy.spatial.transform.rotation', 'scipy.stats._mstats_basic', 'cv2.dnn', 'scipy.linalg._sketches', 'scipy.interpolate._fitpack_py', 'torchmetrics.functional', 'nicr_scene_analysis_datasets.datasets.cityscapes', 'torchmetrics.wrappers.classwise', 'pkg_resources.extern.packaging._parser', 'torchmetrics.functional.retrieval.average_precision', 'nicr_mt_scene_analysis.model.decoder.panoptic', 'scipy.integrate._odepack_py', 'scipy.linalg._solvers', 'scipy.fft._realtransforms', 'scipy.constants.codata', 'nicr_mt_scene_analysis.model', 'scipy.optimize', 'scipy.optimize.moduleTNC', 'scipy.optimize._dual_annealing', 'scipy._lib.uarray', 'numpy.testing._private.utils', 'pkg_resources.extern.packaging.requirements', 'scipy.linalg._solve_toeplitz', 'nicr_scene_analysis_datasets.pytorch', 'nicr_scene_analysis_datasets.datasets.hypersim.dataset', 'torchmetrics.functional.pairwise.linear', 'scipy.optimize._group_columns', 'nicr_mt_scene_analysis.visualization.depth', 'scipy.ndimage.measurements', 'scipy.integrate._ivp.ivp', 'scipy.stats.qmc', 'nicr_mt_scene_analysis.task_helper.instance', 'torchmetrics.classification.matthews_corrcoef', 'pkg_resources.extern.packaging._structures', 'scipy.interpolate._polyint', 'transformers.quantizers.quantizer_bnb_4bit', 'scipy.spatial._kdtree', 'torchmetrics.classification.binned_precision_recall', 'scipy.integrate._quadpack', 'cv2.gapi.ov', 'packaging._manylinux', 'scipy.stats.nbinom_ufunc', 'nicr_mt_scene_analysis.version', 'scipy.stats._boost.ncf_ufunc', 'scipy.stats._stats_pythran', 'scipy.stats.hypergeom_ufunc', 'scipy._lib.array_api_compat.array_api_compat.common._aliases', 'scipy.interpolate._rgi_cython', 'scipy.optimize._pava_pybind', 'scipy.integrate._dop', 'torchmetrics.functional.regression.log_mse', 'torchmetrics.retrieval.precision', 'torchmetrics.functional.regression.tweedie_deviance', 'scipy.optimize._linprog_ip', 'scipy.stats._fit', 'torchmetrics.functional.regression.mape', 'nicr_mt_scene_analysis.model.context_module.ppm', 'scipy.interpolate._ndbspline', 'nicr_scene_analysis_datasets.dataset_base', 'nicr_scene_analysis_datasets', 'scipy.optimize._zeros', 'scipy.stats._censored_data', 'pkg_resources.extern.packaging._tokenizer', 'scipy.constants.constants', 'torchmetrics.functional.retrieval', 'scipy._lib._pep440', 'scipy.fft._fftlog_backend', 'scipy.interpolate.fitpack2', 'nicr_mt_scene_analysis.visualization.panoptic', 'nicr_mt_scene_analysis.loss.l1', 'nicr_scene_analysis_datasets.dataset_base._depth_dataset', 'scipy.optimize.optimize', 'scipy.integrate._ivp', 'scipy.optimize._spectral', 'scipy.optimize._trustregion_constr.tr_interior_point', 'scipy.constants._codata', 'scipy.stats._ansari_swilk_statistics', 'scipy.sparse.linalg._eigen.lobpcg', 'scipy.integrate._ivp.rk', 'scipy.sparse._index', '_ni_label', 'scipy.sparse.linalg._isolve.lgmres', 'nicr_mt_scene_analysis.model.encoder_decoder_fusion', 'scipy.special._ufuncs_cxx', 'scipy.linalg.lapack', 'cv2.ogl', 'scipy.ndimage._ni_support', 'nicr_mt_scene_analysis.data.preprocessing.flip', 'scipy.linalg._matfuncs_sqrtm', 'termcolor._types', 'nicr_mt_scene_analysis.model.utils', 'cv2.gapi.render.ocv', 'scipy.optimize._minpack2', 'torchmetrics.image.psnr', 'scipy.ndimage.interpolation', 'torchmetrics.functional.image.sam', 'torchmetrics.functional.regression.mae', 'scipy.stats.kde', 'torchmetrics.functional.retrieval.precision_recall_curve', 'nicr_mt_scene_analysis.model.decoder', 'scipy.optimize._trustregion_constr.canonical_constraint', 'nicr_mt_scene_analysis.visualization.instance', 'scipy.optimize._isotonic', 'scipy.optimize.slsqp', 'scipy.ndimage._measurements', 'scipy.stats._kde', 'scipy.sparse', 'nicr_mt_scene_analysis.task_helper', 'torchmetrics.classification.precision_recall_curve', 'scipy.sparse.csgraph._tools', 'nicr_mt_scene_analysis.data._dataloader', 'scipy.integrate.lsoda', 'torchmetrics.functional.image.ssim', 'scipy.optimize._lsq.bvls', 'regex.regex', 'scipy.linalg._cythonized_array_utils', 'torchmetrics.text.bert', 'torchmetrics.functional.text.helper', 'scipy.sparse.linalg._eigen.lobpcg.lobpcg', 'pkg_resources.extern.packaging.specifiers', 'scipy.linalg._decomp_cholesky', 'scipy.optimize._minpack', 'cv2.utils.nested', 'transformers.integrations.aqlm', 'scipy.sparse.csgraph._shortest_path', 'scipy._lib.array_api_compat', 'torchmetrics.functional.text.wer', 'torchmetrics.functional.regression.concordance', 'pkg_resources._vendor.platformdirs', 'scipy._lib._ccallback', 'transformers.quantizers.auto', 'cv2', 'scipy.sparse.coo', 'scipy.spatial._qhull', 'torchmetrics.utilities.imports', 'transformers.quantizers.quantizer_gptq', 'scipy.sparse.linalg._dsolve._superlu', 'torchmetrics.functional.retrieval.reciprocal_rank', 'nicr_scene_analysis_datasets.dataset_base._config', 'nicr_mt_scene_analysis.data.preprocessing.torch_transform_wrapper', 'nicr_mt_scene_analysis.model.context_module.none', 'nicr_mt_scene_analysis.task_helper.base', 'scipy.stats._qmc', 'scipy.linalg.decomp_svd', 'scipy.linalg.decomp', 'torchmetrics.classification.accuracy', 'scipy.sparse.linalg._eigen.arpack.arpack', 'scipy.optimize._linesearch', 'scipy.optimize._linprog', 'scipy.stats._tukeylambda_stats', 'nicr_mt_scene_analysis.loss', 'scipy.stats.mvn', 'scipy.linalg.cython_lapack', 'scipy.spatial.transform._rotation_groups', 'nicr_scene_analysis_datasets.datasets.hypersim', 'torchmetrics.functional.regression.r2', 'scipy.stats._stats_py', 'pkg_resources.extern.platformdirs', 'scipy.sparse.csgraph._min_spanning_tree', 'nicr_mt_scene_analysis.loss.base', 'torchmetrics.image.uqi', 'scipy.sparse.linalg._expm_multiply', 'cv2.gapi.render', 'torchmetrics.classification.dice', 'cv2.aruco', 'nicr_scene_analysis_datasets.dataset_base._annotation', 'scipy.optimize._trustregion', 'scipy.sparse.linalg._dsolve._add_newdocs', 'pkg_resources.extern.packaging.utils', 'scipy.linalg._special_matrices', 'scipy.special._support_alternative_backends', 'cv2.gapi.wip.onevpl', 'nicr_mt_scene_analysis.data._utils', 'torchmetrics.retrieval.average_precision', 'torchmetrics.image.ssim', 'scipy.stats.ncf_ufunc', 'pkg_resources.extern.jaraco.functools', 'scipy.interpolate.dfitpack', 'cv2.gapi.wip.draw', 'torchmetrics.retrieval.ndcg', 'cv2.ml', 'scipy.fft._basic_backend', 'cv2.gapi.own.detail', 'scipy.linalg._flinalg_py', 'torchmetrics.regression.mae', 'scipy.optimize._lbfgsb', 'nicr_scene_analysis_datasets.datasets.sunrgbd', 'scipy.linalg.matfuncs', 'scipy.optimize.tnc', 'nicr_scene_analysis_datasets.datasets.scenenetrgbd.dataset', 'torchmetrics.retrieval.hit_rate', 'scipy._lib.array_api_compat.array_api_compat.numpy', 'scipy.linalg._matfuncs_expm', 'scipy.special._spfun_stats', 'nicr_mt_scene_analysis.visualization', 'scipy.linalg.decomp_lu', 'transformers.integrations', 'nicr_scene_analysis_datasets.dataset_base._meta', 'scipy.optimize.zeros', 'scipy.linalg._matfuncs_sqrtm_triu', 'scipy.sparse.linalg', 'cv2.gapi.imgproc.fluid', 'scipy.optimize._linprog_doc', 'nicr_mt_scene_analysis.data.preprocessing.multiscale_supervision', 'pkg_resources', 'transformers.integrations.deepspeed', 'scipy.stats._constants', 'cv2.gapi.video', 'scipy.special._comb', 'scipy.interpolate._cubic', 'scipy.integrate._lsoda', 'scipy.optimize._zeros_py', 'nicr_mt_scene_analysis.data.preprocessing.clone', 'torchmetrics.functional.text.wip', 'scipy._lib', 'cv2.gapi.ot.cpu', 'regex._regex_core', 'torchmetrics.text.cer', 'scipy.sparse.csgraph', 'scipy.stats._rcont.rcont', 'nicr_mt_scene_analysis.visualization.semantic', 'scipy.integrate.quadpack', 'pkg_resources.extern.packaging.tags', 'torchmetrics.regression.mse', 'transformers.generation.beam_search', 'scipy.optimize._highs.cython.src._highs_constants', 'torchmetrics.wrappers.bootstrapping', 'nicr_mt_scene_analysis.data.preprocessing.resize', 'scipy.optimize._trustregion_constr.qp_subproblem', 'scipy.stats._multivariate', 'scipy.stats._continuous_distns', 'scipy.sparse._sparsetools', 'torchmetrics.utilities.prints', 'torchmetrics.functional.regression.explained_variance', 'pkg_resources.extern', 'scipy.optimize._moduleTNC', 'torchmetrics.utilities.distributed', 'torchmetrics.functional.text.bleu', 'scipy.ndimage.fourier', 'scipy.integrate', 'nicr_mt_scene_analysis.model.postprocessing.dense_base', 'nicr_scene_analysis_datasets.datasets.coco.dataset', 'scipy.linalg._basic', 'scipy._lib._uarray', 'scipy.optimize._lsq.lsq_linear', 'torchmetrics.text.chrf', 'scipy.interpolate._rbfinterp_pythran', 'scipy.optimize._slsqp_py', 'scipy.sparse.csgraph._matching', 'scipy.special._specfun', 'nicr_mt_scene_analysis.data.preprocessing.instance', 'scipy._lib._testutils', 'scipy.linalg._decomp_svd'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/EMSANet\")\n",
    "sys.argv = [\"run.py\"]\n",
    "from repositories.EMSANet.emsanet.args import ArgParserEMSANet\n",
    "from repositories.EMSANet.emsanet.data import get_datahelper\n",
    "from repositories.EMSANet.emsanet.model import EMSANet\n",
    "\n",
    "parser = ArgParserEMSANet()\n",
    "args = parser.parse_args()\n",
    "args.dataset = \"nyuv2\"\n",
    "data = get_datahelper(args)\n",
    "model = decorate_model_for_dict_input(\n",
    "    EMSANet(args, dataset_config=data.dataset_config).eval(), [\"rgb\", \"depth\"]\n",
    ")\n",
    "# model(torch.rand(1, 3, 480, 640), torch.rand(1, 1, 480, 640))\n",
    "results[\"nyudv2\"][\"EMSANet\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "del model, parser, args, data\n",
    "\n",
    "parser = ArgParserEMSANet()\n",
    "args = parser.parse_args()\n",
    "args.dataset = \"sunrgbd\"\n",
    "data = get_datahelper(args)\n",
    "model = decorate_model_for_dict_input(\n",
    "    EMSANet(args, dataset_config=data.dataset_config).eval(), [\"rgb\", \"depth\"]\n",
    ")\n",
    "# model(torch.rand(1, 3, 480, 640), torch.rand(1, 1, 480, 640))\n",
    "results[\"sunrgbd\"][\"EMSANet\"] = cal_flops_params(\n",
    "    model, shape=(544, 736), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")  # Since EMSANet only accepts resolution of multiple of 32(2^5), we use 544x736\n",
    "del model, parser, args, data\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2024-09-07 05:16:09.090034: E external/local_xla/xla/stream_executor/cuda/cuda_fft.cc:485] Unable to register cuFFT factory: Attempting to register factory for plugin cuFFT when one has already been registered\n",
      "2024-09-07 05:16:09.151849: E external/local_xla/xla/stream_executor/cuda/cuda_dnn.cc:8454] Unable to register cuDNN factory: Attempting to register factory for plugin cuDNN when one has already been registered\n",
      "2024-09-07 05:16:09.178092: E external/local_xla/xla/stream_executor/cuda/cuda_blas.cc:1452] Unable to register cuBLAS factory: Attempting to register factory for plugin cuBLAS when one has already been registered\n",
      "2024-09-07 05:16:11.418372: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/ESANet/src/build_model.py:29: UserWarning: Argument --channels_decoder is ignored when --decoder_chanels_mode decreasing is set.\n",
      "  warnings.warn('Argument --channels_decoder is ignored when '\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/ESANet/src/models/resnet.py:101: UserWarning: parameters groups, base_width and norm_layer are ignored in NonBottleneck1D\n",
      "  warnings.warn('parameters groups, base_width and norm_layer are '\n",
      "/defaultShare/archive/caojiaolong/DFormer_plus/RGBDBenchmark/repositories/ESANet/src/models/model.py:163: UserWarning: for the context module the learned upsampling is not possible as the feature maps are not upscaled by the factor 2. We will use nearest neighbor instead.\n",
      "  warnings.warn('for the context module the learned upsampling is '\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Device: cuda:0\n",
      "ESANet(\n",
      "  (activation): ReLU(inplace=True)\n",
      "  (encoder_rgb): ResNet(\n",
      "    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (act): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 128, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 256, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (4): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (5): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 512, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (encoder_depth): ResNet(\n",
      "    (conv1): Conv2d(1, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (act): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 128, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 256, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (4): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (5): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 512, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer0): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer1): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer2): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(8, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(8, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer3): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(256, 16, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(16, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(256, 16, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(16, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer4): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(512, 32, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(32, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(512, 32, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(32, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (skip_layer1): Sequential(\n",
      "    (0): ConvBNAct(\n",
      "      (conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (skip_layer2): Sequential(\n",
      "    (0): ConvBNAct(\n",
      "      (conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (skip_layer3): Sequential(\n",
      "    (0): ConvBNAct(\n",
      "      (conv): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (context_module): PyramidPoolingModule(\n",
      "    (features): ModuleList(\n",
      "      (0): Sequential(\n",
      "        (0): AdaptiveAvgPool2d(output_size=1)\n",
      "        (1): ConvBNAct(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Sequential(\n",
      "        (0): AdaptiveAvgPool2d(output_size=5)\n",
      "        (1): ConvBNAct(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "        )\n",
      "      )\n",
      "    )\n",
      "    (final_conv): ConvBNAct(\n",
      "      (conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (decoder): Decoder(\n",
      "    (decoder_module_1): DecoderModule(\n",
      "      (conv3x3): ConvBNAct(\n",
      "        (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "      )\n",
      "      (decoder_blocks): Sequential(\n",
      "        (0): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (1): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (2): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "      )\n",
      "      (upsample): Upsample(\n",
      "        (pad): Identity()\n",
      "        (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512)\n",
      "      )\n",
      "      (side_output): Conv2d(512, 40, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (decoder_module_2): DecoderModule(\n",
      "      (conv3x3): ConvBNAct(\n",
      "        (conv): Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "      )\n",
      "      (decoder_blocks): Sequential(\n",
      "        (0): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (1): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (2): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "      )\n",
      "      (upsample): Upsample(\n",
      "        (pad): Identity()\n",
      "        (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256)\n",
      "      )\n",
      "      (side_output): Conv2d(256, 40, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (decoder_module_3): DecoderModule(\n",
      "      (conv3x3): ConvBNAct(\n",
      "        (conv): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "      )\n",
      "      (decoder_blocks): Sequential(\n",
      "        (0): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (1): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (2): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "      )\n",
      "      (upsample): Upsample(\n",
      "        (pad): Identity()\n",
      "        (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128)\n",
      "      )\n",
      "      (side_output): Conv2d(128, 40, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (conv_out): Conv2d(128, 40, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (upsample1): Upsample(\n",
      "      (pad): Identity()\n",
      "      (conv): Conv2d(40, 40, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=40)\n",
      "    )\n",
      "    (upsample2): Upsample(\n",
      "      (pad): Identity()\n",
      "      (conv): Conv2d(40, 40, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=40)\n",
      "    )\n",
      "  )\n",
      ")\n",
      "Device: cuda:0\n",
      "ESANet(\n",
      "  (activation): ReLU(inplace=True)\n",
      "  (encoder_rgb): ResNet(\n",
      "    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (act): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 128, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 256, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (4): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (5): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 512, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (encoder_depth): ResNet(\n",
      "    (conv1): Conv2d(1, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (act): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(64, 64, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(64, 64, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(64, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(64, 128, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 128, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(128, 256, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(128, 256, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (3): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (4): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (5): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(256, 512, kernel_size=(3, 1), stride=(2, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 2), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "      (2): NonBottleneck1D(\n",
      "        (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "        (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "        (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "        (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (dropout): Dropout2d(p=0, inplace=False)\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer0): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer1): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(64, 4, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(4, 64, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer2): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(8, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(128, 8, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(8, 128, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer3): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(256, 16, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(16, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(256, 16, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(16, 256, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (se_layer4): SqueezeAndExciteFusionAdd(\n",
      "    (se_rgb): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(512, 32, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(32, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "    (se_depth): SqueezeAndExcitation(\n",
      "      (fc): Sequential(\n",
      "        (0): Conv2d(512, 32, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (1): ReLU(inplace=True)\n",
      "        (2): Conv2d(32, 512, kernel_size=(1, 1), stride=(1, 1))\n",
      "        (3): Sigmoid()\n",
      "      )\n",
      "    )\n",
      "  )\n",
      "  (skip_layer1): Sequential(\n",
      "    (0): ConvBNAct(\n",
      "      (conv): Conv2d(64, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (skip_layer2): Sequential(\n",
      "    (0): ConvBNAct(\n",
      "      (conv): Conv2d(128, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (skip_layer3): Sequential(\n",
      "    (0): ConvBNAct(\n",
      "      (conv): Conv2d(256, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (context_module): PyramidPoolingModule(\n",
      "    (features): ModuleList(\n",
      "      (0): Sequential(\n",
      "        (0): AdaptiveAvgPool2d(output_size=1)\n",
      "        (1): ConvBNAct(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Sequential(\n",
      "        (0): AdaptiveAvgPool2d(output_size=5)\n",
      "        (1): ConvBNAct(\n",
      "          (conv): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "        )\n",
      "      )\n",
      "    )\n",
      "    (final_conv): ConvBNAct(\n",
      "      (conv): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "      (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "      (act): ReLU(inplace=True)\n",
      "    )\n",
      "  )\n",
      "  (decoder): Decoder(\n",
      "    (decoder_module_1): DecoderModule(\n",
      "      (conv3x3): ConvBNAct(\n",
      "        (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "      )\n",
      "      (decoder_blocks): Sequential(\n",
      "        (0): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (1): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (2): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(512, 512, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(512, 512, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(512, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "      )\n",
      "      (upsample): Upsample(\n",
      "        (pad): Identity()\n",
      "        (conv): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=512)\n",
      "      )\n",
      "      (side_output): Conv2d(512, 37, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (decoder_module_2): DecoderModule(\n",
      "      (conv3x3): ConvBNAct(\n",
      "        (conv): Conv2d(512, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "      )\n",
      "      (decoder_blocks): Sequential(\n",
      "        (0): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (1): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (2): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(256, 256, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(256, 256, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(256, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "      )\n",
      "      (upsample): Upsample(\n",
      "        (pad): Identity()\n",
      "        (conv): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=256)\n",
      "      )\n",
      "      (side_output): Conv2d(256, 37, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (decoder_module_3): DecoderModule(\n",
      "      (conv3x3): ConvBNAct(\n",
      "        (conv): Conv2d(256, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (act): ReLU(inplace=True)\n",
      "      )\n",
      "      (decoder_blocks): Sequential(\n",
      "        (0): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (1): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "        (2): NonBottleneck1D(\n",
      "          (conv3x1_1): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_1): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn1): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (act): ReLU(inplace=True)\n",
      "          (conv3x1_2): Conv2d(128, 128, kernel_size=(3, 1), stride=(1, 1), padding=(1, 0))\n",
      "          (conv1x3_2): Conv2d(128, 128, kernel_size=(1, 3), stride=(1, 1), padding=(0, 1))\n",
      "          (bn2): BatchNorm2d(128, eps=0.001, momentum=0.1, affine=True, track_running_stats=True)\n",
      "          (dropout): Dropout2d(p=0, inplace=False)\n",
      "        )\n",
      "      )\n",
      "      (upsample): Upsample(\n",
      "        (pad): Identity()\n",
      "        (conv): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=128)\n",
      "      )\n",
      "      (side_output): Conv2d(128, 37, kernel_size=(1, 1), stride=(1, 1))\n",
      "    )\n",
      "    (conv_out): Conv2d(128, 37, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))\n",
      "    (upsample1): Upsample(\n",
      "      (pad): Identity()\n",
      "      (conv): Conv2d(37, 37, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=37)\n",
      "    )\n",
      "    (upsample2): Upsample(\n",
      "      (pad): Identity()\n",
      "      (conv): Conv2d(37, 37, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), groups=37)\n",
      "    )\n",
      "  )\n",
      ")\n",
      "Deleted module: {'pandas.io.parquet', 'tensorflow._api.v2.compat.v1.nest', 'tensorflow.python.debug.cli.cli_shared', 'pandas._libs.tslibs.strptime', 'tensorflow.python.distribute.cluster_resolver.slurm_cluster_resolver', 'tensorflow.python.ops.ragged.ragged_bincount_ops', 'pandas.core.sorting', 'tensorflow._api.v2.compat.v1.bitwise', 'tensorflow.python.ops.distributions.gamma', 'cv2.qt', 'tensorflow.python.keras.optimizer_v2', 'tensorflow.python.tpu.tpu_system_metadata', 'tensorflow.python.keras.saving.saved_model.layer_serialization', 'pandas.core.strings', 'tensorflow._api.v2.nest', 'tensorflow.python.util.dispatch', 'scipy.linalg.flinalg', 'ignite.handlers', 'tensorflow.python.tools.module_util', 'h5py.h5pl', 'pandas.core.dtypes.generic', 'tensorflow.python.training.saving.saveable_object_util', 'tensorflow.python.data.experimental.ops.distribute', 'cv2.gapi.wip.gst', 'tensorflow.python.checkpoint.save_util_v1', 'pandas._config.dates', 'namex.convert', 'tensorflow.python.keras.initializers.initializers_v2', 'keras.src.applications.mobilenet', 'tensorflow.python.client._pywrap_events_writer', 'tensorflow.python.autograph.utils.tensors', 'scipy.linalg._decomp_lu', 'pandas.core.internals', 'tensorflow.python.tpu.ops.tpu_ops', 'keras.src.optimizers.rmsprop', 'pandas.core.array_algos.quantile', 'tensorflow.python.training.input', 'pandas.io.excel._xlrd', 'tensorflow.python.data.ops.optional_ops', 'tensorflow.python.ops.numerics', 'tensorflow.python.data.experimental.ops.resampling', 'keras.api.legacy.saving', 'tensorflow.python.ops.gen_uniform_quant_ops', 'tensorflow.python.data.experimental.ops.io', 'tensorflow.python.autograph.utils.misc', 'keras', 'tensorflow.python.util.fast_module_type', 'tensorflow.core.function.trace_type.trace_type_builder', 'tensorflow.python.autograph.pyct.ast_util', 'packaging.tags', 'pandas.core.arrays.arrow.array', 'keras.src.backend.tensorflow.distribution_lib', 'tensorflow._api.v2.__internal__.test.combinations', 'tensorflow.python.distribute.device_util', 'tensorflow._api.v2.compat.v2.linalg', 'cityscapesscripts.helpers.labels', 'tensorflow.python.data.experimental.ops.random_ops', 'tensorflow._api.v2.compat.v1.data.experimental', 'keras.src.trainers.data_adapters', 'src.prepare_data', 'scipy.sparse.linalg._onenormest', 'keras.src.layers.pooling.max_pooling3d', 'tensorflow.python.debug.wrappers.grpc_wrapper', 'tensorflow', 'keras.src.models.sequential', 'tensorflow._api.v2.compat.v1.manip', 'tensorflow.python.tpu.tpu_feed', 'tensorflow.python.ops.distributions.kullback_leibler', 'tensorflow.core.protobuf.rewriter_config_pb2', 'cv2.gapi.wip', 'tensorflow.python.distribute.mirrored_run', 'pandas.core.arrays.base', 'tensorflow._api.v2.quantization', 'tensorflow.lite.python.metrics.wrapper', 'keras.src.layers.rnn.conv_lstm1d', 'tensorflow.python.eager.polymorphic_function.transform', 'tensorflow.lite.python.tflite_keras_util', 'tensorflow._api.v2.compat.v2.compat.v2', 'keras.src.ops.symbolic_arguments', 'pandas._libs.tslibs.timedeltas', 'keras.src.tree', 'tensorflow.python.ops.linalg.sparse', 'tensorflow.python.ops.gen_io_ops', 'pandas._libs.tslib', 'tensorflow.python.data.experimental', 'attr.converters', 'tensorflow.python.keras.layers', 'pandas.core.arrays._mixins', 'ignite.metrics.regression._base', 'tensorflow.python.debug.cli.cli_config', 'keras.src.layers.regularization.spatial_dropout', 'pandas.core.tools.numeric', 'scipy.linalg._flinalg', 'scipy.sparse._lil', 'tensorflow._api.v2.lookup.experimental', 'tensorflow.python.autograph.pyct.parser', 'tensorflow.python.types.internal', 'tensorflow.python.ops.gen_manip_ops', 'keras.api.ops.numpy', 'tensorflow.python.tpu.tensor_tracer', 'keras.src.utils.progbar', 'keras.src.layers.preprocessing.string_lookup', 'pandas.core.missing', 'h5py._hl.selections2', 'keras.api.regularizers', 'pandas.core.indexes.accessors', 'tensorflow.python.eager.polymorphic_function.polymorphic_function', 'pandas.api.extensions', 'tensorflow.lite.python.analyzer_wrapper', 'pandas._libs.tslibs.timestamps', 'tensorflow.python.keras.saving.saved_model.base_serialization', 'tensorflow.python.ops.ragged.ragged_tensor', 'tensorflow.python.ops.ragged.ragged_math_ops', 'tensorflow.python.ops.template', 'tensorflow.python.ops.gen_state_ops', 'tensorflow.python.autograph.core.converter', 'keras.src.layers.rnn.simple_rnn', 'tensorflow.python.ops.ragged.ragged_tensor_value', 'tensorflow.python.keras.initializers', 'tensorflow.python.ops.ragged.row_partition', 'tensorflow._api.v2.__internal__.autograph', 'pandas.core.computation.common', 'tensorflow._api.v2.test', 'keras.src.layers.pooling.average_pooling3d', 'tensorflow._api.v2.profiler.experimental', 'tensorflow.python.compiler.xla.experimental.xla_sharding', 'termcolor', 'tensorflow.python.ops.ragged.ragged_tensor_shape', 'rich._loop', 'tensorflow.python.ops.gen_tpu_ops', 'tensorflow.python.data.ops.dataset_ops', 'tensorflow.core.util', 'tensorflow.python.feature_column.feature_column_v2', 'keras.src.layers.merging.minimum', 'rich._fileno', 'tensorflow.python.framework._op_def_registry', 'tensorflow.python.training.optimizer', 'tensorflow.python.ops.numpy_ops.np_random', 'pandas.core.arrays._arrow_string_mixins', 'scipy.sparse._bsr', 'tensorflow.python.ops.while_v2', 'tensorflow._api.v2.compat.v1.xla.experimental', 'keras.src.utils.text_dataset_utils', 'tensorflow.python.compiler.xla', 'keras.src.metrics.iou_metrics', 'pandas._testing', 'tensorflow.python.ops.parallel_for.pfor', 'tensorflow._api', 'tensorflow.python.tpu.tpu_embedding_v3', 'pandas.core.util.hashing', 'tensorflow.python.saved_model.simple_save', 'tensorflow.python.ops.distributions.normal', 'tensorflow.python.ops.random_grad', 'keras.src.ops.node', 'keras.src.layers.activations.softmax', 'keras.src.optimizers.lamb', 'keras.src.legacy.saving', 'tensorflow._api.v2.raw_ops', 'tensorflow._api.v2.compat.v2.data.experimental.service', 'tensorflow.python.ops.gradients_util', 'tensorflow.python.debug.wrappers.local_cli_wrapper', 'tensorflow.python.types', 'ignite.handlers.ema_handler', 'tensorflow.python.ops.nccl_ops', 'tensorflow.python.summary.writer.writer_cache', 'tensorflow.tools.compatibility', 'tensorflow._api.v2.__internal__.train', 'keras.api.utils', 'tensorflow.python.keras.layers.dense_attention', 'gast.version', 'tensorflow.python.saved_model.function_serialization', 'tensorflow._api.v2.compat.v2.io', 'tensorflow.python.checkpoint.restore', 'tensorflow.python.data.experimental.service._pywrap_server_lib', 'tensorflow.python.feature_column.feature_column', 'pandas.core._numba', 'tensorflow.compiler.xla.xla_data_pb2', 'tensorflow.python.distribute.distribute_coordinator', 'tensorflow.python.keras.metrics', 'tensorflow.python.distribute.coordinator.remote_value', 'keras.src.utils.model_visualization', 'tensorflow._api.v2.__internal__.decorator', 'ignite.metrics.precision_recall_curve', 'scipy._lib.deprecation', 'pandas.util', 'tensorflow.python.framework', 'keras.src.layers.activations.leaky_relu', 'keras.src.ops', 'keras.src.layers.pooling.global_average_pooling2d', 'keras.src.datasets.cifar', 'scipy.sparse.extract', 'keras.src.utils.sequence_utils', 'tensorflow.python.util.nest_util', 'tensorflow.python.ops.gen_batch_ops', 'numpy.testing.overrides', 'keras.src.models.variable_mapping', 'tensorflow.python.keras.engine.base_layer', 'pandas.io.pytables', 'keras.src.optimizers.adamw', 'pandas.io.parsers', 'torch.distributed.optim.named_optimizer', 'tensorflow.python.keras.mixed_precision.device_compatibility_check', 'h5py._hl.base', 'rich.region', 'xml.sax.handler', 'tensorflow.lite.tools.optimize.debugging.python', 'tensorflow.python.ops.optional_grad', 'h5py._hl.filters', 'pandas.core.groupby.groupby', 'tensorflow.python.debug.cli.profile_analyzer_cli', 'tensorflow.python.training.ftrl', 'tensorflow.python.profiler.trace', 'tensorflow.core.framework.types_pb2', 'keras.src.layers.preprocessing', 'keras.src.layers.rnn.lstm', 'keras.src.layers.pooling.global_max_pooling2d', 'keras.src.layers.rnn.conv_lstm2d', 'tensorflow.distribute.experimental.rpc.proto', 'tensorflow._api.v2.data.experimental.service', 'tensorflow.python.ops.linalg.property_hint_util', 'tensorflow.python.lib.io.python_io', 'tensorflow.lite.python.interpreter_wrapper._pywrap_tensorflow_interpreter_wrapper', 'ignite.metrics.regression.median_absolute_error', 'tensorflow.python.user_ops.ops.gen_user_ops', 'h5py.h5r', 'tensorflow.python.trackable', 'keras.src.ops.image', 'absl.flags._argument_parser', 'keras.src.random.seed_generator', 'tensorflow.python.data.experimental.ops.parsing_ops', 'tensorflow.compiler.mlir.quantization', 'tensorflow.python.ops.gen_spectral_ops', 'tensorflow.python.tpu.feature_column_v2', 'cv2.flann', 'cv2.gapi.ot', 'tensorflow.python.flags_pybind', 'keras.src.layers.normalization.group_normalization', 'rich.default_styles', 'keras.src.regularizers.regularizers', 'tensorflow.core.example.feature_pb2', 'keras.src.layers.preprocessing.center_crop', 'tensorflow.python.ops.control_flow_state', 'torch.distributed.algorithms.model_averaging.utils', 'tensorflow.lite.toco.types_pb2', 'pandas._libs.properties', 'tensorflow.python.ops.gen_training_ops', 'tensorflow._api.v2.saved_model', 'wrapt', 'tensorflow.python.keras.engine.data_adapter', 'tensorflow.python.autograph.pyct.transformer', 'tensorflow.python.summary.writer.event_file_writer_v2', 'ml_dtypes._iinfo', 'keras.src.layers.regularization.gaussian_dropout', 'tensorflow.compiler.mlir.quantization.tensorflow.calibrator', 'keras.src.utils.summary_utils', 'tensorflow.python.data.util.structure', 'tensorflow.python.training.adadelta', 'ignite.metrics.nlp.utils', 'ignite.metrics.metrics_lambda', 'tensorflow.core.function.trace_type.default_types_pb2', 'tensorflow.python.keras.legacy_tf_layers', 'tensorflow.python.ops.gen_stateless_random_ops', 'tensorflow.python.summary.writer.event_file_writer', 'tensorflow.core.framework.resource_handle_pb2', 'tensorflow.tsl', 'tensorflow.python.ops.gen_ragged_math_ops', 'tensorflow.python.ops.bitwise_ops', 'scipy.sparse.linalg._isolve.lsmr', 'ignite.metrics.mean_absolute_error', 'pandas.arrays', 'keras.api.datasets.mnist', 'tensorflow.python.lib', 'keras.src.layers.layer', 'keras.api.applications.efficientnet', 'pandas.core.arrays.masked', 'h5py.h5ds', 'tensorflow._api.v2.io.gfile', 'tensorflow._api.v2.compat.v2.test', 'tensorflow.python.ops.io_ops', 'tensorflow.python.platform._pywrap_stacktrace_handler', 'tensorflow._api.v2.saved_model.experimental', 'tensorflow.python.util', 'tensorflow.python.autograph.operators.slices', 'tensorflow.python.training.session_run_hook', 'pandas.core.internals.api', 'tensorflow.core.protobuf.config_pb2', 'tensorflow.python.client._pywrap_device_lib', 'tensorflow.python.saved_model.load_options', 'tensorflow.python.ops.nn_impl', 'tensorflow.python.ops.gen_logging_ops', 'tensorflow.python.autograph.converters.return_statements', 'pandas._libs.indexing', 'tensorflow.python.keras.utils.tf_utils', 'tensorflow._api.v2.compat.v1.resource_loader', 'tensorflow.python.summary', 'tensorflow.python.autograph.impl.conversion', 'tensorflow.python.framework.versions', 'keras.api.random', 'wrapt.wrappers', 'tensorflow.tools', 'cv2.gapi', 'tensorflow.python.saved_model.fingerprinting', 'keras.src.legacy.saving.serialization', 'tensorflow.python.ops.gen_linalg_ops', 'pandas.io.formats.info', 'tensorflow.python.debug.lib.op_callbacks_common', 'tensorflow.python.data.experimental.ops.prefetching_ops', 'wrapt.importer', 'ignite.engine.events', 'scipy.linalg._decomp_qr', 'keras.src.losses', 'tensorflow.python.eager.execute', 'rich.segment', 'keras.src.layers.input_spec', 'tensorflow.python.ops.linalg', 'tensorflow._api.v2.compat.v1.layers', 'cv2.ipp', 'tensorflow.core.framework.model_pb2', 'tensorflow._api.v2.compat.v1.lite.experimental.authoring', 'tensorflow.python.compiler.tensorrt', 'tensorflow.python.ops.control_flow_switch_case', 'tensorflow.python.pywrap_sanitizers', 'tensorflow.python.framework.tensor_spec', 'rich.scope', 'rich.console', 'tensorflow.python.saved_model.save', 'tensorflow.python.saved_model.constants', 'tensorflow.python.framework.tensor_conversion_registry', 'tensorflow.python.distribute.load_context', 'tensorflow.tsl.protobuf', 'tensorflow._api.v2.experimental', 'tensorflow.python.ops.ragged.ragged_functional_ops', 'tensorflow.python.framework.ops', 'keras.src.layers.convolutional.base_conv_transpose', 'tensorflow.python.ops.linalg.linear_operator_householder', 'pandas.core.arrays.boolean', 'tensorflow.python.framework.type_spec', 'tensorflow.python.framework.kernels', 'keras.src.saving.keras_saveable', 'keras.src.layers.convolutional.depthwise_conv1d', 'tensorflow.python.ops.gen_optional_ops', 'tensorflow.python.feature_column', 'tensorflow.compiler.tf2tensorrt._pywrap_py_utils', 'tensorflow.python.training.adam', 'tensorflow.python.keras.engine.training_utils_v1', 'opt_einsum._version', 'keras.src.applications.resnet', 'tensorflow.python.util.tf_should_use', 'attr._funcs', 'tensorflow.python.ops.ragged.ragged_array_ops', 'tensorflow.python.distribute.central_storage_strategy', 'rich.color_triplet', 'tensorflow.python.lib.io.file_io', 'repositories.ESANet', 'tensorflow._api.v2.compat.v1.distribute.cluster_resolver', 'rich.screen', 'tensorflow._api.v2.__internal__.distribute.interim', 'tensorflow.python.profiler.profiler', 'tensorflow.python.eager.def_function', 'keras.api.applications.mobilenet_v3', 'tensorflow._api.v2.compat.v2.train', 'tensorflow.python.keras.saving.model_config', 'tensorflow._api.v2.compat.v1.train.experimental', 'keras.src.layers.activations.elu', 'tensorflow.python.data.util.convert', 'tensorflow._api.v2.compat.v2.types.experimental', 'pandas.core.array_algos.putmask', 'tensorflow.python.distribute.input_ops', 'tensorflow._api.v2.config.experimental', 'tensorflow.python.framework.tfrt_utils', 'keras.src.layers.core.input_layer', 'scipy.sparse._csr', 'tensorflow._api.v2.compat.v2.__internal__.monitoring', 'pandas.core.arrays.arrow', 'tensorflow.python.checkpoint.sharding.sharding_policies', 'tensorflow.python.keras.layers.pooling', 'tensorflow._api.v2.compat.v2.__internal__.distribute', 'tensorflow.python.ops.losses', 'tensorflow.dtensor.python.save_restore', 'tensorflow.python.distribute.input_lib', 'tensorflow.python.distribute.distribute_lib', 'tensorflow.core.protobuf.tpu.tpu_embedding_configuration_pb2', 'tensorflow.python.keras.utils.tf_inspect', 'tensorflow.python.profiler.tfprof_logger', 'tensorflow.python.distribute.values', 'tensorflow._api.v2.compat.v2.distribute.coordinator', 'tensorflow.python.keras.engine.training_utils', 'keras.src.layers.normalization.batch_normalization', 'tensorflow.core.framework.attr_value_pb2', 'keras.src.optimizers.adafactor', 'tensorflow.python.framework.gpu_util', 'tensorflow._api.v2.compat.v2.experimental.dtensor', 'pandas.api.interchange', 'scipy._lib.array_api_compat.array_api_compat.common._helpers', 'tensorflow._api.v2.compat.v2.__internal__.graph_util', 'tensorflow._api.v2.config.threading', 'tensorflow.python.lib.core._pywrap_py_func', 'tensorflow.python.tpu.tpu_embedding_v2_utils', 'tensorflow._api.v2.tpu.experimental', 'tensorflow.python.autograph.converters.call_trees', 'pandas.io.sas', 'tensorflow.python.keras.engine.sequential', 'pandas.core.dtypes.inference', 'tensorflow.python.util._tf_stack', 'tensorflow.python.profiler', 'tensorflow.python._pywrap_py_exception_registry', 'tensorflow.python.util.is_in_graph_mode', 'tensorflow.python.ops.ragged.ragged_image_ops', 'ignite.metrics.confusion_matrix', 'scipy.sparse.linalg._isolve._gcrotmk', 'keras.src.layers.core.dense', 'tensorflow.python.data.experimental.ops.error_ops', 'tensorflow.python.ops.collective_ops', 'pandas.core.indexes.api', 'tensorflow.python.distribute.cluster_resolver.tfconfig_cluster_resolver', 'tensorflow.python.types.core', 'wrapt.arguments', 'tensorflow.python.saved_model.model_utils', 'h5py._hl.datatype', 'keras.api.callbacks', 'h5py._hl.group', 'tensorflow.python.ops.gen_random_index_shuffle_ops', 'tensorflow.python.framework.device', 'scipy.sparse.data', 'ignite.metrics.regression.mean_absolute_relative_error', 'tensorflow.python.util._pywrap_nest', 'tensorflow.python.util.object_identity', 'tensorflow.python.framework.graph_util_impl', 'tensorflow._api.v2.compat.v2.__internal__.tracking', 'tensorflow.python.autograph.core', 'tensorflow.python.platform.test', 'pandas._libs.ops', 'torch.distributed.optim.post_localSGD_optimizer', 'tensorflow._api.v2.test.experimental', 'keras.src.layers.normalization.layer_normalization', 'pandas._config', 'pandas.tseries.offsets', 'tensorflow.core.protobuf.tpu.optimization_parameters_pb2', 'tensorflow.python.keras.saving.saved_model.save', 'tensorflow.python.autograph.converters', 'tensorflow.core.framework.variable_pb2', 'keras.src.layers.rnn.bidirectional', 'tensorflow.python.framework.type_utils', 'tensorflow.core.framework.versions_pb2', 'tensorflow.compiler.mlir.lite', 'cv2.misc.version', 'scipy.sparse._dia', 'rich.style', 'rich.padding', 'keras.src.metrics', 'tensorflow.python.ops.ragged.ragged_util', 'tensorflow._api.v2.compat.v2.math', 'scipy.sparse.linalg._isolve.iterative', 'pandas.io', 'keras.src.layers.merging.multiply', 'pandas.core.indexers.utils', 'keras.src.layers.normalization.unit_normalization', 'keras.src.layers.preprocessing.resizing', 'ignite.metrics.classification_report', 'tensorflow.python.lib.core', 'keras.src.backend.tensorflow', 'tensorflow.python.ops.structured.structured_tensor', 'rich.panel', 'tensorflow.python.autograph.pyct.origin_info', 'tensorflow._api.v2.compat.v2.queue', 'tensorflow.lite.tools.optimize.debugging.python.debugger', 'numpy._core', 'tensorflow.python.ops.critical_section_ops', 'tensorflow.core.function.capture.capture_container', 'pandas.core.shared_docs', 'tensorflow.python.framework.type_spec_registry', 'tensorflow.python.ops.linalg.linear_operator_identity', 'opt_einsum.backends.tensorflow', 'tensorflow.python.ops.clip_ops', 'keras.src.backend.common.variables', 'tensorflow.python.keras.distribute.distributed_file_utils', 'attr._version_info', 'tensorflow.python.eager.polymorphic_function.composite_tensor_utils', 'keras.api.applications.mobilenet_v2', 'keras.src.random', 'tensorflow._api.v2.compat.v1.sysconfig', 'tensorflow.python.keras.losses', 'six.moves.urllib.request', 'tensorflow.python.util.tf_stack', 'tensorflow.python.framework.graph_to_function_def', 'scipy._lib.array_api_compat.array_api_compat.common._aliases', 'tensorflow.python.ops.losses.losses_impl', 'tensorflow.python.keras.engine.training', 'tensorflow._api.v2.compat.v2.__internal__.distribute.interim', 'tensorflow.python.saved_model.registration.registration', 'keras.src.utils', 'tensorflow.lite.profiling.proto', 'src', 'opt_einsum.path_random', 'tensorflow.distribute', 'tensorflow._api.v2.compat.v1.autograph.experimental', 'pandas.core.dtypes.concat', 'keras.src.tree.optree_impl', 'pandas.core.util', 'tensorflow.python.autograph.operators.py_builtins', 'tensorflow._api.v2.compat.v1.saved_model.utils', 'keras.src.layers.activations', 'pandas.core.ops.array_ops', 'tensorflow.python.ops.distributions.multinomial', 'tensorflow.python.trackable.autotrackable', 'keras.src.saving.saving_api', 'tensorflow.python.distribute.cluster_resolver.cluster_resolver', 'tensorflow.python.framework.flexible_dtypes', 'ignite.metrics.regression.geometric_mean_absolute_error', 'tensorflow.tools.compatibility.all_renames_v2', 'keras.src.utils.argument_validation', 'src.datasets.sunrgbd.pytorch_dataset', 'tensorflow.python.ops.gen_ragged_array_ops', 'tensorflow.python.ops.rnn_grad', 'keras.src.layers.rnn.dropout_rnn_cell', 'tensorflow.python.autograph.converters.variables', 'keras.src.layers.reshaping.cropping2d', 'src.models.resnet', 'tensorflow._api.v2.compat.v2.random.experimental', 'tensorflow._api.v2.compat.v2.xla.experimental', 'tensorflow._api.v2.linalg.experimental', 'tensorflow.python.ops.ragged.ragged_batch_gather_with_default_op', 'keras.src.optimizers.schedules', 'tensorflow.python.saved_model.signature_serialization', 'scipy.sparse._index', 'pandas.core.arrays.datetimelike', 'tensorflow._api.v2.types', 'tensorflow.python.data.experimental.service.server_lib', 'tensorflow.python.distribute.multi_worker_util', 'tensorflow._api.v2.__internal__.types', 'tensorflow._api.v2.compat.v1.mixed_precision', 'tensorflow.python.saved_model.model_utils.export_output', 'tensorflow.compiler.mlir.quantization.tensorflow.python.pywrap_quantize_model', 'scipy.linalg.lapack', 'pandas.core.dtypes.cast', 'pandas.core.arrays._ranges', 'tensorflow._api.v2.distribute.experimental.rpc', 'scipy.linalg._matfuncs_sqrtm', 'tensorflow.python.module', 'tensorflow.python.debug.cli', 'keras.src.backend.tensorflow.layer', 'tensorflow.python.autograph.core.function_wrappers', 'tensorflow.compiler.mlir.quantization.tensorflow.python.representative_dataset', 'tensorflow.python.keras.mixed_precision.policy', 'wrapt.__wrapt__', 'tensorflow.python.training.momentum', 'tensorflow.python.keras.engine.training_v1', 'tensorflow.python.ops.gen_data_flow_ops', 'pandas.core.internals.base', 'tensorflow._api.v2.compat.v1', 'tensorflow.python.trackable.layer_utils', 'cv2.utils.nested', 'absl.testing._pretty_print_reporter', 'tensorflow.python.distribute.coordinator.cluster_coordinator', 'tensorflow.core.protobuf.meta_graph_pb2', 'tensorflow.core.protobuf.tpu', 'tensorflow.lite.experimental.microfrontend.python', 'tensorflow.python.ops.cond', 'tensorflow.compiler.tf2tensorrt', 'repositories.ESANet.train', 'tensorflow.python.ops.metrics_impl', 'ignite.handlers.lr_finder', 'tensorflow.python.keras.layers.core', 'h5py._hl.vds', 'tensorflow.python.keras.saving.saved_model.serialized_attributes', 'tensorflow._api.v2.types.experimental.distributed', 'tensorflow.python.training.basic_session_run_hooks', 'tensorflow.python.util._pywrap_tensor_float_32_execution', 'tensorflow.python.distribute.experimental.multi_worker_mirrored_strategy', 'tensorflow.python.saved_model.signature_def_utils', 'keras.src.applications.efficientnet', 'keras.src.layers.reshaping.cropping3d', 'keras.api.ops', 'tensorflow.python.data.ops.batch_op', 'tensorflow.python.ops.array_ops', 'tensorflow.python.keras.optimizer_v2.adadelta', 'tensorflow.python.framework.auto_control_deps', 'keras.src.layers.regularization.alpha_dropout', 'keras.src.backend.exports', 'tensorflow._api.v2.config.optimizer', 'tensorflow._api.v2.sets', 'tensorflow.python.tpu.tensor_tracer_pb2', 'scipy.sparse.linalg._expm_multiply', 'flatbuffers._version', 'ignite.distributed.utils', 'pandas.core.arrays.string_', 'pandas.io.clipboards', 'ignite.metrics.regression.mean_normalized_bias', 'absl.flags._flagvalues', 'tensorflow.python.training.quantize_training', 'tensorflow.python.ops.numpy_ops.np_dtypes', 'keras.src.utils.traceback_utils', 'keras.src.layers.preprocessing.random_translation', 'tensorflow.python.autograph.pyct.error_utils', 'pandas.core.groupby.base', 'tensorflow.python.ops.control_flow_assert', 'tensorflow.python.client.device_lib', 'pandas.core.indexes.datetimelike', 'cv2.ml', 'tensorflow._api.v2.dtypes.experimental', 'scipy.linalg.matfuncs', 'tensorflow.python.autograph.utils', 'tensorflow._api.v2.compat.v2.__internal__.train', 'tensorflow.lite.python.conversion_metadata_schema_py_generated', 'tensorflow._api.v2.compat.v1.logging', 'tensorflow._api.v2.compat.v2.__operators__', 'keras.src.layers.convolutional.conv2d_transpose', 'keras.src.layers.convolutional.separable_conv2d', 'tensorflow.python.distribute.failure_handling', 'pandas.core.methods', 'ignite.metrics.top_k_categorical_accuracy', 'tensorflow.python.keras.engine.compile_utils', 'tensorflow.python.debug.lib.common', 'ignite.metrics.multilabel_confusion_matrix', 'scipy.sparse.linalg', 'tensorflow.python.data.experimental.ops.cardinality', 'cv2.gapi.imgproc.fluid', 'tensorflow.python.debug', 'attr._compat', 'pandas.io.api', 'tensorflow.python.eager.backprop_util', 'tensorflow.python.training.saving', 'keras.src.utils.image_dataset_utils', 'cv2.gapi.video', 'pandas.core.computation.check', 'keras.src.quantizers', 'tensorflow._api.v2.compat.v2.summary.experimental', 'tensorflow.python.keras.utils.metrics_utils', 'tensorflow.python.framework.composite_tensor_gradient', 'ignite', 'pandas.io.parsers.base_parser', 'keras.src.layers.convolutional.base_separable_conv', 'tensorflow._api.v2.experimental.numpy', 'tensorflow.python.ops.while_loop', 'wrapt.weakrefs', 'pandas.core.base', 'scipy.sparse.csgraph', 'tensorflow._api.v2.compat.v2.config.experimental', 'pandas._libs.tslibs.ccalendar', 'tensorflow.python.eager.polymorphic_function.attributes', 'tensorflow.python.ops.gradient_checker', 'attr.filters', 'tensorflow.security.fuzzing.py', 'absl.logging', 'keras.src.layers', 'pandas.core.indexes.timedeltas', 'tensorflow.python.ops.signal.mfcc_ops', 'tensorflow._api.v2.queue', 'keras.src.callbacks.tensorboard', 'pandas.io.excel._util', 'tensorflow.python.autograph.operators', 'tensorflow._api.v2.compat.v2', 'rich.align', 'pandas._libs.hashtable', 'tensorflow._api.v2.compat.v1.lite.constants', 'tensorflow.tsl.protobuf.coordination_config_pb2', 'pandas.core.computation.pytables', 'tensorflow.python.ops.linalg.slicing', 'tensorflow.python.training.saving.trace_saveable_util', 'rich._pick', 'keras.src.export', 'tensorflow.python.distribute.tpu_values', 'tensorflow.python.ops.gradient_checker_v2', 'tensorflow.python.ops.gen_composite_tensor_ops', 'tensorflow._api.v2.compat.v1.image', 'tensorflow.python.data.ops.test_mode', 'gast.ast3', 'tensorflow.python.grappler._pywrap_tf_optimizer', 'tensorflow.python.framework.byte_swap_tensor', 'cv2.cuda', 'tensorflow.python.keras.utils.tf_contextlib', 'pandas.core.arrays.timedeltas', 'keras.api.distribution', 'scipy._lib.array_api_compat.array_api_compat.numpy._aliases', 'tensorflow.python.ops.gen_debug_ops', 'scipy.sparse.linalg._matfuncs', 'namex.export', 'pandas.core.util.numba_', 'pandas.io.excel._base', 'pandas._libs.tslibs.parsing', 'tensorflow._api.v2.compat.v1.tpu.experimental', 'keras.src.layers.activations.activation', 'optree.registry', 'tensorflow.dtensor.python.dtensor_device', 'pandas.core.internals.ops', 'keras.src.callbacks.terminate_on_nan', 'rich.box', 'tensorflow.lite.python.metrics.metrics_interface', 'tensorflow.python.distribute.parallel_device.parallel_device', 'pandas.core.ops.docstrings', 'tensorflow._api.v2.compat.v1.mlir.experimental', 'tensorflow.python.ops.linalg.linear_operator_block_lower_triangular', 'tensorflow._api.v2.__internal__.mixed_precision', 'tensorflow.distribute.experimental.rpc', 'rich', 'tensorflow.python.saved_model', 'tensorflow.python.autograph.operators.logical', 'keras.src.layers.convolutional.conv1d_transpose', 'keras.api.datasets.fashion_mnist', 'tensorflow.python.debug.lib.debug_utils', 'keras.src.layers.reshaping.repeat_vector', 'tensorflow.python.data.ops.shuffle_op', 'tensorflow.python.ops.cond_v2', 'keras.api.applications.nasnet', 'tensorflow.python.distribute.sharded_variable', 'tensorflow._api.v2.compat.v2.io.gfile', 'absl.flags._defines', 'tensorflow.python.data.experimental.ops.random_access', 'tensorflow._api.v2.feature_column', 'tensorflow.python.debug.cli.ui_factory', 'pandas.testing', 'tensorflow.python.ops.control_flow_v2_func_graphs', 'tensorflow._api.v2.compat.v2.tpu', 'tensorflow._api.v2.sparse', 'tensorflow.python.ops.ragged.ragged_where_op', 'ignite.handlers.state_param_scheduler', 'pandas._libs.sparse', 'tensorflow.python.keras.engine.training_distributed_v1', 'tensorflow._api.v2.compat.v2.lite.experimental.authoring', 'keras.src.export.export_lib', 'pandas.io.orc', 'numpy.testing._private', 'tensorflow.python.data.ops.repeat_op', 'keras.src.callbacks.reduce_lr_on_plateau', 'tensorflow._api.v2.compat.v2.__internal__.feature_column', 'tensorflow._api.v2', 'tensorflow._api.v2.compat.v2.profiler.experimental.server', 'tensorflow._api.v2.__internal__.function', 'tensorflow.python.ops.random_crop_ops', 'tensorflow.python.ops.custom_gradient', 'tensorflow._api.v2.experimental.extension_type', 'tensorflow.python._pywrap_sanitizers', 'tensorflow.python.framework.op_def_library_pybind', 'absl.testing', 'pandas._libs.tslibs.offsets', 'tensorflow.python.framework.function_def_to_graph', 'scipy.sparse._base', 'tensorflow.python.ops.resource_variable_ops', 'tensorflow.python.tpu.tpu_embedding_base', 'tensorflow.python.util.deprecation', 'scipy.sparse.dia', 'keras.src.layers.pooling.global_max_pooling1d', 'tensorflow.python.saved_model.load', 'tensorflow.python.ops.data_flow_grad', 'tensorflow.python.debug.lib.debug_data', 'tensorflow.dtensor.python.d_variable', 'tensorflow.python.checkpoint.trackable_view', 'tensorflow._api.v2.compat.v1.test', 'keras.src.applications.inception_v3', 'pandas.core.frame', 'tensorflow.python.eager.polymorphic_function.compiler_ir', 'scipy._lib.array_api_compat.array_api_compat._internal', 'tensorflow.lite.python.convert_phase', 'tensorflow.python.tpu.topology', 'scipy.sparse.linalg.isolve', 'tensorflow.python.util.decorator_utils', 'tensorflow.python.platform.benchmark', 'ignite.metrics.nlp', 'tensorflow.python.distribute.merge_call_interim', 'tensorflow.python.keras.layers.embeddings', 'tensorflow.python.ops.gen_stateless_random_ops_v2', 'tensorflow.python.training.learning_rate_decay', 'ignite.metrics.gan.inception_score', 'tensorflow.lite', 'tensorflow.python.autograph.pyct.static_analysis.reaching_fndefs', 'tensorflow.lite.python.convert', 'tensorflow.python.checkpoint.checkpoint_adapter', 'tensorflow.python.distribute.strategy_combinations', 'cv2.utils.fs', 'tensorflow.compiler.mlir.quantization.tensorflow.python.quantize_model', 'pandas.core.arrays.datetimes', 'tensorflow.python.ops.inplace_ops', 'pandas.core.algorithms', 'tensorflow.python.ops.gen_list_ops', 'pandas._testing._warnings', 'tensorflow._api.v2.compat.v1.sets', 'tensorflow.python.keras.optimizer_v2.utils', 'tensorflow._api.v2.compat.v1.train', 'src.args', 'tensorflow._api.v2.compat.v2.distribute.experimental.coordinator', 'ignite.metrics.loss', 'tensorflow.python.distribute.cluster_resolver.tpu', 'tensorflow._api.v2.compat.v1.nn', 'pandas.io.stata', 'tensorflow.python.ops.nn_grad', 'tensorflow.dtensor', 'tensorflow._api.v2.compat.v1.metrics', 'pandas.core.strings.object_array', 'flatbuffers.util', 'numpy.testing._private.extbuild', 'tensorflow.python.distribute.distribute_coordinator_context', 'tensorflow.python.keras.optimizer_v1', 'keras.src.metrics.accuracy_metrics', 'flatbuffers.table', 'keras.api.legacy', 'tensorflow.python.autograph.pyct.static_analysis.liveness', 'tensorflow._api.v2.compat.v1.feature_column', 'tensorflow.python.autograph.utils.ag_logging', 'tensorflow.python.autograph.pyct.cache', 'tensorflow.python.keras.backend', 'tensorflow._api.v2.compat.v2.bitwise', 'keras.src.layers.preprocessing.random_flip', 'src.datasets.nyuv2.pytorch_dataset', 'keras.api.applications.efficientnet_v2', 'tensorflow.python.autograph.core.unsupported_features_checker', 'tensorflow.python.keras.legacy_tf_layers.variable_scope_shim', 'tensorflow.python.feature_column.feature_column_lib', 'keras.src.layers.preprocessing.tf_data_layer', 'pandas.io.gbq', 'tensorflow.python.autograph.converters.logical_expressions', 'tensorflow.python.util._pywrap_tfprof', 'pandas.core.tools.times', 'pandas.core.indexes.extension', 'tensorflow.python.util.tf_contextlib', 'tensorflow.python.training.warm_starting_util', 'tensorflow.core.lib', 'ignite.metrics.root_mean_squared_error', 'tensorflow.python.framework.dtypes', 'keras.src.layers.regularization', 'pandas._typing', 'tensorflow.python', 'opt_einsum.paths', 'tensorflow.core.framework.tensor_shape_pb2', 'tensorflow.python.data.experimental.ops.readers', 'tensorflow.python.framework.error_interpolation', 'tensorflow.python.ops.gen_audio_ops', 'tensorflow._api.v2.strings', 'tensorflow._api.v2.compat.v2.__internal__.types.data', 'tensorflow.python.util.compat', 'cv2.gapi.core.cpu', 'scipy.sparse._matrix_io', 'tensorflow.python.training.coordinator', 'pandas.core.reshape.concat', 'tensorflow.python.ops.gen_map_ops', 'tensorflow.python.keras.engine.functional', 'src.datasets.scenenetrgbd.scenenetrgbd', 'xml.sax.xmlreader', 'tensorflow.python.platform.self_check', 'scipy._lib._ccallback_c', 'keras.src.applications', 'keras.api.export', 'scipy.sparse._dok', 'keras.src.optimizers.nadam', 'tensorflow.python.ops.ref_variable', 'tensorflow.dtensor.proto', 'keras.api.constraints', 'keras.src.layers.preprocessing.random_crop', 'opt_einsum.parser', 'rich.errors', 'tensorflow.python.ops.losses.util', 'keras.src.layers.convolutional', 'rich.markup', 'tensorflow._api.v2.__internal__.saved_model.load', 'tensorflow.python.autograph.operators.variables', 'tensorflow.python.framework.smart_cond', 'tensorflow.python.debug.lib.dumping_callback', 'tensorflow.python.framework.override_binary_operator', 'h5py.h5t', 'tensorflow.python.ops.nn_fused_batch_norm_grad', 'h5py.utils', 'keras.src.callbacks.model_checkpoint', 'tensorflow._api.v2.compat.v1.lite', 'pandas.core.groupby', 'tensorflow.python.ops.linalg.linear_operator_diag', 'tensorflow.python.distribute.parallel_device', 'scipy.sparse.linalg.dsolve', 'cv2.gapi.streaming', 'pandas._libs.groupby', 'tensorflow._api.v2.compat', 'pandas.core.reshape', 'tensorflow.python.platform.gfile', 'pandas.api.indexers', 'tensorflow.python.checkpoint.checkpoint_options', 'zoneinfo._common', 'pandas.io.sql', 'tensorflow.python.distribute.tpu_util', 'tensorflow.lite.toco', 'tensorflow.python.framework.tensor_conversion', 'tensorflow.lite.tools.flatbuffer_utils', 'h5py.h5', 'keras.src.backend.tensorflow.trainer', 'tensorflow.python.dlpack', 'tensorflow.python.autograph.converters.continue_statements', 'tensorflow.python.trackable.base', 'tensorflow.python.ops.gen_clustering_ops', 'tensorflow.python.ops.ragged.ragged_operators', 'pandas.io.parsers.c_parser_wrapper', 'tensorflow.python.summary.summary', 'tensorflow.core.distributed_runtime', 'pandas.core.series', 'keras.src.applications.nasnet', 'tensorflow.python.types.trace', 'pandas.io.excel._xlsxwriter', 'tensorflow._api.v2.tpu', 'pandas.core.interchange.utils', 'tensorflow.python.distribute.mirrored_strategy', 'tensorflow.python.framework.common_shapes', 'scipy.sparse.linalg.matfuncs', 'cityscapesscripts.helpers', 'scipy.linalg._decomp_polar', 'torch.distributed.algorithms.model_averaging', 'tensorflow.python.keras.engine.training_arrays_v1', 'tensorflow.core.protobuf.struct_pb2', 'rich.measure', 'ignite.handlers.visdom_logger', 'keras.src.layers.reshaping', 'pandas._libs.json', 'pandas.core.window.online', 'tensorflow._api.v2.compat.v2.image', 'tensorflow.core.config', 'scipy.sparse._extract', 'pytz.exceptions', 'keras.api.ops.nn', 'tensorflow.compiler.mlir.quantization.tensorflow.python', 'tensorflow._api.v2.profiler', 'pandas.tseries.frequencies', 'tensorflow.python.feature_column.utils', 'src.datasets.scenenetrgbd.pytorch_dataset', 'tensorflow.python.ops.rnn', 'keras.api.applications.inception_v3', 'tensorflow.python.autograph.operators.conditional_expressions', 'tensorflow.python.keras.optimizer_v2.adagrad', 'repositories.ESANet.src.build_model', 'keras.src.trainers.compile_utils', 'pandas.core.internals.construction', 'tensorflow._api.v2.train', 'keras.src.metrics.metrics_utils', 'keras.src.ops.math', 'pandas.core.arrays.categorical', 'src.datasets.cityscapes.cityscapes', 'tensorflow.python.training.supervisor', 'scipy.version', 'pandas.io.formats.format', 'tensorflow.python.tpu.tpu_embedding_v3_utils', 'tensorflow.python.profiler.internal.flops_registry', 'tensorflow._api.v2.compat.v2.__internal__', 'tensorflow.python.ops.gen_cudnn_rnn_ops', 'tensorflow.python.keras.engine.keras_tensor', 'keras.src.layers.rnn.stacked_rnn_cells', 'tensorflow.python.eager.remote', 'scipy.sparse.lil', 'tensorflow.python.data.experimental.ops.interleave_ops', 'tensorflow.python.framework.none_tensor', 'tensorflow._api.v2.compat.v2.__internal__.decorator', 'keras.src.layers.attention.multi_head_attention', 'tensorflow._api.v2.profiler.experimental.server', 'tensorflow.python.data.ops.prefetch_op', 'tensorflow.python.trackable.trackable_utils', 'absl.testing.absltest', 'ignite.metrics.fbeta', 'ignite.distributed.comp_models.native', 'rich.themes', 'scipy._lib.array_api_compat.array_api_compat', 'tensorflow.python.autograph.lang.directives', 'keras.src.applications.vgg16', 'tensorflow.core.protobuf.debug_event_pb2', 'tensorflow.python.data.experimental.ops.take_while_ops', 'tensorflow._api.v2.compat.v1.types.experimental', 'tensorflow.python.distribute.v1.input_lib', 'tensorflow.core.function.polymorphism.function_type_pb2', 'tensorflow.python.saved_model.tag_constants', 'tensorflow._api.v2.__internal__.types.data', 'tensorflow.dtensor.python.input_util', 'absl', 'attr.validators', 'tensorflow._api.v2.__internal__.distribute', 'tensorflow.python.tpu.ops.gen_xla_ops', 'tensorflow.python.ops.signal.reconstruction_ops', 'keras.src.layers.convolutional.base_depthwise_conv', 'scipy._lib.array_api_compat.array_api_compat.common._linalg', 'tensorflow.python.summary.plugin_asset', 'ignite.exceptions', 'tensorflow._api.v2.compat.v1.nn.rnn_cell', 'tensorflow.core.function.trace_type', 'tensorflow.python.client._pywrap_debug_events_writer', 'tensorflow._api.v2.__internal__.nest', 'keras.src.legacy.saving.legacy_h5_format', 'absl.flags._helpers', 'pandas.core.arraylike', 'tensorflow.python.data.ops.from_tensor_slices_op', 'tensorflow.security', 'tensorflow.python.ops.linalg.linear_operator_lower_triangular', 'tensorflow.compiler.mlir.quantization.stablehlo.quantization_config_pb2', 'tensorflow.python.saved_model.load_v1_in_v2', 'tensorflow.python.data.experimental.ops.lookup_ops', 'pandas.io.json._table_schema', 'tensorflow.python.grappler.cluster', 'tensorflow._api.v2.compat.v1.__internal__.types.data', 'pandas.io.parsers.readers', 'tensorflow.python.ops.state_ops', 'scipy.sparse.csgraph._laplacian', 'tensorflow.python.ops.gen_resource_variable_ops', 'tensorflow.python.types.doc_typealias', 'keras.src.layers.normalization', 'pandas.core.internals.blocks', 'tensorflow.python.saved_model.utils', 'tensorflow.compiler.mlir.quantization.stablehlo', 'keras.src.backend.tensorflow.trackable', 'tensorflow.python.ops.default_gradient', 'tensorflow._api.v2.compat.v2.version', 'tensorflow.python.ops.signal.window_ops', 'cv2.version', 'keras.src.optimizers.adamax', 'tensorflow.core.protobuf.tensorflow_server_pb2', 'tensorflow.experimental', 'tensorflow.python.ops.rnn_cell_impl', 'pandas._libs.lib', 'tensorflow.python.ops.signal.shape_ops', 'tensorflow.python.distribute.experimental', 'namex', 'tensorflow.compiler.tf2xla', 'tensorflow.python.debug.wrappers.hooks', 'pandas._libs', 'tensorflow.core.util.test_log_pb2', 'tensorflow.python.framework._op_def_library_pybind', 'tensorflow.python.ops.gen_parsing_ops', 'tensorflow.python.tpu.client', 'tensorflow.python.ops.distributions.dirichlet', 'tensorflow._api.v2.graph_util', 'pandas.core.computation.expressions', 'tensorflow._api.v2.errors', 'cityscapesscripts', 'tensorflow.core.grappler.costs.op_performance_data_pb2', 'attr._cmp', 'ignite.metrics.roc_auc', 'tensorflow.python.eager.polymorphic_function', 'rich.repr', 'tensorflow.python.ops.state_grad', 'keras.api.layers', 'tensorflow.dtensor.python.config', 'tensorflow.lite.python.interpreter_wrapper', 'keras.src.datasets.reuters', 'pandas.io.html', 'tensorflow.python.distribute.failure_handling.failure_handling', 'tensorflow.python.training.moving_averages', 'pandas.core.arrays.numeric', 'tensorflow.python.ops.control_flow_util_v2', 'tensorflow.python.ops.ragged.ragged_squeeze_op', 'tensorflow.python.data.experimental.ops.snapshot', 'h5py.h5py_warnings', 'tensorflow.core.framework.op_def_pb2', 'keras.src.layers.merging.dot', 'tensorflow.python.data.experimental.ops.counter', 'keras.src.api_export', 'keras.src.layers.preprocessing.audio_preprocessing', 'tensorflow.python.grappler.tf_optimizer', 'tensorflow.python.ops.linalg.sparse.sparse_csr_matrix_ops', 'tensorflow.compiler.mlir.lite.debug.debug_options_pb2', 'keras.src.distribution.distribution_lib', 'tensorflow.python.ops', 'keras.api.preprocessing', 'cv2.barcode', 'tensorflow.python.keras.saving.saved_model.load', 'scipy.sparse.bsr', 'rich.terminal_theme', 'pandas.plotting._core', 'pandas._testing.compat', 'scipy.sparse.linalg._dsolve.linsolve', 'tensorflow.python.ops.distributions', 'six.moves.urllib', 'keras.api.datasets.imdb', 'ignite.metrics.frequency', 'tensorflow.python.saved_model.utils_impl', 'ignite.metrics.average_precision', 'src.logger', 'keras.src.utils.jax_utils', 'keras.src.trainers', 'rich._extension', 'tensorflow.python.distribute.coordinator.metric_utils', 'pandas.core.indexes.category', 'tensorflow.python.saved_model.pywrap_saved_model.metrics', 'tensorflow.python.ops.distributions.beta', 'tensorflow.python.util._pywrap_checkpoint_reader', 'scipy.sparse.linalg._special_sparse_arrays', 'tensorflow.python.distribute.multi_process_runner', 'scipy.sparse.linalg._dsolve', 'pandas.core.tools', 'tensorflow._api.v2.compat.v2.nest', 'gast.astn', 'keras.api.optimizers', 'tensorflow.python.ops.partitioned_variables', 'scipy._lib._array_api', 'tensorflow.python.tpu.tpu_optimizer', 'tensorflow.compiler.mlir.quantization.stablehlo.quantization_options_pb2', 'keras.src.layers.preprocessing.discretization', 'scipy.sparse._coo', 'optree.utils', 'tensorflow.python.saved_model.loader', 'rich.constrain', 'tensorflow.python.distribute.cross_device_utils', 'pandas.tseries', 'tensorflow.lite.python.lite', 'tensorflow.python.keras.optimizer_v2.nadam', 'pandas.core.interchange.dataframe_protocol', 'ignite.metrics.nlp.rouge', 'tensorflow.python.data.experimental.ops.distributed_save_op', 'ignite.base.mixins', 'tensorflow.python.distribute.values_util', 'tensorflow.python.user_ops.ops', 'pandas.api', 'scipy.linalg.decomp_cholesky', 'cv2.data', 'tensorflow.compiler', 'tensorflow.python.saved_model.builder', 'tensorflow.python.tpu.tpu_replication', 'tensorflow._api.v2.compat.v2.__internal__.dispatch', 'ignite.handlers.terminate_on_nan', 'tensorflow._api.v2.compat.v1.summary', 'ignite.metrics.metric', 'tensorflow.python.framework.func_graph', 'ignite.handlers.tqdm_logger', 'pandas._config.display', 'ignite.engine.utils', 'keras.src.layers.rnn.conv_lstm', 'pandas.core.computation.engines', 'xml.sax.saxutils', 'keras.src.layers.rnn.rnn', 'tensorflow.python.ops.distributions.util', 'keras.src.layers.reshaping.flatten', 'keras.src.applications.vgg19', 'tensorflow.distribute.experimental', 'keras.src.layers.normalization.spectral_normalization', 'src.models.model_one_modality', 'pandas._libs.tslibs.timezones', 'pandas.core.computation.scope', 'tensorflow.python.ops.manip_grad', 'tensorflow.python.keras.engine.input_spec', 'tensorflow.python.saved_model.path_helpers', 'keras.src.layers.core.einsum_dense', 'pandas.core.computation.api', 'tensorflow._api.v2.compat.v1.tpu', 'pandas.io.excel._openpyxl', 'opt_einsum.sharing', 'tensorflow.python.distribute.packed_distributed_variable', 'tensorflow.core.framework.graph_pb2', 'tensorflow.lite.profiling.proto.profiling_info_pb2', 'tensorflow.python.compiler.xla.jit', 'ignite.distributed', 'keras.src.backend.tensorflow.tensorboard', 'keras.src.layers.preprocessing.text_vectorization', 'tensorflow._api.v2.__internal__.test', 'ignite.metrics.running_average', 'tensorflow.python.distribute.ps_values', 'cv2.detail', 'ignite.engine.deterministic', 'tensorflow.core.lib.core', 'tensorflow.python.debug.lib.debug_gradients', 'pandas._libs.internals', 'tensorflow.core.function.polymorphism', 'tensorflow.python.data.experimental.ops.iterator_ops', 'tensorflow.python.ops.confusion_matrix', 'ignite.handlers.neptune_logger', 'pandas._testing.contexts', 'keras.src.layers.reshaping.cropping1d', 'tensorflow.python.pywrap_tensorflow', 'pandas.core.arrays.arrow.accessors', 'tensorflow._api.v2.compat.v2.summary', 'tensorflow._api.v2.compat.v1.compat', 'tensorflow.python.ops.linalg.sparse.gen_sparse_csr_matrix_ops', 'tensorflow.python.debug.cli.command_parser', 'rich.jupyter', 'tensorflow._api.v2.compat.v2.distribute.experimental', 'tensorflow.python.ops.distributions.student_t', 'src.datasets.cityscapes.pytorch_dataset', 'tensorflow.python.data.util.traverse', 'keras.src.utils.jax_layer', 'tensorflow._api.v2.compat.v2.xla', 'pandas.core.interchange.from_dataframe', 'src.datasets', 'tensorflow.python.distribute.cross_device_ops', 'tensorflow.python.keras.engine', 'keras.src.layers.pooling.average_pooling1d', 'tensorflow.python.training.checkpoint_utils', 'torch.distributed.optim.functional_adam', 'h5py._hl.files', 'pandas.io.json._json', 'keras.src.datasets.cifar10', 'pandas.core.arrays.string_arrow', 'pandas._config.config', 'tensorflow.python.keras.utils.io_utils', 'numpy.testing', 'src.models.rgb_depth_fusion', 'tensorflow.python.autograph.pyct', 'wrapt.decorators', 'tensorflow.python.distribute.parameter_server_strategy_v2', 'tensorflow.core.framework.tensor_description_pb2', 'ignite.handlers.early_stopping', 'tensorflow._api.v2.__internal__.smart_cond', 'ignite.handlers.stores', 'tensorflow.python.keras.engine.base_layer_utils', 'attr._make', 'pytz.lazy', 'ignite.metrics.nlp.bleu', 'tensorflow.core.function.trace_type.serialization', 'keras.src.layers.pooling', 'pandas.core.computation.expr', 'pandas.io.parsers.arrow_parser_wrapper', 'tensorflow.compiler.mlir.quantization.tensorflow.python.pywrap_function_lib', 'pandas.io.excel._calamine', 'tensorflow.python.ops.numpy_ops', 'tensorflow.python.util.function_utils', 'pandas.core.reshape.merge', 'tensorflow.python.keras.constraints', 'keras.api.initializers', 'numpy.testing._private.utils', 'pandas.io.formats.console', 'tensorflow.lite.experimental.microfrontend.ops', 'tensorflow.python.distribute.tpu_strategy', 'tensorflow.python.distribute.failure_handling.failure_handling_util', 'tensorflow._api.v2.autograph.experimental', 'keras.src.layers.reshaping.up_sampling3d', 'keras.src.layers.reshaping.up_sampling1d', 'pandas.io.sas.sasreader', 'tensorflow.tools.docs', 'ignite.handlers.utils', 'pandas.core.tools.datetimes', 'tensorflow.python.ops.linalg.linear_operator_full_matrix', 'tensorflow.python.autograph.converters.directives', 'pandas.io.excel._odfreader', 'torch.distributed.algorithms.model_averaging.averagers', 'tensorflow.python.keras.distribute.worker_training_state', 'keras.src.datasets.boston_housing', 'tensorflow.python.keras.layers.convolutional', 'tensorflow.lite.python.analyzer_wrapper._pywrap_analyzer_wrapper', 'tensorflow.python.ops.nn', 'tensorflow._api.v2.compat.v2.math.special', 'pandas.core.window.ewm', 'zoneinfo', 'tensorflow.python.eager.tape', 'tensorflow.python.training.experimental.loss_scale', 'src.datasets.dataset_base', 'tensorflow.python.autograph.pyct.static_analysis.annos', 'tensorflow._api.v2.compat.v2.saved_model', 'tensorflow.python.distribute.combinations', 'pandas.core.groupby.ops', 'tensorflow.python.ops.signal.dct_ops', 'h5py.version', 'tensorflow.python.client.session', 'tensorflow.python.ops.lookup_ops', 'ignite.metrics.regression.canberra_metric', 'tensorflow.python.data', 'keras.src.layers.reshaping.reshape', 'tensorflow.python.keras.distribute.distribute_coordinator_utils', 'tensorflow.python.keras.callbacks', 'keras.src.layers.pooling.base_pooling', 'pandas._testing.asserters', 'tensorflow.python.saved_model.registration', 'tensorflow.python.keras.saving.save', 'tensorflow.python.keras.optimizer_v2.optimizer_v2', 'tensorflow._api.v2.summary', 'keras.src.layers.core.identity', 'tensorflow.core.profiler.tfprof_options_pb2', 'scipy.sparse.linalg._isolve.lgmres', 'tensorflow._api.v2.compat.v1.saved_model.signature_def_utils', 'xml.sax._exceptions', 'keras.src.tree.tree_api', 'tensorflow.python.util.protobuf', 'tensorflow.python._pywrap_mlir', 'tensorflow.python.profiler.profiler_client', 'cv2.ogl', 'tensorflow.python.ops.gen_special_math_ops', 'termcolor._types', 'tensorflow.core.protobuf.device_properties_pb2', 'tensorflow.python.framework.errors', 'absl.testing.parameterized', 'pandas._libs.window', 'tensorflow.python.saved_model.pywrap_saved_model.fingerprinting', 'tensorflow.python.framework.op_def_library', 'tensorflow.python.ops.control_flow_util', 'tensorflow.python.ops.gen_ragged_conversion_ops', 'tensorflow.core.function.polymorphism.type_dispatch', 'pandas.core.reshape.util', 'tensorflow.lite.python.metrics.converter_error_data_pb2', 'tensorflow.python.autograph.impl.api', 'ignite.metrics.js_divergence', 'pandas.core.generic', 'tensorflow._api.v2.lite.experimental.authoring', 'tensorflow.python.training.experimental.mixed_precision', 'tensorflow.python.tpu.tpu_embedding_v3_checkpoint_adapter', 'tensorflow._api.v2.__internal__.eager_context', 'tensorflow.python.ops.batch_ops', 'scipy.linalg._cythonized_array_utils', 'scipy.sparse.linalg._eigen.lobpcg.lobpcg', 'h5py._hl.attrs', 'keras.src.backend.common.backend_utils', 'tensorflow.python.lib.io.tf_record', 'tensorflow.python.framework.convert_to_constants', 'tensorflow.compiler.tf2xla.ops.gen_xla_ops', '_cython_3_0_10', 'tensorflow.python.autograph.converters.conditional_expressions', 'tensorflow._api.v2.compat.v2.__internal__.nest', 'keras.src.backend.tensorflow.math', 'tensorflow.python.eager.context', '_zoneinfo', 'keras.src.datasets.imdb', 'ignite.metrics.accumulation', 'scipy._lib._ccallback', 'tensorflow._api.v2.compat.v2.__internal__.types', 'tensorflow.python.ops.distributions.distributions', 'tensorflow.python.lib.io', 'tensorflow._api.v2.compat.v2.quantization.experimental', 'tensorflow.python.layers', 'tensorflow._api.v2.compat.v1.config.experimental', 'keras.src.layers.pooling.global_max_pooling3d', 'tensorflow._api.v2.types.experimental', 'keras.src.optimizers.sgd', 'tensorflow.python.distribute.coordinator.values', 'pandas.core.array_algos.masked_accumulations', 'pandas.errors', 'torch.distributed.optim.functional_rmsprop', 'tensorflow.python.saved_model.pywrap_saved_model.constants', 'ignite.metrics.maximum_mean_discrepancy', 'tensorflow.python.keras.legacy_tf_layers.base', 'gast', 'keras.src.layers.regularization.dropout', 'pandas.core.array_algos.transforms', 'keras.api.metrics', 'tensorflow._api.v2.compat.v2.errors', 'scipy.sparse.csgraph._min_spanning_tree', 'rich.text', 'tensorflow._api.v2.distribute.experimental.coordinator', 'tensorflow._api.v2.__internal__.distribute.combinations', 'tensorflow._api.v2.compat.v1.compat.v1', 'tensorflow.python.saved_model.pywrap_saved_model', 'tensorflow._api.v2.compat.v1.queue', 'pandas.core.arrays.interval', 'tensorflow._api.v2.compat.v2.data.experimental', 'tensorflow._api.v2.compat.v2.__internal__.ops', 'tensorflow.python.ops.shape_util', 'tensorflow._api.v2.compat.v2.ragged', 'pandas.core.indexers.objects', 'tensorflow.lite.python.authoring', 'tensorflow.python.keras.utils.conv_utils', 'tensorflow.python.eager.graph_only_ops', 'tensorflow._api.v2.compat.v2.debugging', 'pytz', 'tensorflow.python.distribute.input_util', 'tensorflow._api.v2.compat.v2.experimental.numpy', 'tensorflow.python.ops.parsing_ops', 'tensorflow.python.tpu.tpu_sharding', 'tensorflow.dtensor.python.api', 'cv2.gapi.wip.draw', 'keras.src.ops.nn', 'tensorflow.python.ops.candidate_sampling_ops', 'src.datasets.nyuv2', 'ignite.metrics.regression.fractional_bias', 'tensorflow._api.v2.compat.v1.saved_model.constants', 'tensorflow.python.framework.function', 'tensorflow._api.v2.tpu.experimental.embedding', 'tensorflow.tools.docs.doc_controls', 'keras.src', 'ignite.metrics.regression.fractional_absolute_error', 'torch.distributed.optim.optimizer', 'keras.src.models.model', 'pandas.compat.compressors', 'scipy.linalg.decomp_lu', 'attr._next_gen', 'keras.api.applications', 'pandas.core.computation.ops', 'tensorflow._api.v2.compat.v1.experimental.extension_type', 'ignite.metrics.regression.manhattan_distance', 'tensorflow.python.ops.numpy_ops.np_arrays', 'tensorflow._api.v2.compat.v2.tpu.experimental.embedding', 'ignite.engine', 'tensorflow.python.ops.image_ops_impl', 'h5py.h5o', 'keras.src.layers.convolutional.conv2d', 'tensorflow.python.eager.forwardprop', 'tensorflow.python.tpu.tpu_function', 'tensorflow._api.v2.compat.v1.strings', 'tensorflow.python.ops.numpy_ops.np_utils', 'rich.control', 'keras.src.quantizers.quantizers', 'tensorflow.python.debug.cli.tensor_format', 'keras.src.utils.torch_utils', 'tensorflow._api.v2.compat.v1.io.gfile', 'keras.src.callbacks.lambda_callback', 'tensorflow._api.v2.sysconfig', 'scipy.sparse._sparsetools', 'tensorflow.python.ops.control_flow_grad', 'tensorflow._api.v2.compat.v2.profiler.experimental', 'keras.api.preprocessing.image', 'tensorflow.python.ops.weak_tensor_ops', 'tensorflow.tsl.protobuf.rpc_options_pb2', 'pandas.core.ops.invalid', 'tensorflow.python.distribute.v1', 'tensorflow.distribute.experimental.rpc.proto.tf_rpc_service_pb2', 'pandas.io.common', 'pandas.io.parsers.python_parser', 'tensorflow.python.ops.ragged', 'tensorflow.python.user_ops', 'tensorflow._api.v2.compat.v1.quantization', 'keras.api.tree', 'pandas.core.methods.describe', 'rich.color', 'tensorflow.lite.python.op_hint', 'tensorflow.python.platform._pywrap_tf2', 'scipy.sparse.csgraph._matching', 'scipy._lib._testutils', 'pandas.core.dtypes.common', 'ignite.distributed.launcher', 'tensorflow.python.ops.gen_ctc_ops', 'tensorflow.python.util.tf_inspect', 'pandas.core.array_algos', 'tensorflow.python.training.server_lib', 'tensorflow._api.v2.compat.v1.audio', 'tensorflow.python.data.ops.dataset_autograph', 'tensorflow.core.protobuf.debug_pb2', 'keras.src.layers.regularization.activity_regularization', 'tensorflow.python.ops.signal.fft_ops', 'keras.src.utils.naming', 'tensorflow.python.ops.distributions.categorical', 'keras.src.models.cloning', 'tensorflow.python.ops.autograph_ops', 'tensorflow.python.training.experimental', 'keras.src.dtype_policies.dtype_policy', 'tensorflow.python.ops.list_ops', 'tensorflow._api.v2.summary.experimental', 'tensorflow.python.ops.distributions.exponential', 'tensorflow.python.eager.wrap_function', 'keras.src.utils.module_utils', 'tensorflow.core.protobuf.fingerprint_pb2', 'tensorflow.lite.tools.optimize', 'tensorflow._api.v2.ragged', 'repositories', 'keras.src.datasets', 'pandas.core.arrays._utils', 'keras.src.optimizers', 'tensorflow.python.framework.sparse_tensor', 'src.models.model_utils', 'tensorflow.python.keras.layers.legacy_rnn.rnn_cell_impl', 'ignite.distributed.comp_models.xla', 'namex.generate', 'pandas.core.methods.selectn', 'pandas.core.api', 'tensorflow.python.ops.linalg_ops_impl', 'keras.src.layers.preprocessing.random_brightness', 'tensorflow.python.ops.standard_ops', 'tensorflow._api.v2.compat.v2.experimental.numpy.random', 'tensorflow.core.function', 'tensorflow.python.keras.layers.advanced_activations', 'pandas.core.dtypes.missing', 'tensorflow._api.v2.compat.v1.errors', 'scipy', 'pandas.core.indexing', 'attr._config', 'pandas.core.indexes.multi', 'keras.src.metrics.reduction_metrics', 'pandas.io.formats.printing', 'keras.src.losses.loss', 'tensorflow.python.client.timeline', 'tensorflow.python.training.py_checkpoint_reader', 'tensorflow.python.checkpoint.util', 'pandas.core.reshape.encoding', 'tensorflow.python.ops.gen_experimental_dataset_ops', 'torch.distributed.optim.functional_adamax', 'cv2.gapi.core.fluid', 'tensorflow.compiler.xla.service.hlo_pb2', 'tensorflow._api.v2.compat.v1.math.special', 'absl.flags', 'src.datasets.scenenetrgbd', 'tensorflow.python.ops.linalg.linear_operator_composition', 'flatbuffers', 'tensorflow.python.keras.optimizer_v2.adam', 'pandas.core.dtypes', 'keras.src.losses.losses', 'scipy.sparse._sputils', 'h5py.h5l', 'tensorflow.python.tpu.tensor_tracer_flags', 'rich._export_format', 'rich._palettes', 'pandas.core.arrays.period', 'tensorflow.python.ops.linalg.linear_operator_zeros', 'scipy.sparse._construct', 'keras.api.optimizers.schedules', 'tensorflow._api.v2.distribute', 'tensorflow.python._pywrap_tfe', 'tensorflow.core.example.example_pb2', 'tensorflow.python.training.checkpoint_state_pb2', 'tensorflow._api.v2.compat.v1.distributions', 'tensorflow.lite.python.optimize.calibrator', 'torch.distributed.optim.zero_redundancy_optimizer', 'tensorflow.python.data.experimental.ops', 'pandas.compat._constants', 'tensorflow.python.dlpack.dlpack', 'tensorflow.python.keras.utils.version_utils', 'tensorflow._api.v2.compat.v2.saved_model.experimental', 'tensorflow.python.data.experimental.service._pywrap_utils_exp', 'tensorflow.python.framework.traceable_stack', 'tensorflow._api.v2.compat.v1.io', 'keras.api.preprocessing.sequence', 'scipy.sparse.linalg._eigen._svds', 'tensorflow.compiler.mlir.lite.python._pywrap_converter_api', 'rich.protocol', 'ignite.handlers.time_limit', 'tensorflow.python.util.nest', 'tensorflow.python.keras.distribute', 'tensorflow.python.platform.remote_utils', 'keras.src.layers.pooling.max_pooling2d', 'tensorflow.python.ops.lookup_grad', 'tensorflow.python.ops.summary_ops_v2', 'keras.src.utils.io_utils', 'pandas.compat._optional', 'tensorflow._api.v2.compat.v2.tpu.experimental', 'keras.src.backend.common.stateless_scope', 'tensorflow.python.data.experimental.ops.unique', 'tensorflow.python.keras.distribute.distributed_training_utils_v1', 'pandas.core.indexes.datetimes', 'tensorflow._api.v2.compat.v2.__internal__.autograph', 'tensorflow.python.keras.protobuf.versions_pb2', 'keras.src.backend.tensorflow.core', 'tensorflow.python.keras.layers.legacy_rnn', 'tensorflow.python.framework.strict_mode', 'scipy._lib.array_api_compat.array_api_compat.numpy.linalg', 'tensorflow.python.eager.polymorphic_function.tracing_compilation', 'tensorflow.python.distribute.one_device_strategy', 'tensorflow._api.v2.compat.v2.linalg.experimental', 'tensorflow.compiler.mlir.lite.python', 'tensorflow.python.ops.distributions.bernoulli', 'pandas.core._numba.executor', 'keras.api.dtype_policies', 'tensorflow.python.autograph.converters.asserts', 'tensorflow.python.autograph.converters.functions', 'optree.functools', 'keras.src.ops.operation', 'cv2.mat_wrapper', 'scipy.sparse.linalg._isolve.utils', 'tensorflow.python.distribute.cluster_resolver.kubernetes_cluster_resolver', 'tensorflow.python.ops.numpy_ops.np_math_ops', 'tensorflow.python.keras.saving.saved_model.load_context', 'scipy.sparse.linalg.eigen', 'tensorflow._api.v2.__internal__.monitoring', 'tensorflow.python.ops.gen_functional_ops', 'torch.distributed.optim.functional_rprop', 'pybind11_abseil.status', 'tensorflow.python.ops.tensor_getitem_override', 'h5py._hl.selections', 'pandas._libs.pandas_datetime', 'tensorflow.python.ops.sparse_grad', 'tensorflow.python.platform', 'tensorflow.python.tpu.tpu_hardware_feature', 'tensorflow.python.util.tf_export', 'pandas._libs.algos', 'tensorflow.dtensor.python.tpu_util', 'tensorflow.python.ops.ragged.ragged_config', 'ignite.metrics.regression.median_relative_absolute_error', 'cv2.gapi.imgproc', 'pandas.core.sample', 'scipy.linalg.misc', 'keras.src.layers.pooling.global_average_pooling3d', 'tensorflow.python.distribute.test_util', 'keras.src.backend.tensorflow.optimizer', 'tensorflow.python.distribute.failure_handling.preemption_watcher', 'ignite.distributed.comp_models', 'tensorflow.python.keras.mixed_precision', 'pandas.plotting', '_cython_3_0_5', 'rich.pretty', 'tensorflow._api.v2.compat.v1.quantization.experimental', 'tensorflow.python.keras.mixed_precision.autocast_variable', 'keras.src.utils.timeseries_dataset_utils', 'tensorflow.python.framework.importer', 'tensorflow.python.framework.composite_tensor', 'pandas.core.reshape.tile', 'tensorflow.python.client', 'tensorflow.core.protobuf.rpc_options_pb2', 'tensorflow.python.util.tf_decorator_export', 'keras.src.layers.core.embedding', 'tensorflow.python.training.saver', 'pandas.core.window.doc', 'tensorflow.python.keras.utils', 'tensorflow.core.protobuf.saved_model_pb2', 'tensorflow._api.v2.compat.v1.raw_ops', 'keras.src.applications.xception', 'tensorflow.core.protobuf.service_config_pb2', 'tensorflow.lite.experimental.microfrontend.python.ops', 'tensorflow.python.distribute.distribute_utils', 'tensorflow.python.data.ops.multi_device_iterator_ops', 'tensorflow.core.framework', 'tensorflow.python.distribute.cluster_resolver.tpu.tpu_cluster_resolver', 'scipy.linalg.special_matrices', 'tensorflow.python.keras.utils.mode_keys', 'pandas.compat', 'tensorflow.python.ops.gen_lookup_ops', 'tensorflow.python.ops.ragged.ragged_ops', 'tensorflow.python.ops.variable_v1', 'ignite.distributed.comp_models.base', 'tensorflow.python.checkpoint.checkpoint_context', 'keras.api.activations', 'pandas._libs.interval', 'tensorflow._api.v2.version', 'keras.api.datasets', 'tensorflow.python.ops.parsing_config', 'pandas.core.internals.concat', 'keras.src.callbacks.backup_and_restore', 'tensorflow.python.saved_model.revived_types', 'tensorflow.python.ops.numpy_ops.np_config', 'tensorflow.python.ops.sort_ops', 'opt_einsum.backends.theano', 'tensorflow._api.v2.audio', 'tensorflow._api.v2.compat.v1.signal', 'pandas.core.internals.managers', 'src.datasets.cityscapes', 'keras.src.trainers.data_adapters.py_dataset_adapter', 'tensorflow.python.ops.control_flow_ops', 'tensorflow.python.keras.optimizers', 'tensorflow._api.v2.compat.v1.random', 'tensorflow.python.distribute.reduce_util', 'tensorflow._api.v2.quantization.experimental', 'pandas.core.indexes.frozen', 'keras.api.applications.xception', 'tensorflow.python.framework.indexed_slices', 'tensorflow._api.v2.compat.v1.test.experimental', 'pandas.core.arrays', 'tensorflow.python.keras.engine.partial_batch_padding_handler', 'rich.cells', 'h5py._objects', 'tensorflow.python.ops.histogram_ops', 'tensorflow.lite.experimental.microfrontend.python.ops.audio_microfrontend_op', 'keras.src.layers.attention', 'tensorflow._api.v2.__internal__.ops', 'torch.distributed.optim.functional_adadelta', 'tensorflow._api.v2.xla', 'ignite.handlers.clearml_logger', 'tensorflow.python.data.ops.readers', 'tensorflow._api.v2.__internal__.tf2', 'tensorflow._api.v2.compat.v2.sysconfig', 'tensorflow.python.ops.sets_impl', 'keras.api.mixed_precision', 'tensorflow._api.v2.compat.v1.app', 'tensorflow._api.v2.compat.v2.nn.experimental', 'keras.src.backend.config', 'tensorflow.python.keras.saving.saved_model.json_utils', 'pandas.core.arrays.integer', 'tensorflow.python.training.summary_io', 'keras.src.layers.merging.subtract', 'keras.api.utils.legacy', 'keras.src.ops.function', 'flatbuffers.compat', 'cv2.segmentation', 'tensorflow._api.v2.compat.v2.__internal__.saved_model.load', 'tensorflow.core.util.event_pb2', 'ignite.engine.engine', 'rich._emoji_replace', 'tensorflow._api.v2.compat.v1.compat.v2', 'keras.src.utils.image_utils', 'keras.src.models', 'tensorflow.python.checkpoint.tensor_callable', 'pandas.core.groupby.indexing', 'pandas._libs.tslibs.base', 'tensorflow.python.ops.signal.spectral_ops', 'keras.src.ops.core', 'pandas.io.spss', 'keras.src.callbacks.early_stopping', 'keras.src.optimizers.base_optimizer', 'tensorflow.python.eager.monitoring', 'tensorflow.python.tpu.tensor_tracer_report', 'tensorflow._api.v2.compat.v2.__internal__.test', 'tensorflow.python.data.experimental.ops.shuffle_ops', 'tensorflow.python.framework._pywrap_python_op_gen', 'keras.src.distribution', 'tensorflow.python.framework.device_spec', 'tensorflow._api.v2.compat.v2.profiler', 'scipy.linalg._flapack', 'keras.src.backend.tensorflow.sparse', 'pandas.core.window', 'tensorflow.python.framework.combinations', 'pandas.api.types', 'opt_einsum.blas', 'pandas.core.interchange', 'pandas._libs.tslibs.conversion', 'tensorflow.core.framework.summary_pb2', 'tensorflow.core.protobuf', 'tensorflow.compiler.xla.service', 'pandas.util.version', 'keras.src.metrics.confusion_metrics', 'tensorflow._api.v2.autodiff', 'keras.src.initializers.constant_initializers', 'tensorflow.python.keras.protobuf', 'tensorflow._api.v2.compat.v2.autograph', 'scipy.linalg.basic', 'keras.src.layers.core.masking', 'keras.api.losses', 'pandas._libs.window.aggregations', 'tensorflow.python.ops.gen_candidate_sampling_ops', 'cv2.gapi.own', 'pytz.tzinfo', 'keras.src.layers.merging.base_merge', 'keras.src.applications.convnext', 'ignite.metrics.mutual_information', 'tensorflow.python.training.proximal_adagrad', 'tensorflow.python.keras.activations', 'pandas.compat.numpy', 'tensorflow.python.compiler.mlir', 'opt_einsum.backends.object_arrays', 'tensorflow.compiler.jit', 'tensorflow.python.framework._test_metrics_util', 'attr', 'tensorflow.python.ops.summary_op_util', 'h5py.h5a', 'pandas.core.dtypes.base', 'pandas.core.arrays.sparse.accessor', 'pandas._version_meson', 'keras.src.initializers.random_initializers', 'h5py.h5fd', 'h5py._hl.compat', 'cv2.fisheye', 'pandas._libs.parsers', 'tensorflow._api.v2.linalg', 'rich.pager', 'tensorflow.python.tpu.feature_column', 'keras.src.layers.activations.prelu', 'tensorflow.python.ops.gen_random_ops', 'cv2.videoio_registry', 'tensorflow.python._pywrap_dtensor_device', 'tensorflow._api.v2.compat.v1.dtypes.experimental', 'tensorflow.python.distribute.multi_worker_test_base', 'absl.app', 'ignite.metrics.metric_group', 'tensorflow.core.protobuf.device_filters_pb2', 'tensorflow.python.ops.gen_summary_ops', 'tensorflow.core.protobuf.queue_runner_pb2', 'keras.src.utils.dataset_utils', 'keras.src.optimizers.adam', 'tensorflow.core.protobuf.snapshot_pb2', 'keras.src.layers.preprocessing.random_rotation', 'tensorflow.core.protobuf.saved_object_graph_pb2', 'keras.src.callbacks.swap_ema_weights', 'tensorflow.python.ops.gen_tpu_partition_ops', 'scipy._lib.decorator', 'tensorflow.lite.python.metrics.metrics', 'tensorflow.python.training.slot_creator', 'tensorflow._api.v2.compat.v1.python_io', 'tensorflow.python.ops.op_selector', 'tensorflow.python.ops.nn_impl_distribute', 'keras.src.regularizers', 'tensorflow._api.v2.compat.v2.profiler.experimental.client', 'tensorflow.python.proto_exports', 'tensorflow._api.v2.compat.v1.dtypes', 'tensorflow.python.debug.wrappers.dumping_wrapper', 'tensorflow.compiler.tf2tensorrt.ops.gen_trt_ops', 'keras.src.dtype_policies.dtype_policy_map', 'tensorflow.python.ops.resource_variables_toggle', 'tensorflow.python.eager.polymorphic_function.tf_method_target', 'tensorflow.python.platform.sysconfig', 'tensorflow._api.v2.compat.v2.lite', 'tensorflow.python.keras.distribute.distributed_training_utils', 'keras.src.backend.tensorflow.random', 'tensorflow.python.distribute.experimental.mirrored_strategy', 'tensorflow.python.tpu.tpu_embedding_v2', 'keras.src.utils.python_utils', 'tensorflow.python.keras.regularizers', 'tensorflow.python.framework.load_library', 'tensorflow.core.function.trace_type.default_types', 'keras.api.applications.mobilenet', 'tensorflow.python.profiler.internal._pywrap_profiler', 'tensorflow.python.ops.ragged.ragged_string_ops', 'tensorflow.python.keras.protobuf.saved_metadata_pb2', 'tensorflow.core.protobuf.verifier_config_pb2', 'cv2.load_config_py3', 'pandas.core.computation.eval', 'tensorflow._api.v2.__internal__', 'tensorflow.python.platform.device_context', 'tensorflow.python.ops.gen_image_ops', 'torch.distributed.optim.functional_sgd', 'tensorflow.python.data.util.options', 'tensorflow.python.eager.polymorphic_function.function_type_utils', 'tensorflow.python.debug.cli.debugger_cli_common', 'tensorflow.python._pywrap_tensorflow_internal', 'absl.logging.converter', 'pandas.core.ops.dispatch', 'tensorflow.python.ops.cudnn_rnn_grad', 'cv2.gapi.ie', 'tensorflow.python.keras.saving.saved_model.metric_serialization', 'keras.src.legacy.saving.saving_utils', 'tensorflow.python.tpu.tpu_embedding_for_serving', 'tensorflow.python.keras.mixed_precision.loss_scale_optimizer', 'tensorflow.core.distributed_runtime.preemption.gen_check_preemption_op', 'keras.src.optimizers.ftrl', 'tensorflow.python.ops.math_grad', 'tensorflow.python.summary.tb_summary', 'tensorflow.python.keras.saving', 'tensorflow.python.ops.session_ops', 'scipy.linalg._decomp_qz', 'tensorflow.python.keras.utils.data_utils', 'tensorflow.compiler.mlir.lite.debug', 'tensorflow.compiler.jit.ops.xla_ops', 'tensorflow.python.keras.optimizer_v2.rmsprop', 'keras.src.trainers.data_adapters.data_adapter_utils', 'scipy.sparse.base', 'opt_einsum.backends.cupy', 'tensorflow.python.keras.optimizer_v2.ftrl', 'tensorflow.python.keras.layers.merge', 'tensorflow.python.framework.tensor_shape', 'tensorflow.python.debug.cli.evaluator', 'tensorflow.python.training.rmsprop', 'tensorflow.python.keras.saving.saved_model', 'keras.src.layers.merging.average', 'tensorflow.python.ops.tensor_array_ops', 'tensorflow.python.ops.functional_ops', 'tensorflow._api.v2.experimental.dtensor', 'tensorflow.python.ops.array_ops_stack', 'tensorflow.python.framework.tensor_util', 'tensorflow.core.protobuf.data_service_pb2', 'keras.src.saving.object_registration', 'tensorflow.python.framework.errors_impl', 'ml_dtypes._ml_dtypes_ext', 'tensorflow._api.v2.__internal__.saved_model', 'tensorflow.python.util._pywrap_determinism', 'tensorflow.tools.compatibility.renames_v2', 'tensorflow.lite.python.metrics', 'tensorflow.python._pywrap_parallel_device', 'tensorflow.python.distribute.numpy_dataset', 'tensorflow.python.data.ops.debug_mode', 'cv2.gapi.ie.detail', 'tensorflow.python.compiler.mlir.mlir', 'tensorflow.python.autograph.pyct.transpiler', 'tensorflow._api.v2.compat.v2.distribute.experimental.partitioners', 'tensorflow.python.data.experimental.ops.writers', 'tensorflow.python.platform.flags', 'tensorflow.python.ops.ctc_ops', 'scipy.sparse._data', 'pandas.compat.numpy.function', 'tensorflow.python.distribute.coordinator.watchdog', 'tensorflow._api.v2.compat.v2.dtypes', 'keras.src.layers.preprocessing.hashed_crossing', 'tensorflow.python.util.protobuf.compare', 'pandas.core.arrays.floating', 'tensorflow._api.v2.profiler.experimental.client', 'tensorflow.python.util.variable_utils', 'pandas.io.excel._odswriter', 'tensorflow.python.ops.distributions.laplace', 'tensorflow._api.v2.compat.v1.random.experimental', 'scipy.linalg._expm_frechet', 'tensorflow._api.v2.compat.v1.types', 'keras.src.version', 'tensorflow.python.framework.extension_type', 'tensorflow.python.keras.backend_config', 'tensorflow.python.ops.composite_tensor_ops', 'tensorflow.dtensor.python.accelerator_util', 'pandas.io._util', 'absl.command_name', 'tensorflow.python.util._pywrap_util_port', 'tensorflow.python.training.checkpoint_ops', 'pandas.io.pickle', 'tensorflow.python.trackable.constants', 'tensorflow._api.v2.compat.v2.experimental.tensorrt', 'tensorflow.core.framework.node_def_pb2', 'tensorflow.python.ops.string_ops', 'tensorflow._api.v2.compat.v1.saved_model', 'keras.src.datasets.california_housing', 'tensorflow.python.framework.weak_tensor', 'tensorflow.python.eager.polymorphic_function.autograph_util', 'tensorflow._api.v2.__internal__.distribute.multi_process_runner', 'tensorflow.python.ops.ragged.segment_id_ops', 'scipy.sparse.linalg._norm', 'h5py.h5f', 'cv2.typing', 'pandas.core.ops.missing', 'tensorflow.core.tpu.kernels', 'keras.src.metrics.metric', 'ignite.metrics.regression.pearson_correlation', 'tensorflow.python.framework._dtypes', 'tensorflow._api.v2.data.experimental', 'tensorflow.python.grappler', 'tensorflow.python.distribute.tpu_replicated_variable', 'tensorflow.python.ops.linalg.linear_operator_addition', 'keras.api', 'tensorflow._api.v2.compat.v1.linalg', 'opt_einsum.backends.dispatch', 'tensorflow.python.tpu._pywrap_sparse_core_layout', 'tensorflow.python.eager.function', 'keras.src.backend.tensorflow.rnn', 'tensorflow.compiler.mlir.quantization.tensorflow.quantization_options_pb2', 'tensorflow.python.ops.data_flow_ops', 'tensorflow._api.v2.compat.v2.nn', 'tensorflow.python.layers.utils', 'tensorflow.python.ops.variable_scope', 'tensorflow.python.ops.gen_sendrecv_ops', 'tensorflow.python.saved_model.nested_structure_coder', 'tensorflow._api.v2.train.experimental', 'pandas.core.indexers', 'tensorflow.python.autograph.operators.data_structures', 'tensorflow.python.framework.c_api_util', 'tensorflow._api.v2.compat.v1.sparse', 'pandas.core.reshape.pivot', 'tensorflow._api.v2.compat.v1.data.experimental.service', 'keras.api.applications.vgg19', 'keras.src.activations', 'tensorflow.python.tpu.tpu', 'scipy.linalg', 'tensorflow.python.ops.random_ops', 'tensorflow._api.v2.compat.v2.train.experimental', 'tensorflow.dtensor.proto.layout_pb2', 'rich._emoji_codes', 'keras.src.backend.common.keras_tensor', 'tensorflow._api.v2.bitwise', 'tensorflow.python.tpu', 'tensorflow.compiler.mlir.quantization.tensorflow.calibrator.calibration_statistics_pb2', 'tensorflow.python.eager.polymorphic_function.saved_model_utils', 'tensorflow.python.framework.constant_op', 'tensorflow.python.keras.engine.base_preprocessing_layer', 'ignite.handlers.tensorboard_logger', 'keras.src.layers.core.wrapper', 'tensorflow._api.v2.compat.v2.feature_column', 'tensorflow.python.distribute.parameter_server_strategy', 'xml.sax', 'keras.src.initializers', 'tensorflow.python.saved_model.method_name_updater', 'torch.distributed.optim', 'tensorflow._api.v2.compat.v2.compat.v1', 'tensorflow.python.trackable.python_state', 'tensorflow.core.function.capture', 'tensorflow._api.v2.math', 'keras.src.backend.common.global_state', 'tensorflow._api.v2.compat.v2.__internal__.test.combinations', 'tensorflow._api.v2.compat.v2.__internal__.smart_cond', 'tensorflow.distribute.experimental.rpc.kernels.gen_rpc_ops', 'tensorflow.python.ops.structured', 'tensorflow.python.autograph.pyct.cfg', 'tensorflow.python.training.saving.saveable_object', 'tensorflow.python.distribute.coordinator', 'cv2.dnn', 'tensorflow.python.keras.utils.object_identity', 'ignite.handlers.timing', 'pandas', 'tensorflow.python.ops.gen_count_ops', 'tensorflow.python.summary.writer.writer', 'tensorflow._api.v2.compat.v2.config', 'tensorflow.python.framework.cpp_shape_inference_pb2', 'tensorflow._api.v2.compat.v2.sparse', 'pandas.io.json._normalize', 'tensorflow.python.ops.linalg.sparse.sparse_csr_matrix_grad', 'scipy.linalg._solvers', 'tensorflow._api.v2.lite', 'h5py._hl.dataset', 'attr.exceptions', 'tensorflow.python.ops.linalg.sparse.conjugate_gradient', 'tensorflow.python.autograph.pyct.templates', 'tensorflow.python.keras.layers.legacy_rnn.rnn_cell_wrapper_impl', 'tensorflow._api.v2.compat.v2.lite.experimental', 'tensorflow.python.distribute.multi_process_lib', 'tensorflow.python.ops.ragged.ragged_dispatch', 'tensorflow.python.ops.linalg_grad', 'tensorflow.python.distribute.shared_variable_creator', 'h5py.h5z', 'tensorflow.python.summary.writer', 'tensorflow.python.ops.linalg.linear_operator_inversion', 'keras.api.datasets.reuters', 'tensorflow._api.v2.compat.v1.config.optimizer', 'tensorflow.compiler.mlir.quantization.tensorflow.python.py_function_lib', 'h5py.h5ac', 'keras.src.datasets.cifar100', 'pandas.core.window.numba_', 'torch.distributed.optim.apply_optimizer_in_backward', 'pandas.core.reshape.api', 'tensorflow._api.v2.compat.v1.tpu.experimental.embedding', 'h5py.h5g', 'ml_dtypes', 'tensorflow.python.autograph.lang.special_functions', 'keras.src.layers.convolutional.base_conv', 'pandas.core.indexes.interval', 'tensorflow.python.ops.gen_control_flow_ops', 'optree', 'scipy._lib._pep440', 'torch.distributed.optim.utils', 'tensorflow._api.v2.image', 'keras.src.optimizers.schedules.learning_rate_schedule', 'ignite.handlers.base_logger', 'keras.src.layers.activations.relu', 'ignite.handlers.mlflow_logger', 'tensorflow.compiler.jit.ops', 'opt_einsum.contract', 'tensorflow.python.autograph.pyct.static_analysis', 'tensorflow._api.v2.nn', 'tensorflow.core.profiler.tfprof_log_pb2', 'pandas.core.flags', 'tensorflow.python.eager.core', 'keras.src.layers.reshaping.zero_padding1d', 'tensorflow._api.v2.io', 'keras.src.backend', 'keras.src.applications.resnet_v2', 'tensorflow._api.v2.compat.v2.__internal__.mixed_precision', 'optree._C', 'cv2.gapi.render.ocv', 'tensorflow._api.v2.compat.v2.graph_util', 'keras.src.trainers.data_adapters.array_data_adapter', 'ignite.distributed.auto', 'tensorflow.python.distribute', 'tensorflow._api.v2.__internal__.dispatch', 'tensorflow.dtensor.python.mesh_util', 'optree.typing', 'tensorflow.python.data.ops', 'tensorflow.core.grappler.costs', 'tensorflow.python.eager.test', 'tensorflow._api.v2.compat.v1.lite.experimental', 'tensorflow.python.ops.ragged.ragged_conversion_ops', 'tensorflow.python.ops.gen_boosted_trees_ops', 'rich._null_file', 'scipy.sparse', 'pandas.util._tester', 'pandas.core.dtypes.api', 'keras.src.layers.core.lambda_layer', 'tensorflow._api.v2.__operators__', 'tensorflow.python.debug.wrappers', 'tensorflow.python.checkpoint.checkpoint_view', 'tensorflow.python.distribute.cluster_resolver.gce_cluster_resolver', 'tensorflow.python.framework.constant_tensor_conversion', 'tensorflow._api.v2.compat.v2.dtypes.experimental', 'tensorflow.python.ops.gen_sdca_ops', 'scipy.linalg._decomp_cholesky', 'tensorflow._api.v2.compat.v2.config.optimizer', 'tensorflow._api.v2.compat.v1.mlir', 'scipy.sparse.csgraph._shortest_path', 'tensorflow.python.ops.math_ops', 'cv2', 'tensorflow.python.autograph.pyct.inspect_utils', 'tensorflow.python.eager.forwardprop_util', 'pandas.core.nanops', 'tensorflow._api.v2.compat.v2.raw_ops', 'optree.accessor', 'tensorflow.python.trackable.resource', 'scipy.linalg.decomp', 'keras.src.applications.efficientnet_v2', 'tensorflow._api.v2.autograph', 'scipy.linalg.decomp_svd', 'scipy.sparse.linalg._eigen.arpack.arpack', 'tensorflow.lite.python.convert_saved_model', 'tensorflow.dtensor.python.gen_dtensor_ops', 'tensorflow.core.tpu.kernels.sparse_core_layout_pb2', 'keras.src.layers.preprocessing.integer_lookup', 'tensorflow.python.profiler.profiler_v2', 'tensorflow.python.ops.init_ops_v2', 'tensorflow.python.ops.check_ops', 'cv2.gapi.render', 'tensorflow.security.fuzzing', 'tensorflow.core.lib.core.error_codes_pb2', 'keras.api.optimizers.legacy', 'src.models', 'tensorflow.python.saved_model.signature_constants', 'rich._log_render', 'tensorflow.tsl.protobuf.test_log_pb2', 'tensorflow.python.distribute.experimental.dtensor_util', 'tensorflow.python.ops.sparse_ops', 'tensorflow.python.util.type_annotations', 'tensorflow.python.ops.bincount_ops', 'scipy.linalg._special_matrices', 'tensorflow.python.user_ops.user_ops', 'ignite.metrics.regression.median_absolute_percentage_error', 'tensorflow.python.ops.random_ops_util', 'optree.integration', 'tensorflow._api.v2.compat.v1.version', 'tensorflow.python.keras.mixed_precision.loss_scale', 'tensorflow._api.v2.compat.v2.distribute.experimental.rpc', 'tensorflow.python.ops.special_math_ops', 'tensorflow.python.layers.base', 'h5py._proxy', 'pandas.core', 'ignite.metrics.epoch_metric', 'h5py.h5s', 'scipy.linalg._flinalg_py', 'src.preprocessing', 'scipy._lib.array_api_compat.array_api_compat.numpy', 'tensorflow.python.distribute.cluster_resolver.tpu_cluster_resolver', 'keras.src.layers.convolutional.conv3d', 'scipy.linalg._matfuncs_expm', 'keras.src.trainers.data_adapters.generator_data_adapter', 'tensorflow.python.util.tf_decorator', 'keras.src.layers.reshaping.up_sampling2d', 'tensorflow.python.ops.linalg.linear_operator_util', 'pandas.core.array_algos.take', 'scipy.linalg._matfuncs_sqrtm_triu', 'ignite.metrics.accuracy', 'tensorflow.python.feature_column.sequence_feature_column', 'absl.flags._validators', 'tensorflow.python.ops.linalg.linear_operator_low_rank_update', 'tensorflow.python.ops.map_fn', 'keras.src.ops.numpy', 'rich.theme', 'ignite.metrics.mean_pairwise_distance', 'tensorflow.python.eager.polymorphic_function.function_context', 'tensorflow._api.v2.nn.experimental', 'tensorflow.python.eager', 'tensorflow.lite.python.analyzer', 'cv2.gapi.ot.cpu', 'pandas.io.feather_format', 'torch.distributed.optim.functional_adamw', 'keras.src.layers.rnn.time_distributed', 'h5py._hl', 'tensorflow.python.ops.linalg.linear_operator', 'pandas._libs.join', 'tensorflow.python.autograph.pyct.errors', 'keras.src.applications.imagenet_utils', 'tensorflow.security.fuzzing.py.annotation_types', 'tensorflow.python.saved_model.save_options', 'tensorflow.python.data.experimental.service', 'tensorflow.python.ops.linalg.linear_operator_permutation', 'keras.src.optimizers.lion', 'zoneinfo._tzpath', 'keras.src.legacy', 'tensorflow.python.data.ops.iterator_ops', 'tensorflow._api.v2.xla.experimental', 'tensorflow.python.data.ops.random_op', 'tensorflow.python.ops.signal', 'scipy.linalg._basic', 'tensorflow.lite.python.metrics.wrapper.metrics_wrapper', 'keras.src.callbacks.history', 'tensorflow.python.ops.losses.losses', 'keras.src.legacy.saving.saving_options', 'tensorflow.python.training.queue_runner_impl', 'tensorflow._api.v2.mlir.experimental', 'numpy._core._multiarray_umath', 'ignite.handlers.param_scheduler', 'keras.src.layers.pooling.average_pooling2d', 'tensorflow._api.v2.signal', 'tensorflow._api.v2.compat.v1.saved_model.loader', 'tensorflow._api.v2.compat.v1.debugging', 'scipy.linalg._decomp_svd', 'tensorflow.python.ops.manip_ops', 'tensorflow.python.training.proximal_gradient_descent', 'wrapt._wrappers', 'tensorflow.compiler.mlir.lite.python.wrap_converter', 'tensorflow.python.platform.resource_loader', 'keras.src.ops.linalg', 'tensorflow.python.autograph.pyct.pretty_printer', 'h5py.h5d', 'tensorflow._api.v2.compat.v2.__internal__.eager_context', 'tensorflow.python.compiler.xla.xla', 'pandas._libs.tslibs.fields', 'tensorflow.python.keras.engine.training_generator_v1', 'pandas._config.localization', 'tensorflow._api.v2.compat.v2.autograph.experimental', 'tensorflow.python.ops.gen_rnn_ops', 'tensorflow.python.ops.ragged.ragged_gather_ops', 'opt_einsum.backends.jax', 'tensorflow.python.data.experimental.ops.batching', 'keras.src.applications.densenet', 'pandas.core.tools.timedeltas', 'tensorflow.core.function.trace_type.serialization_pb2', 'tensorflow.lite.experimental', 'tensorflow.python.autograph.core.config_lib', 'tensorflow.python.data.util.nest', 'tensorflow.core.framework.step_stats_pb2', 'cv2.samples', 'tensorflow.python.autograph.utils.tensor_list', 'tensorflow.python.ops.gen_checkpoint_ops', 'pandas.io.excel', 'rich.table', 'scipy.sparse.sparsetools', 'pandas.core.ops', 'tensorflow.python.ops.gen_dataset_ops', 'scipy.sparse.csgraph._traversal', 'tensorflow.python.util.custom_nest_protocol', 'tensorflow.python.ops.signal.util_ops', 'tensorflow.compiler.mlir.quantization.tensorflow.python.save_model', 'tensorflow.python.util.keras_deps', 'tensorflow.python.ops.gen_script_ops', 'keras.src.callbacks.remote_monitor', 'scipy.sparse.linalg.interface', 'tensorflow.python.checkpoint.checkpoint_management', 'keras.src.callbacks.learning_rate_scheduler', 'tensorflow.lite.python.util', 'tensorflow._api.v2.random.experimental', 'pandas.compat.pyarrow', 'keras.src.applications.mobilenet_v2', 'ignite.contrib', 'tensorflow.python.autograph.pyct.static_analysis.reaching_definitions', 'tensorflow.core.framework.graph_debug_info_pb2', 'pandas.core.indexes.base', 'tensorflow.core.protobuf.tpu.dynamic_padding_pb2', 'scipy.sparse.construct', 'keras.src.backend.tensorflow.image', 'keras.src.backend.tensorflow.numpy', 'pandas.io.xml', 'src.utils', 'tensorflow._api.v2.compat.v1.saved_model.builder', 'tensorflow._api.v2.compat.v2.experimental', 'tensorflow.python.ops.linalg.linear_operator_tridiag', 'keras.src.layers.pooling.max_pooling1d', 'keras.src.layers.attention.attention', 'pandas.core.dtypes.dtypes', 'keras.src.metrics.regression_metrics', 'optree.version', 'keras.api.applications.imagenet_utils', 'keras.src.trainers.epoch_iterator', 'h5py._selector', 'pandas._libs.ops_dispatch', 'tensorflow.python.eager.polymorphic_function.concrete_function', 'tensorflow._api.v2.experimental.dlpack', 'opt_einsum.backends', 'tensorflow.python.ops.gen_math_ops', 'tensorflow.python.ops.gen_sync_ops', 'tensorflow._api.v2.dtypes', 'tensorflow.python.ops.gradients', 'keras.src.backend.common.dtypes', 'scipy.linalg.blas', 'tensorflow._api.v2.__internal__.feature_column', 'tensorflow.python.util._pywrap_utils', 'pandas._testing._io', 'opt_einsum.helpers', 'cv2.misc', 'tensorflow.python.platform._pywrap_cpu_feature_guard', 'tensorflow.python.ops.image_ops', 'tensorflow._api.v2.compat.v1.ragged', 'tensorflow.python.framework.stack', 'tensorflow.python.ops.linalg_ops', 'tensorflow.python.ops.ragged.ragged_factory_ops', 'scipy.sparse.linalg._isolve', 'scipy.sparse.dok', 'keras.src.layers.convolutional.separable_conv1d', 'tensorflow.python.ops.ragged.ragged_getitem', 'tensorflow.python.tpu.tpu_name_util', 'tensorflow.compiler.mlir', 'tensorflow.python.util.traceback_utils', 'pandas.util._print_versions', '97d2165b493fecec79c65b5c6254dffd4e375528', 'tensorflow.python.keras.saving.saved_model.network_serialization', 'keras.api.datasets.california_housing', 'google.protobuf.any_pb2', 'tensorflow.python.compiler', 'keras.src.layers.preprocessing.normalization', 'keras.src.optimizers.optimizer', 'cv2.parallel', 'pandas.core.array_algos.masked_reductions', 'tensorflow.python.ops.sdca_ops', 'tensorflow.python.framework.config', 'keras.src.layers.convolutional.depthwise_conv2d', 'tensorflow.python.training.experimental.loss_scale_optimizer', 'tensorflow._api.v2.compat.v2.mlir.experimental', 'keras.src.callbacks.callback_list', 'keras.api.ops.linalg', 'tensorflow.python.ops.parallel_for.control_flow_ops', 'tensorflow.python.saved_model.fingerprinting_utils', 'keras.src.utils.rng_utils', 'tensorflow.python.keras.saving.saved_model.save_impl', 'tensorflow.dtensor.python.d_checkpoint', 'tensorflow.python.keras.optimizer_v2.gradient_descent', 'tensorflow.python.saved_model.model_utils.mode_keys', 'keras.src.trainers.data_adapters.data_adapter', 'tensorflow.core.framework.dataset_options_pb2', 'tensorflow.python.tpu.tpu_strategy_util', 'tensorflow._api.v2.mlir', 'keras.src.constraints.constraints', 'pandas.core.window.rolling', 'tensorflow.python.ops.gen_collective_ops', 'tensorflow.python.ops.gen_decode_proto_ops', 'tensorflow._api.v2.compat.v1.mixed_precision.experimental', 'tensorflow._api.v2.compat.v2.__internal__.distribute.combinations', 'keras.src.layers.preprocessing.hashing', 'pandas._libs.tslibs.nattype', 'keras.src.utils.file_utils', 'keras.src.backend.common.symbolic_scope', 'keras.src.applications.inception_resnet_v2', 'keras.src.layers.merging', 'tensorflow.dtensor.python', 'tensorflow.python.checkpoint.saveable_compat', 'ignite.metrics.regression.r2_score', 'tensorflow.python.checkpoint.functional_saver', 'tensorflow.python.framework.tensor', 'tensorflow.python.lib.io._pywrap_file_io', 'tensorflow._api.v2.compat.v2.distribute.cluster_resolver', 'tensorflow.python.ops.tensor_array_grad', 'tensorflow.python.ops.distributions.dirichlet_multinomial', 'keras.src.layers.preprocessing.category_encoding', 'tensorflow.python.keras.utils.generic_utils', 'tensorflow.python.ops.gen_nn_ops', 'h5py', 'tensorflow._api.v2.compat.v1.debugging.experimental', 'scipy.linalg._decomp', 'keras.src.layers.reshaping.zero_padding2d', 'tensorflow._api.v2.compat.v2.__internal__.tf2', 'ignite.metrics.kl_divergence', 'tensorflow.python.framework.immutable_dict', 'tensorflow._api.v2.compat.v2.test.experimental', 'ml_dtypes._finfo', 'keras.src.layers.attention.grouped_query_attention', 'keras.src.metrics.f_score_metrics', 'tensorflow._api.v2.experimental.tensorrt', 'ignite.metrics.regression', 'pandas.core.common', 'tensorflow._api.v2.compat.v1.initializers', 'keras.src.legacy.saving.json_utils', 'tensorflow.python.keras.optimizer_v2.learning_rate_schedule', 'src.models.context_modules', 'keras.src.layers.preprocessing.random_zoom', 'tensorflow.python.ops.sets', 'keras.src.layers.pooling.global_average_pooling1d', 'rich.containers', 'tensorflow._api.v2.config', 'tensorflow.python.debug.cli.analyzer_cli', 'keras.src.optimizers.adadelta', 'pandas.core.dtypes.astype', 'tensorflow._api.v2.__internal__.graph_util', 'tensorflow.python.distribute.experimental.rpc', 'tensorflow.core.profiler.tfprof_output_pb2', 'tensorflow.python.pywrap_tfe', 'tensorflow.python.autograph.pyct.loader', 'pandas._libs.tslibs.np_datetime', 'scipy.sparse.linalg._eigen', 'scipy.linalg.decomp_schur', 'tensorflow.core.framework.dataset_metadata_pb2', 'tensorflow._api.v2.compat.v2.__internal__.function', 'cv2.gapi.core.ocl', 'tensorflow.python.autograph.impl', 'scipy.linalg._decomp_lu_cython', 'ignite.metrics.gan.utils', 'tensorflow.python.autograph.core.ag_ctx', 'tensorflow.python.training.adagrad', 'tensorflow.python.ops.gen_string_ops', 'scipy.linalg._decomp_ldl', 'cv2.gapi.oak', 'pandas._libs.tslibs.vectorized', 'tensorflow._api.v2.compat.v1.distribute', 'tensorflow.python.data.ops.map_op', 'pandas.core.computation', 'tensorflow.python.training.monitored_session', 'tensorflow.python.ops.proto_ops', 'keras.api.applications.inception_resnet_v2', 'tensorflow.python.summary.summary_iterator', 'src.datasets.sunrgbd', 'tensorflow.python.autograph.core.config', 'tensorflow._api.v2.compat.v1.linalg.experimental', 'rich._wrap', 'keras.api.applications.resnet50', 'tensorflow.python.keras.utils.layer_utils', 'scipy._distributor_init', 'keras.src.optimizers.adagrad', 'tensorflow.python.keras.engine.training_eager_v1', 'tensorflow.python.keras.layers.convolutional_recurrent', 'tensorflow.python.data.experimental.ops.grouping', 'pandas.util._exceptions', 'keras.src.applications.mobilenet_v3', 'opt_einsum', 'ignite.metrics.gan', 'tensorflow.core.config.flags', 'tensorflow.lite.python', 'tensorflow.python.tpu.bfloat16', 'gast.gast', 'scipy.linalg._matfuncs', 'scipy._lib._threadsafety', 'scipy.sparse.compressed', 'src.models.model', 'absl.testing.xml_reporter', 'tensorflow._api.v2.compat.v1.saved_model.main_op', 'tensorflow.python._pywrap_quantize_training', 'pandas.core.groupby.grouper', 'tensorflow.python.profiler.internal._pywrap_traceme', 'scipy.sparse.csc', 'keras.api.applications.resnet', 'scipy.sparse.linalg._isolve.lsqr', 'tensorflow.python.ops.ragged.ragged_map_ops', 'tensorflow.python.keras.engine.node', 'scipy.__config__', 'tensorflow.python.data.experimental.ops.get_single_element', 'tensorflow._api.v2.compat.v1.train.queue_runner', 'pandas._libs.tslibs.tzconversion', 'tensorflow.lite.experimental.microfrontend.ops.gen_audio_microfrontend_op', 'tensorflow.python.compiler.tensorrt.utils', 'tensorflow.python.saved_model.tracing_utils', 'tensorflow.python.platform.app', 'tensorflow.python.ops.ragged.ragged_concat_ops', 'tensorflow._api.v2.compat.v1.profiler', 'rich.highlighter', 'keras.api.saving', 'tensorflow.python.eager.executor', 'tensorflow._api.v2.distribute.experimental', 'pandas.core.ops.mask_ops', 'tensorflow.python.saved_model.builder_impl', 'tensorflow.distribute.experimental.rpc.kernels', 'tensorflow._api.v2.compat.v1.xla', 'keras.src.utils.dtype_utils', 'pandas.core.array_algos.replace', 'tensorflow.core.framework.kernel_def_pb2', 'tensorflow.python.feature_column.serialization', 'tensorflow.python.checkpoint', 'tensorflow._api.v2.data', 'ignite.metrics.gpu_info', 'tensorflow.python.compiler.xla.experimental', 'pandas.core.construction', 'tensorflow.lite.python.authoring.authoring', 'scipy.linalg._procrustes', 'keras.src.utils.backend_utils', 'tensorflow.python.data.experimental.ops.data_service_ops', 'keras.src.trainers.trainer', 'tensorflow.python.util.lazy_loader', 'tensorflow.core.framework.api_def_pb2', 'keras.src.trainers.data_adapters.torch_data_loader_adapter', 'ignite.metrics', 'tensorflow.python.pywrap_mlir', 'tensorflow.python.saved_model.model_utils.export_utils', 'keras.src.trainers.data_adapters.array_slicing', 'pandas.api.typing', 'tensorflow.core.framework.device_attributes_pb2', 'h5py.h5i', 'keras.src.saving', 'tensorflow.python.framework.graph_io', 'pandas.core.indexes', 'ignite.distributed.comp_models.horovod', 'tensorflow.python.debug.lib.source_utils', 'keras.src.backend.tensorflow.nn', 'scipy.sparse.linalg._eigen.arpack._arpack', 'pandas.core.arrays.numpy_', 'keras.src.metrics.probabilistic_metrics', 'tensorflow.python.distribute.collective_all_reduce_strategy', 'tensorflow.python.framework.test_util', 'tensorflow.core.example', 'tensorflow.python.data.experimental.ops.compression_ops', 'tensorflow.python.distribute.experimental.rpc.rpc_ops', 'keras.api.datasets.boston_housing', 'pandas.core.internals.array_manager', 'tensorflow.python.ops.handle_data_util', 'scipy.linalg._decomp_cossin', 'tensorflow._api.v2.compat.v1.losses', 'tensorflow._api.v2.__internal__.tracking', 'tensorflow.python.tf2', 'keras.api.datasets.cifar10', 'tensorflow.python.autograph.converters.lists', 'tensorflow.core.function.polymorphism.function_cache', 'tensorflow.compiler.jit.ops.xla_ops_grad', 'tensorflow.python.tpu.training_loop', 'scipy.sparse._csc', 'tensorflow.python.ops.linalg.linear_operator_kronecker', 'ignite.metrics.regression.wave_hedges_distance', 'keras.api.ops.image', 'cv2.gapi.core', 'keras.api.applications.resnet_v2', 'tensorflow.python.framework.random_seed', 'optree.ops', 'tensorflow.python.keras.initializers.initializers_v1', 'tensorflow._api.v2.compat.v2.mlir', 'tensorflow.core.framework.function_pb2', 'tensorflow._api.v2.compat.v1.autograph', 'tensorflow._api.v2.lite.experimental', 'tensorflow.python.ops.linalg.linear_operator_adjoint', 'tensorflow.python.ops.logging_ops', 'keras.src.layers.preprocessing.rescaling', 'keras.src.layers.convolutional.conv3d_transpose', 'tensorflow.python.platform.tf_logging', 'ignite.metrics.cohen_kappa', 'tensorflow.python.ops.ragged.dynamic_ragged_shape', 'pandas.core.arrays.sparse.array', 'tensorflow.python.autograph.utils.context_managers', 'tensorflow.python.training.training_util', 'ignite.metrics.mean_squared_error', 'tensorflow._api.v2.compat.v2.strings', 'tensorflow.python.ops.init_ops', 'tensorflow.python.ops.gen_array_ops', 'pytz.tzfile', 'pandas.core.computation.align', 'tensorflow.python.ops.ragged.ragged_batch_gather_ops', 'tensorflow.core.protobuf.error_codes_pb2', 'tensorflow.lite.tools.optimize.debugging', 'keras.src.layers.preprocessing.feature_space', 'tensorflow.lite.python.lite_constants', 'pandas.core.accessor', 'pandas._libs.pandas_parser', 'tensorflow._api.v2.compat.v1.graph_util', 'tensorflow.python.autograph', 'keras.src.backend.tensorflow.linalg', 'pandas.core.strings.base', 'cv2.utils', 'tensorflow.python.profiler.internal', 'absl.flags._validators_classes', 'keras.src.callbacks.csv_logger', 'tensorflow.python.debug.lib', 'scipy.linalg._decomp_schur', 'tensorflow.python.training.session_manager', 'tensorflow.python.distribute.summary_op_util', 'tensorflow.python.autograph.operators.control_flow', 'tensorflow.python.keras.layers.serialization', 'absl.flags._flag', 'tensorflow.python.debug.lib.debug_graphs', 'tensorflow.python.ops.linalg.linear_operator_circulant', 'tensorflow.python.debug.lib.profiling', 'tensorflow._api.v2.compat.v2.random', 'scipy.linalg._decomp_update', 'tensorflow.python.data.util', 'tensorflow._api.v2.compat.v1.__internal__', 'tensorflow.python.ops.unconnected_gradients', 'tensorflow.compiler.mlir.quantization.tensorflow', 'tensorflow._api.v2.compat.v2.audio', 'multiprocessing.pool', 'tensorflow.python.ops.array_grad', 'pandas._libs.window.indexers', 'tensorflow.python.client.pywrap_tf_session', 'tensorflow.python.framework.test_combinations', 'keras.src.initializers.initializer', 'scipy._lib._util', 'pandas._libs.reshape', 'tensorflow.python.keras', 'tensorflow.python.framework.op_callbacks', 'tensorflow.python.distribute.cluster_resolver', 'pandas.core.arrays.sparse', 'tensorflow._api.v2.compat.v1.data', 'tensorflow.python.eager.record', 'scipy.sparse._csparsetools', 'keras.src.callbacks.progbar_logger', 'pandas.io.formats', 'tensorflow.python.autograph.converters.break_statements', 'keras.src.layers.rnn.gru', 'attr.setters', 'tensorflow._api.v2.compat.v1.lookup', 'scipy.linalg.decomp_qr', 'tensorflow.python.saved_model.loader_impl', 'scipy._lib.array_api_compat.array_api_compat.common', 'tensorflow.python.data.experimental.ops.enumerate_ops', 'tensorflow.python.ops.nn_ops', 'rich._ratio', 'scipy.sparse.linalg._interface', 'src.confusion_matrix', 'tensorflow.python.ops.gradients_impl', 'pandas.core.config_init', 'tensorflow._api.v2.compat.v2.signal', 'keras.src.random.random', 'tensorflow.python.data.experimental.ops.scan_ops', 'tensorflow._api.v2.compat.v2.sets', 'tensorflow.python.ops.signal.mel_ops', 'tensorflow.python.keras.utils.losses_utils', 'tensorflow.compiler.tf2xla.python.xla', 'src.build_model', 'cv2.Error', 'tensorflow.core.framework.cost_graph_pb2', 'scipy.sparse.linalg._isolve.minres', 'tensorflow._api.v2.compat.v1.__internal__.types', 'tensorflow.lite.python.schema_py_generated', 'tensorflow.core.protobuf.composite_tensor_variant_pb2', 'tensorflow.python.distribute.coordinator.utils', 'tensorflow._api.v2.compat.v1.saved_model.tag_constants', 'pandas._libs.tslibs.period', 'tensorflow.python.tools', 'tensorflow.python.ops.parallel_for', 'pandas.core.window.expanding', 'tensorflow.python.ops.resources', 'scipy.sparse.csgraph._flow', 'tensorflow.python.keras.optimizer_v2.adamax', 'tensorflow.python.autograph.utils.type_registry', 'tensorflow.core.protobuf.trackable_object_graph_pb2', 'pandas.core.indexes.range', 'tensorflow.core.framework.allocation_description_pb2', 'flatbuffers.packer', 'tensorflow.python.ops.gen_filesystem_ops', 'h5py.defs', 'tensorflow.python.data.util.random_seed', 'tensorflow.python.framework.meta_graph', 'ignite.base', 'tensorflow.python.ops.image_grad', 'tensorflow.python.ops.ragged.ragged_autograph', 'tensorflow.python.keras.optimizer_v2.legacy_learning_rate_decay', 'tensorflow.python.eager.polymorphic_function.saved_model_exported_concrete', 'keras.src.layers.merging.maximum', 'tensorflow.python.compat.compat', 'pandas._libs.missing', 'keras.api.models', 'tensorflow.python.saved_model.save_context', 'ignite.handlers.checkpoint', 'h5py._errors', 'tensorflow._api.v2.compat.v1.gfile', 'tensorflow.python.ops.control_flow_v2_toggles', 'tensorflow.python.framework.extension_type_field', 'scipy.sparse.csr', 'ignite.metrics.regression.maximum_absolute_error', 'tensorflow._api.v2.compat.v2.autodiff', 'tensorflow.python.types.distribute', 'tensorflow.python.autograph.lang', 'keras.src.layers.rnn', 'tensorflow.python.framework._pywrap_python_api_dispatcher', 'tensorflow.lite.python.interpreter', 'scipy.sparse.csgraph._reordering', 'ignite.handlers.time_profilers', 'keras.src.layers.convolutional.conv1d', 'scipy.sparse._matrix', 'tensorflow.python.framework.op_def_registry', 'keras.src.dtype_policies', 'tensorflow._api.v2.debugging.experimental', 'tensorflow.python.checkpoint.sharding', 'cv2.gapi.onnx', 'keras.src.utils.tf_utils', 'tensorflow.python.autograph.pyct.anno', 'tensorflow.python.ops.while_v2_indexed_slices_rewriter', 'tensorflow.core.protobuf.control_flow_pb2', 'scipy.sparse.linalg._eigen.arpack', 'tensorflow.python.ops.gen_nccl_ops', 'tensorflow.lite.profiling', 'rich.emoji', 'tensorflow.python.autograph.pyct.qual_names', 'pandas.core.roperator', 'keras.src.backend.common.name_scope', 'tensorflow._api.v2.compat.v2.quantization', 'tensorflow._api.v2.compat.v2.compat', 'pandas._libs.index', 'src.datasets.nyuv2.nyuv2', 'scipy.linalg._fblas', 'scipy.linalg._misc', 'tensorflow._api.v2.random', 'pandas.core.indexes.period', 'tensorflow._api.v2.compat.v2.debugging.experimental', 'cv2.gapi.onnx.ep', 'tensorflow._api.v2.compat.v2.config.threading', 'tensorflow.python.lib.io._pywrap_record_io', 'tensorflow.python.trackable.converter', 'pandas.core.groupby.numba_', 'tensorflow._api.v2.compat.v1.nn.experimental', 'tensorflow.python.training.adagrad_da', 'tensorflow.python.checkpoint.async_checkpoint_helper', 'tensorflow.python.ops.variables', 'tensorflow.compiler.tf2tensorrt.ops', 'scipy.sparse._compressed', 'tensorflow.core.function.trace_type.custom_nest_trace_type', 'rich.palette', 'ignite.metrics.regression.mean_error', 'tensorflow.python.profiler.model_analyzer', 'tensorflow.python.keras.layers.rnn_cell_wrapper_v2', 'ignite.metrics.cosine_similarity', 'pandas.core.ops.common', 'keras.src.metrics.hinge_metrics', 'keras.src.layers.core', 'tensorflow.python.client._pywrap_tf_session', 'tensorflow.python.module.module', 'tensorflow.python.compiler.tensorrt.trt_convert', 'cv2.ocl', 'tensorflow.python.keras.saving.saving_utils', 'ignite.metrics.ssim', 'tensorflow.python.data.experimental.ops.from_list', 'torch.distributed.optim.functional_adagrad', 'tensorflow.python.ops.stateless_random_ops', 'tensorflow.python.keras.engine.input_layer', 'keras.src.layers.preprocessing.random_contrast', 'tensorflow.compiler.xla', 'tensorflow._api.v2.experimental.numpy.random', 'wrapt.patches', 'tensorflow._api.v2.compat.v1.lookup.experimental', 'tensorflow.python.autograph.operators.exceptions', 'termcolor.termcolor', 'scipy.sparse.linalg._isolve.tfqmr', 'pandas._libs.tslibs.dtypes', 'tensorflow.python.distribute.coordinator.coordinator_context', 'tensorflow.python.ops.stateful_random_ops', 'tensorflow.python.training.device_setter', 'keras.src.layers.merging.add', 'tensorflow.python.tpu.tpu_embedding_v1', 'tensorflow._api.v2.compat.v1.saved_model.signature_constants', 'keras.src.layers.pooling.base_global_pooling', 'tensorflow.python.debug.lib.debug_events_writer', 'tensorflow.python.util.all_util', 'tensorflow.python.data.experimental.ops.pad_to_cardinality', 'tensorflow.core.profiler', 'tensorflow.tsl.protobuf.error_codes_pb2', 'tensorflow.python.util.lock_util', 'ignite.handlers.wandb_logger', 'keras.api.applications.densenet', 'tensorflow.python.ops.gen_stateful_random_ops', 'pandas._libs.arrays', 'tensorflow.python.trackable.data_structures', 'tensorflow.lite.toco.toco_flags_pb2', 'keras.api.applications.vgg16', 'tensorflow.python.tpu.device_assignment', 'tensorflow.python.ops.numpy_ops.np_array_ops', 'keras.src.layers.regularization.gaussian_noise', 'tensorflow.python.grappler._pywrap_tf_cluster', 'keras.src.callbacks.callback', 'tensorflow.lite.python.schema_util', 'keras.src.layers.reshaping.permute', 'flatbuffers.builder', 'pandas.core.groupby.generic', 'h5py._conv', 'tensorflow.python.feature_column.feature_column_v2_types', 'pandas._libs.hashing', 'tensorflow.python.training.experimental.mixed_precision_global_state', 'tensorflow.compiler.mlir.quantization.tensorflow.calibrator.calibration_algorithm', 'scipy.linalg._sketches', 'keras.api.datasets.cifar100', 'tensorflow._api.v2.math.special', 'tensorflow.compiler.tf2xla.ops', 'tensorflow.python.saved_model.main_op_impl', 'keras.src.layers.preprocessing.index_lookup', 'tensorflow.lite.python.optimize', 'tensorflow.python.ops.distributions.uniform', 'tensorflow.python.tpu.client.client', 'keras.src.layers.reshaping.zero_padding3d', 'tensorflow.python.platform.build_info', 'opt_einsum.backends.torch', 'pandas.core.window.common', 'keras.api.applications.convnext', 'tensorflow.python.ops.weights_broadcast_ops', 'absl.flags._exceptions', 'pandas._libs.writers', 'tensorflow._api.v2.compat.v2.lookup', 'tensorflow.core.function.polymorphism.function_type', 'tensorflow.python.keras.saving.saved_model.constants', 'scipy.linalg._solve_toeplitz', 'keras.src.activations.activations', 'tensorflow.python.autograph.converters.slices', 'pandas._libs.tslibs', 'tensorflow._api.v2.compat.v2.experimental.extension_type', 'tensorflow._api.v2.compat.v2.types.experimental.distributed', 'keras.src.trainers.data_adapters.tf_dataset_adapter', 'tensorflow.core.protobuf.cluster_pb2', 'cv2.gapi.ov', 'scipy.sparse.sputils', 'tensorflow.python.keras.utils.control_flow_util', 'pandas.util._validators', 'rich.styled', 'ignite.metrics.recall', 'tensorflow.python.framework._proto_comparators', 'keras.src.saving.serialization_lib', 'tensorflow._api.v2.lookup', 'tensorflow.python.ops.linalg.linalg_impl', 'tensorflow._api.v2.compat.v2.__internal__.saved_model', 'tensorflow.core.distributed_runtime.preemption', 'pandas.io.json', 'tensorflow.python.checkpoint.save_util', 'tensorflow._api.v2.compat.v1.math', 'tensorflow._api.v2.compat.v1.config', 'tensorflow.python.trackable.base_delegate', 'tensorflow.python.eager.backprop', 'ignite.metrics.entropy', 'ignite.metrics.precision', 'tensorflow.python.types.data', 'tensorflow._api.v2.compat.v2.lookup.experimental', 'flatbuffers.encode', 'tensorflow.python.platform.googletest', 'tensorflow.python.saved_model.signature_def_utils_impl', 'keras.src.models.functional', 'tensorflow.python.data.ops.iterator_autograph', 'scipy.sparse.linalg._eigen.lobpcg', 'tensorflow._api.v2.distribute.cluster_resolver', 'ignite.metrics.psnr', 'tensorflow.python.training.gradient_descent', 'tensorflow.python.ops.gen_encode_proto_ops', 'tensorflow.python.keras.saving.hdf5_format', 'keras.src.backend.common', 'tensorflow._api.v2.compat.v2.data', 'tensorflow.python.framework.auto_control_deps_utils', 'tensorflow.python.saved_model.function_deserialization', 'tensorflow._api.v2.compat.v2.distribute', 'keras.src.datasets.fashion_mnist', 'tensorflow.python.framework.graph_util', 'tensorflow.python.checkpoint.sharding.sharding_util', 'keras.api.config', 'pandas.core.reshape.melt', 'pandas.core.array_algos.datetimelike_accumulations', 'scipy.sparse.csgraph._tools', 'tensorflow.python.distribute.experimental.dtensor_strategy_extended', 'tensorflow.python.ops.control_flow_case', 'tensorflow._api.v2.compat.v1.saved_model.experimental', 'pandas.core.computation.parsing', 'ignite.utils', 'src.datasets.sunrgbd.sunrgbd', 'tensorflow.python.compat.v2_compat', 'scipy._lib.array_api_compat', 'keras.src.ops.operation_utils', 'tensorflow.python.debug.lib.check_numerics_callback', 'keras.src.utils.audio_dataset_utils', 'tensorflow.python.ops.linalg.linear_operator_block_diag', 'tensorflow.python.eager.cancellation', 'keras.api.quantizers', 'tensorflow.python.platform.analytics', 'scipy.sparse.coo', 'tensorflow.python.keras.layers.recurrent', 'tensorflow.python.ops.ragged.ragged_embedding_ops', 'tensorflow.python.framework.registry', 'tensorflow.python.keras.utils.dataset_creator', 'scipy.sparse.linalg._dsolve._superlu', 'pandas.io.excel._pyxlsb', 'tensorflow.python.debug.wrappers.framework', 'h5py.h5p', 'tensorflow.python.training.queue_runner', 'tensorflow.dtensor.python.layout', 'keras.src.constraints', 'tensorflow.python.training.training', 'tensorflow.python.keras.models', 'keras.src.optimizers.loss_scale_optimizer', 'tensorflow.python.ops.parsing_grad', 'tensorflow.python.autograph.converters.control_flow', 'cv2.aruco', 'keras.src.layers.rnn.conv_lstm3d', 'tensorflow._api.v2.compat.v2.experimental.dlpack', 'keras.api.backend', 'pandas.compat.pickle_compat', 'pandas.core.strings.accessor', 'tensorflow.python.ops.script_ops', 'tensorflow._api.v2.compat.v2.__internal__.distribute.multi_process_runner', 'tensorflow.python.ops.gen_set_ops', 'pandas._libs.testing', 'tensorflow.core.framework.full_type_pb2', 'scipy.sparse.linalg._dsolve._add_newdocs', 'tensorflow.lite.tools', 'tensorflow.python.ops.linalg.linear_operator_toeplitz', 'tensorflow.lite.toco.model_flags_pb2', 'tensorflow.tsl.protobuf.histogram_pb2', 'pandas.core.apply', 'tensorflow.python.eager.polymorphic_function.atomic_function', 'tensorflow.python.ops.ragged.ragged_check_ops', 'cv2.gapi.wip.onevpl', 'tensorflow.python.autograph.pyct.naming', 'tensorflow._api.v2.distribute.coordinator', 'tensorflow.python.ops.embedding_ops', 'cv2.gapi.own.detail', 'tensorflow.python.training', 'keras.src.utils.numerical_utils', 'tensorflow.python.data.ops.options', 'tensorflow.python.ops.distributions.distribution', 'tensorflow.core.tpu', 'tensorflow.python.eager.imperative_grad', 'tensorflow._api.v2.compat.v1.distribute.experimental', 'tensorflow._api.v2.distribute.experimental.partitioners', 'rich.abc', 'keras.src.callbacks', 'tensorflow.python.ops.gen_sparse_ops', 'tensorflow.python.autograph.pyct.static_analysis.activity', 'tensorflow.python.eager.lift_to_graph', 'tensorflow._api.v2.compat.v1.user_ops', 'tensorflow.python.checkpoint.checkpoint', 'tensorflow.core.protobuf.saver_pb2', 'ignite.metrics.gan.fid', 'tensorflow.core.framework.tensor_pb2', 'tensorflow.core.protobuf.tpu.topology_pb2', 'pandas.tseries.api', 'tensorflow.compiler.tf2xla.python', 'tensorflow.lite.python.metrics._pywrap_tensorflow_lite_metrics_wrapper', 'tensorflow._api.v2.compat.v1.config.threading', 'tensorflow.python.checkpoint.graph_view', 'tensorflow.python.ops.distributions.special_math', 'tensorflow.python.profiler.option_builder', 'tensorflow._api.v2.compat.v2.types', 'ignite.handlers.polyaxon_logger', 'tensorflow.python.ops.gen_bitwise_ops', 'pandas.core.groupby.categorical', 'tensorflow.python.compat', 'ignite.metrics.regression.geometric_mean_relative_absolute_error', 'scipy._lib', 'tensorflow.python.training.sync_replicas_optimizer', 'repositories.ESANet.src', 'tensorflow._api.v2.compat.v1.experimental', 'tensorflow.python.trackable.asset', 'tensorflow._api.v2.debugging', 'keras.src.saving.saving_lib', 'tensorflow.python.tpu.ops', 'pandas.util._decorators', 'tensorflow.compiler.mlir.quantization.tensorflow.exported_model_pb2', 'tensorflow.python.training.basic_loops', 'keras.src.layers.attention.additive_attention', 'pandas.core.resample', 'keras.src.layers.merging.concatenate', 'tensorflow.python.util.module_wrapper', 'keras.src.datasets.mnist', 'tensorflow.lite.experimental.microfrontend', 'tensorflow.python.data.ops.structured_function', 'tensorflow.python.keras.saving.saved_model.utils', 'tensorflow.python.distribute.collective_util', 'tensorflow.core.function.capture.restore_captures', 'keras.src.utils.tracking', 'tensorflow._api.v2.compat.v1.spectral', 'tensorflow.python.eager.polymorphic_function.eager_function_run', 'tensorflow.core.function.trace_type.util', 'flatbuffers.number_types', 'tensorflow.core.grappler', 'tensorflow.python.ops.tensor_math_operator_overrides', 'pandas.plotting._misc', 'tensorflow.core', 'tensorflow.python.keras.saving.saved_model.model_serialization', 'rich._cell_widths'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/ESANet\")\n",
    "sys.argv = [\"run.py\"]\n",
    "from repositories.ESANet.src.build_model import build_model\n",
    "from repositories.ESANet.train import parse_args\n",
    "\n",
    "args = parse_args()\n",
    "args.pretrained_on_imagenet = False\n",
    "model, _ = build_model(args, n_classes=40)\n",
    "results[\"nyudv2\"][\"ESANet\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "del model, args\n",
    "\n",
    "args = parse_args()\n",
    "args.pretrained_on_imagenet = False\n",
    "model, _ = build_model(args, n_classes=37)\n",
    "results[\"sunrgbd\"][\"ESANet\"] = cal_flops_params(\n",
    "    model, shape=(480, 640), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")  # Since ESANet use same resolution as NYUDv2\n",
    "del model, args\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
      "  warnings.warn(\n",
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet34_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet34_Weights.DEFAULT` to get the most up-to-date weights.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'repositories.FRNet', 'toolbox.models.FRNet', 'cv2.gapi.oak', 'cv2.misc.version', 'cv2.gapi.imgproc', 'toolbox.models.FRNet.refine', 'cv2.gapi.imgproc.fluid', 'cv2.ipp', 'repositories.FRNet.toolbox.metrics', 'cv2.segmentation', 'cv2.load_config_py3', 'cv2.gapi.onnx.ep', 'cv2.version', 'cv2.gapi.core.fluid', 'cv2.typing', 'cv2.cuda', 'cv2.gapi.video', 'cv2.qt', 'cv2.ogl', 'cv2.gapi.ie.detail', 'cv2.gapi.render.ocv', 'cv2.gapi.ov', 'cv2.gapi.ot.cpu', 'cv2.barcode', 'toolbox.models.FRNet.ResNet', 'cv2.gapi.streaming', 'repositories', 'repositories.FRNet.toolbox.log', 'cv2.gapi.own', 'repositories.FRNet.toolbox.models.FRNet.FRNet', 'toolbox.log', 'repositories.FRNet.toolbox.utils', 'cv2.gapi.render', 'cv2.parallel', 'cv2.utils', 'cv2.aruco', 'cv2.gapi.wip.gst', 'zmq.utils.garbage', 'repositories.FRNet.toolbox.models', 'cv2.gapi.wip', 'cv2.Error', 'cv2.samples', 'toolbox.utils', 'cv2.gapi.ie', 'cv2.ocl', 'cv2.gapi.core', 'cv2.gapi.onnx', 'cv2.gapi', 'cv2.dnn', 'cv2.misc', 'cv2.utils.fs', 'repositories.FRNet.toolbox', 'cv2.gapi.wip.onevpl', 'cv2.mat_wrapper', 'cv2.fisheye', 'cv2.gapi.own.detail', 'cv2.utils.nested', 'cv2.ml', 'cv2.gapi.core.cpu', 'cv2.gapi.wip.draw', 'toolbox.models.FRNet.fusion', 'cv2.detail', 'cv2.data', 'cv2.gapi.core.ocl', 'cv2.flann', 'cv2.videoio_registry', 'cv2.gapi.ot', 'toolbox.models', 'cv2', 'toolbox.metrics', 'toolbox', 'repositories.FRNet.toolbox.models.FRNet'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/FRNet\")\n",
    "from repositories.FRNet.toolbox.models.FRNet.FRNet import FRNet\n",
    "\n",
    "model = FRNet(n_class=40).eval()\n",
    "results[\"nyudv2\"][\"FRNet\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = FRNet(n_class=37).eval()\n",
    "results[\"sunrgbd\"][\"FRNet\"] = cal_flops_params(\n",
    "    model, shape=(544, 736), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")  # Since FRNet only accepts resolution of multiple of 32(2^5), we use 544x736\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----------------Model Params--------------------------------------\n",
      "backbone: mit_b3\n",
      "dpr: 0.4\n",
      "--------------------------------------------------------------\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/calflops/pytorch_ops.py:364: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.\n",
      "  return oldFunc(*args, **kwds)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "-----------------Model Params--------------------------------------\n",
      "backbone: mit_b5\n",
      "dpr: 0.4\n",
      "--------------------------------------------------------------\n",
      "-----------------Model Params--------------------------------------\n",
      "backbone: mit_b3\n",
      "dpr: 0.4\n",
      "--------------------------------------------------------------\n",
      "-----------------Model Params--------------------------------------\n",
      "backbone: mit_b5\n",
      "dpr: 0.4\n",
      "--------------------------------------------------------------\n",
      "Deleted module: {'mmcv.cnn.bricks.depthwise_separable_conv_module', 'yapf_third_party._ylib2to3.pgen2.pgen', 'mmengine.optim.optimizer.base', 'mmcv.visualization', 'mmengine.optim.scheduler', 'cv2.cuda', 'cv2.qt', 'mmengine._strategy.deepspeed', 'mmengine.optim.scheduler.momentum_scheduler', 'mmengine.runner', 'repositories', 'mmcv', 'mmengine.registry.build_functions', 'cv2.gapi.wip.gst', 'mmengine.fileio.backends', 'cv2.samples', 'importlib_metadata.compat', 'rich.table', 'mmcv.cnn.bricks.norm', 'rich.box', 'mmengine.optim.optimizer.builder', 'yapf_third_party._ylib2to3.patcomp', 'mmcv.cnn', 'mmcv.visualization.image', 'attr._config', 'mmengine._strategy.fsdp', 'rich', 'mmengine.utils.misc', 'mmcv.cnn.bricks.drop', 'yapf_third_party._ylib2to3.pgen2.tokenize', 'torch.distributed.optim.functional_adamax', 'cv2.gapi.core.fluid', 'mmengine._strategy', 'mmengine.hooks.runtime_info_hook', 'mmengine.utils.path', 'zipp', 'rich._export_format', 'rich._palettes', 'mmcv.transforms.base', 'cv2.gapi.wip', 'torch.distributed.optim.zero_redundancy_optimizer', 'cv2.misc', 'rich.live', 'rich.protocol', 'mmengine.optim.optimizer.amp_optimizer_wrapper', 'mmcv.transforms.builder', 'attr.converters', 'mmengine.fileio.io', 'yapf_third_party._ylib2to3.fixer_util', 'mmengine.fileio.backends.registry_utils', 'mmcv_custom', 'mmengine.registry.default_scope', 'yapf.pyparser.pyparser_utils', 'mmcv.cnn.rfsearch', 'cv2.parallel', 'mmengine.utils.dl_utils.setup_env', 'mmengine.visualization.vis_backend', 'mmengine.model.wrappers.distributed', 'yapf_third_party._ylib2to3.pgen2.parse', 'mmengine.model.wrappers.seperate_distributed', 'cv2.utils.fs', 'mmengine.registry', 'cv2.mat_wrapper', 'mmengine.runner.base_loop', 'torch.distributed.optim.functional_rprop', 'yapf.pytree.comment_splicer', 'mmengine.utils.dl_utils.parrots_wrapper', 'mmengine.dist', 'yapf.pytree.pytree_utils', 'addict.addict', 'rich.file_proxy', 'termcolor', 'mmengine.structures.base_data_element', 'cv2.gapi.imgproc', 'rich._loop', 'mmengine.runner.loops', 'yapf.pytree', 'rich.filesize', 'rich._fileno', 'yapf.yapflib.format_token', 'rich.pretty', 'importlib_metadata._meta', 'yapf.yapflib.identify_container', 'rich.containers', 'mmengine.device', 'yapf.yapflib.file_resources', 'cv2.gapi.core.cpu', 'mmengine.model.weight_init', 'mmengine.device.utils', 'mmengine.optim.optimizer.zero_optimizer', 'cv2.gapi.core.ocl', 'mmengine.runner.priority', 'mmcv.image.photometric', 'mmengine.fileio', 'mmengine.hooks.naive_visualization_hook', 'cv2.gapi.oak', 'yapf.yapflib.line_joiner', 'rich.errors', 'rich.markup', 'mmengine.structures.instance_data', 'importlib_metadata.compat.py39', 'yapf.yapflib', 'mmengine.utils.version_utils', 'mmcv_custom.checkpoint', 'rich._wrap', 'yapf.pytree.blank_line_calculator', 'mmengine.evaluator', 'cv2.gapi.streaming', 'mmcv.transforms.formatting', 'mmengine.hooks.early_stopping_hook', 'mmcv.cnn.bricks.generalized_attention', 'mmcv.arraymisc', 'mmengine.dataset.utils', 'rich.cells', 'yapf_third_party._ylib2to3', 'mmengine.model.wrappers.utils', 'mmengine.fileio.handlers.pickle_handler', 'mmcv.video', 'mmengine.visualization', 'mmcv.cnn.bricks.upsample', 'mmengine.structures', 'torch.distributed.optim.functional_adadelta', 'torch.distributed.optim.named_optimizer', 'rich.region', 'torch.distributed.algorithms.model_averaging', 'mmengine.model.averaged_model', 'rich.measure', 'cv2.segmentation', 'rich._emoji_replace', 'mmcv.cnn.resnet', 'yapf.pytree.subtype_assigner', 'rich.highlighter', 'yapf.pyparser.pyparser', 'mmcv.transforms.loading', 'yapf_third_party._ylib2to3.pgen2.driver', 'mmcv.cnn.bricks.conv_ws', 'repositories.GeminiFusion.models.swin_transformer', 'importlib_metadata._functools', 'mmengine.hooks.empty_cache_hook', 'tifffile._imagecodecs', 'mmengine.hooks.sampler_seed_hook', 'mmengine.runner.checkpoint', 'addict', 'cv2.flann', 'cv2.gapi.ot', 'mmengine.utils.dl_utils.time_counter', 'rich.default_styles', 'mmcv.cnn.utils.flops_counter', 'rich.themes', 'torch.distributed.algorithms.model_averaging.utils', 'importlib_metadata._text', 'mmcv.cnn.bricks.swish', 'mmengine.fileio.handlers.yaml_handler', 'tifffile.tifffile', 'mmcv.transforms.wrappers', 'mmengine.registry.registry', 'attr.validators', 'cv2.gapi.own', 'mmengine.evaluator.metric', 'mmengine.model.base_model.data_preprocessor', 'mmengine.registry.root', 'attr', 'mmengine.model.efficient_conv_bn_eval', 'mmcv.visualization.optflow', 'cv2.gapi.core', 'mmengine.dataset.sampler', 'cv2.fisheye', 'mmengine.optim', 'rich.spinner', 'mmengine.fileio.backends.local_backend', 'rich.pager', 'mmengine.fileio.handlers.base', 'mmengine.logging.logger', 'cv2.videoio_registry', 'mmengine.dist.utils', 'mmengine', 'mmengine.hooks.profiler_hook', 'mmengine.config.lazy', 'cv2.version', 'mmengine.fileio.backends.http_backend', 'mmengine.model.utils', 'mmcv.image.misc', 'yapf_third_party._ylib2to3.pytree', 'mmcv.version', 'rich.live_render', 'cv2.utils', 'mmengine.utils.timer', 'mmengine.model.base_module', 'mmcv.cnn.bricks.conv2d_adaptive_padding', 'yapf.pyparser', 'cv2.gapi', 'mmengine.fileio.handlers', 'attr._cmp', 'rich.repr', 'mmengine._strategy.base', 'mmengine.logging.message_hub', 'rich.segment', 'cv2.ipp', 'cv2.load_config_py3', 'mmengine.model', 'mmengine.utils.dl_utils.collect_env', 'attr.setters', 'mmengine.model.test_time_aug', 'rich.scope', 'rich._ratio', 'rich.console', 'mmengine.hooks.checkpoint_hook', 'mmcv.cnn.bricks.scale', 'mmcv.cnn.rfsearch.utils', 'torch.distributed.optim.functional_sgd', 'yapf.pytree.split_penalty', 'cv2.barcode', 'mmcv.cnn.bricks.non_local', 'mmcv.cnn.bricks.padding', 'mmengine.hooks.ema_hook', 'rich.terminal_theme', 'mmengine.fileio.file_client', 'cv2.Error', 'mmcv.image.io', 'cv2.gapi.ie', 'mmcv.video.optflow', 'attr._funcs', 'rich.color_triplet', 'mmengine.utils.package_utils', 'rich.screen', 'mmengine.model.wrappers', 'rich._extension', 'importlib_metadata._compat', 'mmcv.video.processing', 'mmengine.fileio.handlers.registry_utils', 'mmengine.version', 'mmengine.optim.scheduler.lr_scheduler', 'mmengine.optim.optimizer.optimizer_wrapper_dict', 'cv2.gapi.ie.detail', 'zipp.glob', 'rich.ansi', 'yapf_third_party._ylib2to3.pygram', 'rich.constrain', 'mmengine.runner._flexible_runner', 'mmcv.image', 'mmengine.evaluator.utils', 'mmengine.dataset', 'yapf.pytree.continuation_splicer', 'cv2.gapi.onnx', 'mmengine.hooks.logger_hook', 'rich.emoji', 'yapf', 'mmengine._strategy.utils', 'cv2.data', 'mmengine.config', 'mmengine.utils.dl_utils', 'mmcv.cnn.bricks.hswish', 'mmengine.fileio.backends.memcached_backend', 'cv2.gapi.onnx.ep', 'cv2.typing', 'yapf.yapflib.format_decision_state', 'mmengine.utils.dl_utils.torch_ops', 'mmengine.runner.log_processor', 'yapf.yapflib.subtypes', 'yapf.yapflib.yapf_api', 'mmengine._strategy.colossalai', 'rich.palette', 'mmengine.utils.dl_utils.hub', 'mmengine.utils', 'mmengine.model.wrappers.fully_sharded_distributed', 'mmcv.cnn.bricks.hsigmoid', 'cv2.ocl', 'mmengine.utils.dl_utils.misc', 'torch.distributed.optim.post_localSGD_optimizer', 'torch.distributed.optim.functional_adagrad', 'mmengine.fileio.backends.base', 'mmcv.image.geometric', 'timm.models.layers', 'mmcv.cnn.vgg', 'cv2.detail', 'mmengine.visualization.visualizer', 'mmcv.cnn.bricks.conv', 'termcolor.termcolor', 'mmengine.structures.label_data', 'rich._emoji_codes', 'cv2.misc.version', 'mmengine.utils.dl_utils.trace', 'rich.style', 'mmengine.optim.optimizer.apex_optimizer_wrapper', 'rich.padding', 'mmengine.fileio.handlers.json_handler', 'rich.jupyter', 'mmengine.optim.optimizer', 'torch.distributed.optim.functional_adam', 'mmengine.dist.dist', 'torch.distributed.optim', 'rich.progress', 'mmcv.cnn.rfsearch.operator', 'rich.panel', 'yapf.yapflib.errors', 'yapf_third_party._ylib2to3.pgen2.literals', 'mmengine.logging', 'cv2.dnn', 'yapf.pyparser.split_penalty_visitor', 'mmengine.runner.runner', 'attr._make', 'mmcv.transforms.utils', 'yapf.pytree.pytree_visitor', 'mmengine.utils.progressbar', 'mmcv.cnn.bricks.plugin', 'mmcv.transforms', 'attr.exceptions', 'mmengine.dataset.base_dataset', 'attr._version_info', 'mmengine.logging.history_buffer', 'repositories.GeminiFusion.models', 'yapf.yapflib.reformatter', 'torch.distributed.optim.apply_optimizer_in_backward', 'cv2.gapi.ov', 'torch.distributed.algorithms.model_averaging.averagers', 'rich.styled', 'mmengine._strategy.distributed', 'mmcv.video.io', 'zipp.compat.py310', 'mmengine.model.base_model.base_model', 'yapf_third_party._ylib2to3.pgen2.token', 'mmengine.registry.utils', 'torch.distributed.optim.utils', 'importlib_metadata._collections', 'mmengine.config.utils', 'mmengine.utils.manager', 'mmengine.fileio.parse', 'mmcv.image.colorspace', 'mmengine.structures.pixel_data', 'repositories.GeminiFusion.models.mix_transformer', 'cv2.ogl', 'yapf.yapflib.style', 'mmengine.utils.progressbar_rich', 'termcolor._types', 'rich._spinners', 'cv2.gapi.render.ocv', 'mmcv.cnn.bricks', 'mmcv.cnn.utils', 'yapf_third_party._ylib2to3.pgen2.grammar', 'rich._null_file', 'mmengine.dataset.dataset_wrapper', 'rich.progress_bar', 'mmengine.optim.scheduler.param_scheduler', 'mmcv.visualization.color', 'repositories.GeminiFusion.models.modules', 'mmengine.runner.activation_checkpointing', 'cv2.utils.nested', 'mmengine.hooks.hook', 'mmengine.hooks.test_time_aug_hook', 'mmcv.cnn.bricks.wrappers', 'cv2', 'mmengine.hooks', 'yapf_third_party', 'mmengine.config.config', 'mmcv.cnn.rfsearch.search', 'torch.distributed.optim.functional_rmsprop', 'yapf.pytree.pytree_unwrapper', 'mmengine.runner.amp', 'mmengine.optim.optimizer.optimizer_wrapper', 'repositories.GeminiFusion', 'mmcv.cnn.bricks.conv_module', 'mmcv.cnn.bricks.context_block', 'mmengine.fileio.backends.lmdb_backend', 'yapf.yapflib.split_penalty', 'cv2.gapi.render', 'rich.text', 'cv2.aruco', 'rich._log_render', 'mmengine.model.base_model', 'zipp.compat', 'mmengine.visualization.utils', 'repositories.GeminiFusion.models.segformer', 'cv2.gapi.wip.onevpl', 'mmengine.hooks.iter_timer_hook', 'importlib_metadata._adapters', 'cv2.gapi.own.detail', 'cv2.gapi.wip.draw', 'cv2.ml', 'mmengine.hooks.param_scheduler_hook', 'importlib_metadata._itertools', 'mmcv.cnn.utils.fuse_conv_bn', 'rich.abc', 'torch.distributed.optim.optimizer', 'mmcv.transforms.processing', 'cv2.gapi.imgproc.fluid', 'attr._next_gen', 'mmengine.runner.utils', 'attr._compat', 'mmengine.hooks.sync_buffer_hook', 'importlib_metadata', 'cv2.gapi.video', 'rich.theme', 'yapf.yapflib.logical_line', 'yapf.yapflib.object_state', 'mmcv.cnn.bricks.activation', 'mmengine.optim.optimizer.default_constructor', 'cv2.gapi.ot.cpu', 'torch.distributed.optim.functional_adamw', 'mmcv.arraymisc.quantization', 'yapf_third_party._ylib2to3.pgen2', 'attr.filters', 'rich.control', 'mmcv.cnn.alexnet', 'rich.align', 'mmengine.evaluator.evaluator', 'mmengine.fileio.backends.petrel_backend', 'tifffile', 'rich.color', 'mmengine._strategy.single_device', 'rich._pick', 'rich._cell_widths'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/GeminiFusion\")\n",
    "from repositories.GeminiFusion.models.segformer import WeTr\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    WeTr(backbone=\"mit_b3\", num_classes=40, n_heads=8, dpr=0.4, drop_rate=0.0).eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"GeminiFusion-B3\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    WeTr(backbone=\"mit_b5\", num_classes=40, n_heads=8, dpr=0.4, drop_rate=0.0).eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"GeminiFusion-B5\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    WeTr(backbone=\"mit_b3\", num_classes=37, n_heads=8, dpr=0.4, drop_rate=0.0).eval()\n",
    ")\n",
    "results[\"sunrgbd\"][\"GeminiFusion-B3\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(\n",
    "    WeTr(backbone=\"mit_b5\", num_classes=37, n_heads=8, dpr=0.4, drop_rate=0.0).eval()\n",
    ")\n",
    "results[\"sunrgbd\"][\"GeminiFusion-B5\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "albumentations not installed\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/torch/functional.py:504: UserWarning: torch.meshgrid: in an upcoming release, it will be required to pass the indexing argument. (Triggered internally at /opt/conda/conda-bld/pytorch_1702400366987/work/aten/src/ATen/native/TensorShape.cpp:3526.)\n",
      "  return _VF.meshgrid(tensors, **kwargs)  # type: ignore[attr-defined]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'pandas.io.parquet', 'pandas.core.array_algos', 'utils.cross_entropy', 'pandas._libs.tslibs.strptime', 'pandas.core.sorting', 'pandas._libs.tslibs.fields', 'cv2.cuda', 'pandas._config.localization', 'cv2.qt', 'pandas.core.arrays.timedeltas', 'opt_einsum.backends.jax', 'pandas.core.strings', 'pandas.core.util.numba_', 'pandas.io.excel._base', 'pandas.core.tools.timedeltas', 'pandas._libs.tslibs.parsing', 'repositories', 'pandas.core.dtypes.generic', 'repositories.MultiMAE.utils.checkpoint', 'pandas.core.arrays._utils', 'cv2.gapi.wip.gst', 'cv2.samples', 'pandas._config.dates', 'pandas.io.excel', 'pandas.core.internals.ops', 'pandas.core.ops', 'pandas.core.methods.selectn', 'pandas.core.api', 'pandas.core.internals', 'pandas.core.array_algos.quantile', 'pandas.core.ops.docstrings', 'pandas.io.excel._xlrd', 'pandas.core.dtypes.missing', 'pandas.core.indexing', 'pandas.core.indexes.multi', 'pandas.io.formats.printing', 'pandas.compat.pyarrow', 'pandas.core.reshape.encoding', 'pandas.core.arrays.arrow.array', 'cv2.gapi.core.fluid', 'pandas.core.indexes.base', 'utils', 'pandas.io.xml', 'repositories.MultiMAE.utils', 'repositories.MultiMAE.multimae.criterion', 'pandas.core.dtypes.dtypes', 'pandas.core.dtypes', 'pandas.testing', 'pandas.core.arrays.period', 'pandas._libs.sparse', 'pandas._libs.ops_dispatch', 'cv2.gapi.wip', 'pandas.io.orc', 'pandas.core.arrays.base', 'opt_einsum.backends', 'pandas.compat._constants', 'pandas._testing._io', 'opt_einsum.helpers', 'cv2.misc', 'pandas._libs.tslibs.timedeltas', 'pandas._libs.tslibs.offsets', 'utils.model_builder', 'pandas._libs.tslib', 'pandas.core.arrays._mixins', 'pandas.core.frame', 'pandas.core.tools.numeric', 'pandas.util._print_versions', 'pandas.compat._optional', 'einops.einops', 'pandas.core.indexes.datetimes', 'pandas.core.missing', 'pandas.core.indexes.accessors', 'utils.taskonomy.task_configs', 'cv2.parallel', 'pandas.core.array_algos.masked_reductions', 'pandas.api.extensions', 'pandas._libs.tslibs.timestamps', 'pandas.core._numba.executor', 'utils.task_balancing', 'cv2.utils.fs', 'pandas.core.arrays.datetimes', 'cv2.mat_wrapper', 'pandas.core.algorithms', 'pandas._testing._warnings', 'repositories.MultiMAE.utils.auto_augment', 'pandas.core.window.rolling', 'pandas._libs.pandas_datetime', 'pandas.io.stata', 'pandas._libs.tslibs.nattype', 'pandas.core.computation.common', 'pandas.core.strings.object_array', 'pandas._libs.algos', 'repositories.MultiMAE.multimae.multimae', 'cv2.gapi.imgproc', 'pandas.core.sample', 'utils.metrics', 'pandas.plotting', 'pandas.io.gbq', 'pandas.core.tools.times', 'pandas.core.indexes.extension', 'pandas.core.arrays._arrow_string_mixins', 'pandas.core.common', 'repositories.MultiMAE.utils.data_constants', 'pandas._typing', 'pandas.core.reshape.tile', 'opt_einsum.paths', 'pandas._testing', 'pandas.core.util.hashing', 'pandas.core.window.doc', 'pandas.core.dtypes.astype', 'utils.random_erasing', 'cv2.gapi.core.cpu', 'pandas._libs.tslibs.np_datetime', 'pandas.core.reshape.concat', 'utils.taskonomy.taskonomy_dataset', 'utils.dataset_folder', 'repositories.MultiMAE.utils.dist', 'cv2.gapi.core.ocl', 'pandas.compat', 'cv2.gapi.oak', 'multimae', 'multimae.criterion', 'pandas._libs.tslibs.vectorized', 'einops._backends', 'pandas.core.computation', 'utils.semseg_metrics', 'opt_einsum.parser', 'pandas._libs.interval', 'pandas.core.internals.concat', 'pandas.core.groupby', 'repositories.MultiMAE.utils.registry', 'pandas.core._numba', 'opt_einsum.backends.theano', 'cv2.gapi.streaming', 'pandas._libs.groupby', 'pandas.core.reshape', 'pandas.core.internals.managers', 'pandas.api.indexers', 'pandas.io.sql', 'pandas.util._exceptions', 'pandas.core.indexes.frozen', 'repositories.MultiMAE', 'opt_einsum', 'pandas.util', 'pandas.core.arrays', 'pandas.io.parsers.c_parser_wrapper', 'pandas.core.series', 'pandas.io.excel._xlsxwriter', 'pandas.core.groupby.grouper', 'pandas.core.interchange.utils', 'pandas.io.pytables', 'pandas.io.parsers', 'repositories.MultiMAE.utils.taskonomy', 'pandas.core.arrays.integer', 'cv2.segmentation', 'pandas._libs.tslibs.tzconversion', 'pandas._libs.json', 'pandas.core.window.online', 'pytz.exceptions', 'pandas.tseries.frequencies', 'pandas.core.groupby.indexing', 'pandas.core.ops.mask_ops', 'pandas.core.internals.construction', 'pandas.core.groupby.groupby', 'pandas.io.spss', 'pandas.core.arrays.categorical', 'repositories.MultiMAE.utils.optim_factory', 'pandas.io.formats.format', 'pandas.core.array_algos.replace', 'repositories.MultiMAE.utils.taskonomy.taskonomy_dataset', 'pandas.core.construction', 'repositories.MultiMAE.run_finetuning_semseg', 'pandas.core.window', 'cv2.flann', 'cv2.gapi.ot', 'pandas.api.types', 'opt_einsum.blas', 'pandas.api.typing', 'pandas.core.interchange', 'pandas._libs.tslibs.conversion', 'pandas.core.indexes', 'pandas._libs.properties', 'pandas.core.arrays.numpy_', 'pandas.util.version', 'pandas.core.internals.array_manager', 'multimae.input_adapters', 'pandas._libs.window.aggregations', 'utils.datasets_semseg', 'cv2.gapi.own', 'repositories.MultiMAE.utils.cross_entropy', 'pytz.tzinfo', 'einops._torch_specific', 'pandas.compat.numpy', 'opt_einsum.backends.object_arrays', 'cv2.gapi.core', 'pandas.core.dtypes.base', 'pandas.core.arrays.sparse.accessor', 'pandas._version_meson', 'pandas.core.arraylike', 'cv2.fisheye', 'pandas._libs.parsers', 'pandas.arrays', 'utils.taskonomy', 'pandas.io.json._table_schema', 'pandas.core.arrays.masked', 'pandas.io.parsers.readers', 'cv2.videoio_registry', 'repositories.MultiMAE.utils.model_builder', 'pandas.core.internals.blocks', 'pandas.core.arrays.sparse.array', 'pandas.core.internals.api', 'utils.transforms_factory', 'cv2.version', 'pytz.tzfile', 'pandas.core.computation.align', 'pandas._libs.lib', 'pandas.core.accessor', 'pandas._libs.pandas_parser', 'pandas.core.strings.base', 'cv2.utils', 'repositories.MultiMAE.utils.taskonomy.transforms', 'pandas._libs', 'pandas._libs.indexing', 'utils.logger', 'cv2.gapi', 'pandas.core.computation.expressions', 'pandas.io.formats.info', 'pandas._libs.window.indexers', 'pandas._libs.reshape', 'pandas.io.html', 'pandas.core.arrays.sparse', 'pandas.core.arrays.numeric', 'pandas.io.formats', 'cv2.ipp', 'cv2.load_config_py3', 'pandas.core.computation.eval', 'pandas.core.config_init', 'cv2.barcode', 'pandas.core.arrays.boolean', 'repositories.MultiMAE.utils.mixup', 'pandas.plotting._core', 'pandas._testing.compat', 'cv2.Error', 'pandas.core.ops.dispatch', 'repositories.MultiMAE.utils.transforms_factory', 'cv2.gapi.ie', 'utils.log_images', 'opt_einsum._version', 'pandas._libs.tslibs.period', 'utils.model', 'pandas.core.window.expanding', 'multimae.multimae_utils', 'opt_einsum.backends.cupy', 'utils.transforms', 'pandas.core.indexes.category', 'einops.packing', 'utils.mixup', 'pandas.core.indexes.range', 'pandas.core.array_algos.putmask', 'pandas.core.tools', 'pandas.core.arrays.arrow', 'cv2.gapi.ie.detail', 'pandas._libs.missing', 'pandas.compat.numpy.function', 'pandas.tseries', 'pandas.core.arrays.floating', 'pandas.api.interchange', 'pandas.io.excel._odswriter', 'pandas.core.interchange.dataframe_protocol', 'cv2.gapi.onnx', 'repositories.MultiMAE.utils.transforms', 'utils.checkpoint', 'utils.pos_embed', 'pandas.io.sas', 'pandas.api', 'pandas.core.dtypes.inference', 'pandas.core.roperator', 'pandas.io._util', 'cv2.data', 'pandas.io.pickle', 'pandas._libs.index', 'repositories.MultiMAE.utils.model', 'pandas._config.display', 'torch.backends.opt_einsum', 'pandas.core.indexes.api', 'multimae.output_adapters', 'pandas.core.indexes.period', 'cv2.gapi.onnx.ep', 'pandas.core.computation.engines', 'cv2.typing', 'utils.auto_augment', 'pandas.core.groupby.numba_', 'utils.model_ema', 'utils.native_scaler', 'pandas.core.ops.missing', 'pandas._libs.tslibs.timezones', 'pandas.core.computation.scope', 'pandas.core.computation.api', 'repositories.MultiMAE.utils.random_erasing', 'pandas.io.excel._openpyxl', 'opt_einsum.sharing', 'pandas.core.ops.common', 'opt_einsum.backends.dispatch', 'cv2.ocl', 'pandas._libs.ops', 'pandas._config', 'pandas.tseries.offsets', 'pandas.core.indexers', 'einops.parsing', 'pandas.core.reshape.pivot', 'cv2.detail', 'pandas._libs.tslibs.dtypes', 'pandas._libs.internals', 'utils.registry', 'pandas._testing.contexts', 'cv2.misc.version', 'pandas.core.arrays.arrow.accessors', 'repositories.MultiMAE.multimae.multimae_utils', 'pandas._libs.arrays', 'repositories.MultiMAE.multimae.output_adapters', 'pandas.io', 'pandas.core.interchange.from_dataframe', 'pandas.core.indexers.utils', 'pandas.io.json._json', 'repositories.MultiMAE.multimae.input_adapters', 'pandas.core.arrays.string_arrow', 'pandas._config.config', 'pandas.core.groupby.generic', 'pandas._libs.hashing', 'cv2.dnn', 'pytz.lazy', 'pandas', 'pandas.core.shared_docs', 'pandas.core.computation.expr', 'pandas.io.json._normalize', 'pandas.io.parsers.arrow_parser_wrapper', 'pandas.io.excel._calamine', 'opt_einsum.backends.torch', 'pandas.core.window.common', 'opt_einsum.backends.tensorflow', 'pandas._libs.writers', 'pandas.core.reshape.merge', 'pandas.io.formats.console', 'utils.data_constants', 'pandas.io.sas.sasreader', 'pandas._libs.tslibs', 'pandas.core.window.numba_', 'multimae.multimae', 'pandas.core.tools.datetimes', 'cv2.gapi.ov', 'pandas.io.excel._odfreader', 'pandas.util._validators', 'pandas.core.reshape.api', 'pandas.core.window.ewm', 'opt_einsum.path_random', 'pandas.core.dtypes.concat', 'pandas.io.json', 'pandas.core.util', 'pandas.core.groupby.ops', 'pandas.core.indexes.interval', 'pandas.core.ops.array_ops', 'utils.dist', 'opt_einsum.contract', 'pandas._testing.asserters', 'pandas.core.arrays.datetimelike', 'pandas.core.flags', 'cv2.ogl', 'pandas.core.dtypes.cast', 'pandas.core.arrays._ranges', 'cv2.gapi.render.ocv', 'pandas._libs.window', 'pandas.core.reshape.util', 'pandas.core.reshape.melt', 'pandas.core.array_algos.datetimelike_accumulations', 'repositories.MultiMAE.utils.model_ema', 'pandas.util._tester', 'pandas.core.generic', 'pandas.core.dtypes.api', 'repositories.MultiMAE.utils.logger', 'pandas.core.computation.parsing', 'pandas.core.internals.base', 'cv2.utils.nested', 'repositories.MultiMAE.utils.native_scaler', 'cv2', 'pandas.core.nanops', 'pandas.io.excel._pyxlsb', 'pandas.core.array_algos.masked_accumulations', 'pandas.errors', 'multimae.output_adapter_utils', 'pandas.core.array_algos.transforms', 'cv2.gapi.render', 'cv2.aruco', 'pandas.core.arrays.interval', 'pandas.compat.pickle_compat', 'pandas.core.strings.accessor', 'pandas.core.arrays.string_', 'pandas.core.indexers.objects', 'pandas.io.clipboards', 'pandas._libs.testing', 'pandas.core.apply', 'repositories.MultiMAE.utils.task_balancing', 'cv2.gapi.wip.onevpl', 'pandas.core.groupby.base', 'pytz', 'cv2.gapi.own.detail', 'cv2.gapi.wip.draw', 'pandas.core.indexes.datetimelike', 'pandas.core', 'cv2.ml', 'repositories.MultiMAE.multimae', 'pandas.core.methods', 'pandas.compat.compressors', 'einops', 'pandas.core.array_algos.take', 'cv2.gapi.imgproc.fluid', 'pandas.io.api', 'pandas.tseries.api', 'pandas.core.computation.ops', 'cv2.gapi.video', 'pandas.core.computation.check', 'pandas.io.parsers.base_parser', 'utils.optim_factory', 'utils.taskonomy.transforms', 'repositories.MultiMAE.multimae.output_adapter_utils', 'pandas.core.groupby.categorical', 'pandas.io.feather_format', 'cv2.gapi.ot.cpu', 'pandas.core.base', 'pandas._libs.tslibs.ccalendar', 'pandas.util._decorators', 'pandas._libs.join', 'pandas.core.resample', 'repositories.MultiMAE.utils.taskonomy.task_configs', 'pandas.core.indexes.timedeltas', 'pandas.io.excel._util', 'pandas.core.ops.invalid', 'repositories.MultiMAE.utils.metrics', 'pandas._libs.hashtable', 'pandas.io.common', 'pandas.io.parsers.python_parser', 'pandas.plotting._misc', 'pandas.core.methods.describe', 'pandas.core.computation.pytables', 'pandas.core.dtypes.common'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/MultiMAE\")\n",
    "sys.argv = [\n",
    "    \"run.py\",\n",
    "    \"-c\",\n",
    "    \"repositories/MultiMAE/cfgs/finetune/semseg/nyu/ft_nyu_200e_multimae-b_rgb-depth.yaml\",\n",
    "]\n",
    "\n",
    "from repositories.MultiMAE.multimae import multimae\n",
    "from repositories.MultiMAE.multimae.input_adapters import (\n",
    "    PatchedInputAdapter,\n",
    "    SemSegInputAdapter,\n",
    ")\n",
    "from repositories.MultiMAE.utils.registry import register_model\n",
    "from repositories.MultiMAE.utils import create_model\n",
    "from repositories.MultiMAE.utils.data_constants import COCO_SEMSEG_NUM_CLASSES\n",
    "from functools import partial\n",
    "from repositories.MultiMAE.multimae.output_adapters import (\n",
    "    ConvNeXtAdapter,\n",
    "    DPTOutputAdapter,\n",
    "    SegmenterMaskTransformerAdapter,\n",
    ")\n",
    "from repositories.MultiMAE.run_finetuning_semseg import get_args\n",
    "\n",
    "register_model(multimae.multivit_base)\n",
    "args = get_args()\n",
    "args.in_domains = args.in_domains.split(\"-\")\n",
    "args.num_classes_with_void = (\n",
    "    args.num_classes + 1 if args.seg_use_void_label else args.num_classes\n",
    ")\n",
    "DOMAIN_CONF = {\n",
    "    \"rgb\": {\n",
    "        \"channels\": 3,\n",
    "        \"stride_level\": 1,\n",
    "        \"aug_type\": \"image\",\n",
    "        \"input_adapter\": partial(PatchedInputAdapter, num_channels=3),\n",
    "    },\n",
    "    \"depth\": {\n",
    "        \"channels\": 1,\n",
    "        \"stride_level\": 1,\n",
    "        \"aug_type\": \"mask\",\n",
    "        \"input_adapter\": partial(PatchedInputAdapter, num_channels=1),\n",
    "    },\n",
    "    \"semseg\": {\n",
    "        \"stride_level\": 4,\n",
    "        \"aug_type\": \"mask\",\n",
    "        \"input_adapter\": partial(\n",
    "            SemSegInputAdapter,\n",
    "            num_classes=COCO_SEMSEG_NUM_CLASSES,\n",
    "            dim_class_emb=64,\n",
    "            interpolate_class_emb=False,\n",
    "            emb_padding_idx=COCO_SEMSEG_NUM_CLASSES,\n",
    "        ),\n",
    "    },\n",
    "    \"pseudo_semseg\": {\"aug_type\": \"mask\"},\n",
    "    \"mask_valid\": {\n",
    "        \"stride_level\": 1,\n",
    "        \"aug_type\": \"mask\",\n",
    "    },\n",
    "}\n",
    "input_adapters = {\n",
    "    domain: DOMAIN_CONF[domain][\"input_adapter\"](\n",
    "        stride_level=DOMAIN_CONF[domain][\"stride_level\"],\n",
    "        patch_size_full=args.patch_size,\n",
    "        image_size=args.input_size,\n",
    "        learnable_pos_emb=args.learnable_pos_emb,\n",
    "    )\n",
    "    for domain in args.in_domains\n",
    "}\n",
    "adapters_dict = {\n",
    "    \"segmenter\": partial(\n",
    "        SegmenterMaskTransformerAdapter,\n",
    "        depth=args.decoder_depth,\n",
    "        drop_path_rate=args.drop_path_decoder,\n",
    "    ),\n",
    "    \"convnext\": partial(\n",
    "        ConvNeXtAdapter,\n",
    "        preds_per_patch=args.decoder_preds_per_patch,\n",
    "        depth=args.decoder_depth,\n",
    "        interpolate_mode=args.decoder_interpolate_mode,\n",
    "        main_tasks=args.decoder_main_tasks.split(\"-\"),\n",
    "    ),\n",
    "    \"dpt\": partial(\n",
    "        DPTOutputAdapter,\n",
    "        stride_level=1,\n",
    "        main_tasks=args.decoder_main_tasks.split(\"-\"),\n",
    "        head_type=\"semseg\",\n",
    "    ),\n",
    "}\n",
    "\n",
    "output_adapters = {\n",
    "    \"semseg\": adapters_dict[args.output_adapter](\n",
    "        num_classes=args.num_classes_with_void,\n",
    "        embed_dim=args.decoder_dim,\n",
    "        patch_size=args.patch_size,\n",
    "    ),\n",
    "}\n",
    "model = decorate_model_for_dict_input(\n",
    "    create_model(\n",
    "        \"multivit_base\",\n",
    "        input_adapters=input_adapters,\n",
    "        output_adapters=output_adapters,\n",
    "        drop_path_rate=args.drop_path_encoder,\n",
    "    ).eval(),\n",
    "    [\"rgb\", \"depth\"],\n",
    ")\n",
    "results[\"nyudv2\"][\"MultiMAE\"] = cal_flops_params(\n",
    "    model, shape=(640, 640), dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "\n",
    "results[\"sunrgbd\"][\"MultiMAE\"] = cal_flops_params(\n",
    "    model, shape=(640, 640), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")\n",
    "del model, args, input_adapters, adapters_dict, output_adapters, DOMAIN_CONF\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using cache found in /root/.cache/torch/hub/facebookresearch_omnivore_main\n",
      "Using cache found in /root/.cache/torch/hub/facebookresearch_omnivore_main\n",
      "Using cache found in /root/.cache/torch/hub/facebookresearch_omnivore_main\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'cv2.gapi.oak', 'cv2.ipp', 'hydra.core.global_hydra', 'einops._backends', 'cv2.load_config_py3', 'attr._version_info', 'hydra._internal.config_loader_impl', 'hydra.grammar.gen', 'portalocker', 'antlr4.atn.ATNDeserializer', 'attr.setters', 'antlr4.IntervalSet', 'antlr4.atn.ATNConfigSet', 'cv2.cuda', 'antlr4.atn.ATN', 'cv2.qt', 'antlr4.atn.LexerATNSimulator', 'hydra.grammar', 'iopath.common', 'hydra.core.config_loader', 'hydra.utils', 'antlr4.tree.Chunk', 'cv2.gapi.ov', 'cv2.barcode', 'hydra._internal.grammar', 'omnivore.models', 'iopath.common.non_blocking_io', 'cv2.gapi.streaming', 'omnivision.utils', 'omegaconf.grammar', 'omnivision.models', 'cv2.gapi.wip.gst', 'hydra.core.config_store', 'cv2.Error', 'cv2.samples', 'antlr4.error.DiagnosticErrorListener', 'cv2.gapi.ie', 'antlr4.atn.Transition', 'attr._funcs', 'antlr4.tree.Trees', 'antlr4.atn.PredictionMode', 'hydra.main', 'iopath.common.download', 'antlr4.Parser', 'hydra.core', 'hydra.core.singleton', 'antlr4.tree.ParseTreePatternMatcher', 'portalocker.portalocker', 'portalocker.utils', 'einops.packing', 'attr._config', 'antlr4.InputStream', 'hydra.plugins.launcher', 'omegaconf.errors', 'antlr4.tree.TokenTagToken', 'hydra._internal.config_search_path_impl', 'hydra.core.default_element', 'antlr4', 'hydra.plugins.plugin', 'antlr4.dfa', 'antlr4.error.ErrorStrategy', 'iopath.common.file_io', 'cv2.segmentation', 'cv2.gapi.core.fluid', 'antlr4.error.Errors', 'omegaconf.grammar.gen', 'hydra.conf', 'cv2.ogl', 'omnivision.models.swin_transformer', 'cv2.gapi.ie.detail', 'hydra.compose', 'hydra.core.object_type', 'cv2.gapi.render.ocv', 'hydra.plugins.sweeper', 'hydra.core.override_parser.overrides_visitor', 'hydra.plugins.config_source', 'antlr4.Token', 'hydra.plugins.completion_plugin', 'iopath.version', 'antlr4.Lexer', 'antlr4.atn.ATNType', 'antlr4.RuleContext', 'cv2.gapi.wip', 'omegaconf.resolvers', 'hydra.initialize', 'hydra.core.config_search_path', 'antlr4.atn.LexerActionExecutor', 'omnivision.utils.checkpoint', 'antlr4.atn', 'hydra._internal', 'cv2.gapi.onnx', 'omnivore.models.omnivore_model', 'antlr4.tree.RuleTagToken', 'hydra.plugins', 'cv2.misc', 'antlr4.dfa.DFAState', 'cv2.utils.nested', 'omegaconf.resolvers.oc', 'hydra._internal.grammar.functions', 'cv2.data', 'antlr4.tree.Tree', 'cv2.flann', 'cv2.gapi.ot', 'hydra._internal.deprecation_warning', 'iopath', 'cv2', 'omegaconf.basecontainer', 'hydra.core.override_parser', 'attr.converters', 'hydra.version', 'antlr4.atn.ParserATNSimulator', 'hydra._internal.config_repository', 'hydra._internal.utils', 'omegaconf.grammar_parser', 'hydra._internal.sources_registry', 'omegaconf.grammar.gen.OmegaConfGrammarLexer', 'cv2.gapi.onnx.ep', 'cv2.typing', 'antlr4.PredictionContext', 'hydra._internal.grammar.utils', 'omegaconf.resolvers.oc.dict', 'einops.einops', 'hydra._internal.hydra', 'hydra.core.utils', 'hydra.types', 'iopath.common.event_logger', 'attr.validators', 'cv2.gapi.render', 'cv2.gapi.own', 'omegaconf.nodes', 'omegaconf.grammar.gen.OmegaConfGrammarParserVisitor', 'cv2.parallel', 'hydra.core.override_parser.overrides_parser', 'einops._torch_specific', 'antlr4.Utils', 'antlr4.BufferedTokenStream', 'portalocker.__about__', 'cv2.aruco', 'omegaconf.dictconfig', 'antlr4.atn.ATNDeserializationOptions', 'iopath.tabular', 'cv2.ocl', 'omnivision.utils.distributed', 'attr', 'hydra.core.override_parser.types', 'hydra.grammar.gen.OverrideParser', 'antlr4.atn.SemanticContext', 'cv2.gapi.core', 'cv2.utils.fs', 'cv2.gapi.wip.onevpl', 'cv2.mat_wrapper', 'antlr4.Recognizer', 'hydra._internal.grammar.grammar_functions', 'cv2.fisheye', 'omegaconf.version', 'timm.models.layers', 'einops.parsing', 'iopath.tabular.tabular_io', 'cv2.ml', 'antlr4.ParserRuleContext', 'cv2.gapi.wip.draw', 'cv2.gapi.own.detail', 'antlr4.dfa.DFA', 'cv2.detail', 'portalocker.exceptions', 'hydra._internal.callbacks', 'antlr4.atn.ATNConfig', 'cv2.videoio_registry', 'omegaconf', 'hydra.errors', 'omegaconf.grammar_visitor', 'einops', 'cv2.misc.version', 'antlr4.CommonTokenStream', 'cv2.gapi.imgproc', 'cv2.gapi.imgproc.fluid', 'attr._next_gen', 'antlr4.CommonTokenFactory', 'attr._compat', 'antlr4.error.ErrorListener', 'antlr4.FileStream', 'cv2.version', 'antlr4.ListTokenSource', 'antlr4.error', 'omegaconf.omegaconf', 'cv2.gapi.video', 'omegaconf.base', 'cv2.gapi.ot.cpu', 'antlr4.StdinStream', 'cv2.utils', 'hydra', 'hydra._internal.instantiate._instantiate2', 'portalocker.constants', 'omegaconf._utils', 'hydra.grammar.gen.OverrideParserVisitor', 'antlr4.atn.ATNSimulator', 'hydra.grammar.gen.OverrideLexer', 'attr.filters', 'hydra._internal.defaults_list', 'cv2.gapi', 'cv2.dnn', 'antlr4.atn.LexerAction', 'omegaconf.listconfig', 'omnivision', 'antlr4.atn.ATNState', 'attr._make', 'hydra._internal.instantiate', 'hydra.plugins.search_path_plugin', 'attr._cmp', 'cv2.gapi.core.cpu', 'omegaconf.grammar.gen.OmegaConfGrammarParser', 'antlr4.tree', 'cv2.gapi.core.ocl', 'omnivore', 'hydra.core.hydra_config', 'attr.exceptions', 'hydra.core.plugins'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/omnivore\")\n",
    "import torch\n",
    "\n",
    "# use cuda to accalerate the calculation\n",
    "model = decorate_model_for_omnivore(\n",
    "    torch.hub.load(\"facebookresearch/omnivore\", model=\"omnivore_swinB\").eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"omnivore_swinB\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=True\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_omnivore(\n",
    "    torch.hub.load(\"facebookresearch/omnivore\", model=\"omnivore_swinS\").eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"omnivore_swinS\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=True\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_omnivore(\n",
    "    torch.hub.load(\"facebookresearch/omnivore\", model=\"omnivore_swinT\").eval()\n",
    ")\n",
    "results[\"nyudv2\"][\"omnivore_swinT\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=True\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/torchvision/models/_utils.py:208: UserWarning: The parameter 'pretrained' is deprecated since 0.13 and may be removed in the future, please use 'weights' instead.\n",
      "  warnings.warn(\n",
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and may be removed in the future. The current behavior is equivalent to passing `weights=ResNet34_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet34_Weights.DEFAULT` to get the most up-to-date weights.\n",
      "  warnings.warn(msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'repositories.PGDENet.toolbox.models.BBSnetmodel', 'cv2.gapi.oak', 'cv2.misc.version', 'cv2.gapi.imgproc', 'cv2.gapi.imgproc.fluid', 'cv2.ipp', 'cv2.segmentation', 'cv2.load_config_py3', 'cv2.gapi.onnx.ep', 'cv2.version', 'toolbox.models.BBSnetmodel', 'cv2.gapi.core.fluid', 'cv2.typing', 'cv2.cuda', 'cv2.gapi.video', 'cv2.qt', 'cv2.ogl', 'repositories.PGDENet.toolbox.models.BBSnetmodel.BBSnet', 'cv2.gapi.ie.detail', 'toolbox.models.BBSnetmodel.fusion', 'cv2.gapi.render.ocv', 'cv2.gapi.ov', 'cv2.gapi.ot.cpu', 'repositories.PGDENet', 'toolbox.models.BBSnetmodel.ResNet', 'cv2.barcode', 'repositories', 'cv2.gapi.streaming', 'cv2.gapi.own', 'cv2.gapi.render', 'cv2.utils', 'cv2.parallel', 'toolbox.models.BBSnetmodel.DEM', 'repositories.PGDENet.toolbox.models', 'repositories.PGDENet.toolbox', 'cv2.aruco', 'cv2.gapi.wip.gst', 'cv2.gapi.wip', 'cv2.Error', 'cv2.samples', 'cv2.gapi.ie', 'cv2.ocl', 'cv2.gapi.core', 'cv2.gapi.onnx', 'cv2.gapi', 'cv2.dnn', 'cv2.misc', 'cv2.utils.fs', 'cv2.gapi.wip.onevpl', 'cv2.mat_wrapper', 'cv2.fisheye', 'cv2.gapi.own.detail', 'cv2.utils.nested', 'cv2.ml', 'cv2.gapi.core.cpu', 'cv2.gapi.wip.draw', 'cv2.detail', 'toolbox.models.BBSnetmodel.SG', 'cv2.data', 'cv2.gapi.core.ocl', 'cv2.flann', 'cv2.videoio_registry', 'cv2.gapi.ot', 'toolbox.models', 'cv2', 'toolbox'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/PGDENet\")\n",
    "from repositories.PGDENet.toolbox.models.BBSnetmodel.BBSnet import BBSNet\n",
    "\n",
    "model = BBSNet(n_class=40).eval()\n",
    "results[\"nyudv2\"][\"PGDENet\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=True, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = BBSNet(n_class=37).eval()\n",
    "results[\"sunrgbd\"][\"PGDENet\"] = cal_flops_params(\n",
    "    model, shape=(544, 736), dataset=\"sunrgbd\", single_channel=True, cuda=False\n",
    ")  # Since FRNet only accepts resolution of multiple of 32(2^5), we use 544x736\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'cv2.gapi.oak', 'cv2.misc.version', 'cv2.gapi.imgproc', 'cv2.gapi.imgproc.fluid', 'cv2.ipp', 'cv2.segmentation', 'cv2.load_config_py3', 'utils.pyt_utils', 'cv2.version', 'cv2.gapi.core.fluid', 'engine.logger', 'cv2.typing', 'cv2.gapi.onnx.ep', 'utils', 'cv2.cuda', 'cv2.gapi.video', 'cv2.qt', 'engine', 'cv2.ogl', 'net_util', 'cv2.gapi.ie.detail', 'cv2.gapi.render.ocv', 'cv2.gapi.ov', 'cv2.gapi.ot.cpu', 'cv2.barcode', 'cv2.gapi.streaming', 'cv2.gapi.own', 'cv2.gapi.render', 'network', 'cv2.parallel', 'cv2.utils', 'cv2.aruco', 'cv2.gapi.wip.gst', 'cv2.gapi.wip', 'cv2.Error', 'cv2.samples', 'cv2.gapi.ie', 'cv2.ocl', 'easydict', 'cv2.gapi.core', 'cv2.gapi', 'cv2.dnn', 'cv2.misc', 'cv2.utils.fs', 'cv2.gapi.onnx', 'cv2.gapi.wip.onevpl', 'config', 'cv2.mat_wrapper', 'cv2.fisheye', 'cv2.gapi.own.detail', 'cv2.utils.nested', 'cv2.ml', 'dual_resnet', 'cv2.gapi.core.cpu', 'cv2.gapi.wip.draw', 'cv2.detail', 'cv2.data', 'cv2.gapi.core.ocl', 'cv2.flann', 'cv2.videoio_registry', 'cv2.gapi.ot', 'cv2'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/SAGate/model/SA-Gate.nyu\")\n",
    "sys.path.append(\"repositories/SAGate/furnace\")\n",
    "from network import DeepLab\n",
    "import torch.nn as nn\n",
    "\n",
    "model = DeepLab(40, criterion=None, norm_layer=nn.BatchNorm2d)\n",
    "results[\"nyudv2\"][\"SA-Gate\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = DeepLab(37, criterion=None, norm_layer=nn.BatchNorm2d)\n",
    "results[\"sunrgbd\"][\"SA-Gate\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# SGNet only supports old machine, therefore we skip it\n",
    "\n",
    "# sys.path.append(\"repositories/SGNet\")\n",
    "# from repositories.SGNet.graphs.models.SGNet.SGNet_ASPP import SGNet\n",
    "\n",
    "# model = SGNet(n_class=40).eval()\n",
    "\n",
    "# envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\u001b[32m07 05:18:11 \u001b[0mResNet init weights from pretreain\n",
      "\u001b[32m07 05:18:12 \u001b[0mASPP init weights\n",
      "\u001b[32m07 05:18:12 \u001b[0mGFPN init weights\n",
      "\u001b[32m07 05:18:12 \u001b[0mHead init weights\n",
      "\u001b[32m07 05:18:16 \u001b[0mResNet init weights from pretreain\n",
      "\u001b[32m07 05:18:16 \u001b[0mASPP init weights\n",
      "\u001b[32m07 05:18:16 \u001b[0mGFPN init weights\n",
      "\u001b[32m07 05:18:16 \u001b[0mHead init weights\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'cv2.gapi.oak', 'repositories.ShapeConv.rgbd_seg.utils.checkpoint', 'cv2.ipp', 'cv2.load_config_py3', 'repositories.ShapeConv.rgbd_seg.models.utils.norm', 'opt_einsum.parser', 'cv2.cuda', 'cv2.qt', 'rgbd_seg.models.utils.builder', 'opt_einsum.backends.jax', 'cv2.gapi.ov', 'rgbd_seg.models.decoders.builder', 'rgbd_seg.models.encoders.enhance_modules.aspp', 'opt_einsum.backends.theano', 'cv2.barcode', 'repositories', 'cv2.gapi.streaming', 'rgbd_seg.models.builder', 'cv2.gapi.wip.gst', 'cv2.Error', 'repositories.ShapeConv.rgbd_seg.models.encoders.enhance_modules.ppm', 'cv2.samples', 'opt_einsum.path_random', 'rgbd_seg.models.encoders.enhance_modules.ppm', 'cv2.gapi.ie', 'opt_einsum', 'rgbd_seg.models.utils.act', 'rgbd_seg.models.utils.shape_conv', 'repositories.ShapeConv.rgbd_seg.models.encoders.backbones.builder', 'opt_einsum._version', 'repositories.ShapeConv.rgbd_seg.models.utils.registry', 'rgbd_seg.models', 'rgbd_seg.models.registry', 'rgbd_seg.models.heads.registry', 'opt_einsum.backends.cupy', 'repositories.ShapeConv', 'opt_einsum.contract', 'cv2.segmentation', 'cv2.gapi.core.fluid', 'rgbd_seg.models.encoders.builder', 'cv2.ogl', 'rgbd_seg.models.decoders.registry', 'cv2.gapi.ie.detail', 'cv2.gapi.render.ocv', 'repositories.ShapeConv.rgbd_seg.models.encoders', 'rgbd_seg.models.heads.builder', 'cv2.gapi.wip', 'opt_einsum.backends', 'opt_einsum.backends.torch', 'rgbd_seg.models.heads.head', 'rgbd_seg.models.encoders.backbones.builder', 'cv2.gapi.onnx', 'opt_einsum.helpers', 'cv2.misc', 'rgbd_seg.models.encoders.backbones.registry', 'rgbd_seg.models.heads', 'addict', 'repositories.ShapeConv.rgbd_seg.models.utils.conv_module', 'cv2.utils.nested', 'repositories.ShapeConv.rgbd_seg.models.builder', 'cv2.data', 'cv2.flann', 'cv2.gapi.ot', 'rgbd_seg.models.encoders.backbones', 'cv2', 'opt_einsum.blas', 'torch.backends.opt_einsum', 'rgbd_seg.models.decoders', 'cv2.gapi.onnx.ep', 'repositories.ShapeConv.rgbd_seg.models', 'cv2.typing', 'cv2.gapi.render', 'cv2.gapi.own', 'repositories.ShapeConv.rgbd_seg.models.encoders.builder', 'repositories.ShapeConv.rgbd_seg.utils.registry', 'cv2.parallel', 'repositories.ShapeConv.rgbd_seg.models.encoders.enhance_modules.registry', 'rgbd_seg.models.encoders.enhance_modules.registry', 'opt_einsum.sharing', 'cv2.aruco', 'rgbd_seg.models.decoders.gfpn.gfpn', 'rgbd_seg.models.utils.conv_module', 'opt_einsum.backends.object_arrays', 'opt_einsum.backends.dispatch', 'cv2.ocl', 'rgbd_seg.models.utils.registry', 'cv2.gapi.core', 'rgbd_seg.models.decoders.bricks', 'repositories.ShapeConv.rgbd_seg.models.encoders.enhance_modules.builder', 'cv2.utils.fs', 'cv2.gapi.wip.onevpl', 'repositories.ShapeConv.rgbd_seg.models.encoders.backbones', 'cv2.mat_wrapper', 'rgbd_seg.utils.dist_utils', 'repositories.ShapeConv.rgbd_seg.models.encoders.backbones.resnet', 'cv2.fisheye', 'cv2.gapi.own.detail', 'cv2.gapi.wip.draw', 'cv2.ml', 'nyu40_deeplabv3plus_resnext101_shape', 'rgbd_seg', 'cv2.detail', 'rgbd_seg.utils.registry', 'rgbd_seg.models.utils.norm', 'cv2.videoio_registry', 'rgbd_seg.models.encoders.backbones.resnet', 'addict.addict', 'repositories.ShapeConv.rgbd_seg.models.weight_init', 'cv2.misc.version', 'repositories.ShapeConv.rgbd_seg.models.registry', 'cv2.gapi.imgproc', 'rgbd_seg.models.encoders.enhance_modules', 'cv2.gapi.imgproc.fluid', 'repositories.ShapeConv.rgbd_seg.models.encoders.enhance_modules.aspp', 'rgbd_seg.models.utils.upsample', 'cv2.version', 'rgbd_seg.utils.config', 'cv2.gapi.video', 'repositories.ShapeConv.rgbd_seg.models.utils.act', 'repositories.ShapeConv.rgbd_seg.utils.config', 'rgbd_seg.models.utils', 'cv2.gapi.ot.cpu', 'cv2.utils', 'repositories.ShapeConv.rgbd_seg.utils.dist_utils', 'repositories.ShapeConv.rgbd_seg.models.utils', 'repositories.ShapeConv.rgbd_seg.models.encoders.enhance_modules', 'opt_einsum.paths', 'repositories.ShapeConv.rgbd_seg.utils', 'rgbd_seg.models.weight_init', 'repositories.ShapeConv.rgbd_seg.models.utils.builder', 'cv2.gapi', 'cv2.dnn', 'rgbd_seg.models.encoders', 'rgbd_seg.models.decoders.gfpn', 'repositories.ShapeConv.rgbd_seg.models.utils.upsample', 'cv2.gapi.core.cpu', 'repositories.ShapeConv.rgbd_seg.models.encoders.backbones.registry', 'cv2.gapi.core.ocl', 'sun_deeplabv3plus_resnext101_shape', 'rgbd_seg.utils', 'repositories.ShapeConv.rgbd_seg', 'opt_einsum.backends.tensorflow', 'rgbd_seg.utils.checkpoint', 'rgbd_seg.models.encoders.enhance_modules.builder'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/ShapeConv\")\n",
    "from repositories.ShapeConv.rgbd_seg.models.builder import build_model\n",
    "# to register the model, we need to import the backbone\n",
    "import repositories.ShapeConv.rgbd_seg.models.encoders.backbones\n",
    "from repositories.ShapeConv.rgbd_seg.utils import Config\n",
    "\n",
    "cfg = Config.fromfile(\n",
    "    \"repositories/ShapeConv/configs/nyu/nyu40_deeplabv3plus_resnext101_shape.py\"\n",
    ")\n",
    "model = decorate_model_for_stack_input(build_model(cfg[\"inference\"][\"model\"]).eval())\n",
    "results[\"nyudv2\"][\"ShapeConv\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model, cfg\n",
    "\n",
    "# importlib.reload(repositories.ShapeConv.rgbd_seg.models.encoders.backbones)\n",
    "cfg = Config.fromfile(\n",
    "    \"repositories/ShapeConv/configs/sun/sun_deeplabv3plus_resnext101_shape.py\"\n",
    ")\n",
    "model = decorate_model_for_stack_input(build_model(cfg[\"inference\"][\"model\"]).eval())\n",
    "results[\"sunrgbd\"][\"ShapeConv\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model, cfg\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/envs/dformer_new/lib/python3.10/site-packages/calflops/pytorch_ops.py:364: UserWarning: Implicit dimension choice for softmax has been deprecated. Change the call to include dim=X as an argument.\n",
      "  return oldFunc(*args, **kwds)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Deleted module: {'mmcv.cnn.bricks.depthwise_separable_conv_module', 'yapf_third_party._ylib2to3.pgen2.pgen', 'mmengine.optim.optimizer.base', 'mmcv.visualization', 'mmengine.optim.scheduler', 'cv2.cuda', 'cv2.qt', 'mmengine._strategy.deepspeed', 'mmengine.optim.scheduler.momentum_scheduler', 'mmengine.runner', 'repositories', 'mmcv', 'mmengine.registry.build_functions', 'cv2.gapi.wip.gst', 'mmengine.fileio.backends', 'cv2.samples', 'importlib_metadata.compat', 'rich.table', 'mmcv.cnn.bricks.norm', 'rich.box', 'repositories.TokenFusion.semantic_segmentation.models', 'mmengine.optim.optimizer.builder', 'yapf_third_party._ylib2to3.patcomp', 'mmcv.cnn', 'mmcv.visualization.image', 'attr._config', 'mmengine._strategy.fsdp', 'rich', 'mmengine.utils.misc', 'mmcv.cnn.bricks.drop', 'yapf_third_party._ylib2to3.pgen2.tokenize', 'torch.distributed.optim.functional_adamax', 'cv2.gapi.core.fluid', 'mmengine._strategy', 'mmengine.hooks.runtime_info_hook', 'mmengine.utils.path', 'zipp', 'rich._export_format', 'rich._palettes', 'mmcv.transforms.base', 'cv2.gapi.wip', 'torch.distributed.optim.zero_redundancy_optimizer', 'cv2.misc', 'repositories.TokenFusion.semantic_segmentation.models.segformer', 'rich.live', 'rich.protocol', 'mmengine.optim.optimizer.amp_optimizer_wrapper', 'mmcv.transforms.builder', 'attr.converters', 'mmengine.fileio.io', 'yapf_third_party._ylib2to3.fixer_util', 'mmengine.fileio.backends.registry_utils', 'mmengine.registry.default_scope', 'yapf.pyparser.pyparser_utils', 'mmcv.cnn.rfsearch', 'cv2.parallel', 'mmengine.utils.dl_utils.setup_env', 'mmengine.visualization.vis_backend', 'mmengine.model.wrappers.distributed', 'yapf_third_party._ylib2to3.pgen2.parse', 'mmengine.model.wrappers.seperate_distributed', 'cv2.utils.fs', 'mmengine.registry', 'cv2.mat_wrapper', 'mmengine.runner.base_loop', 'torch.distributed.optim.functional_rprop', 'yapf.pytree.comment_splicer', 'mmengine.utils.dl_utils.parrots_wrapper', 'mmengine.dist', 'yapf.pytree.pytree_utils', 'addict.addict', 'rich.file_proxy', 'termcolor', 'mmengine.structures.base_data_element', 'cv2.gapi.imgproc', 'rich._loop', 'mmengine.runner.loops', 'yapf.pytree', 'rich.filesize', 'rich._fileno', 'yapf.yapflib.format_token', 'rich.pretty', 'importlib_metadata._meta', 'yapf.yapflib.identify_container', 'rich.containers', 'mmengine.device', 'yapf.yapflib.file_resources', 'cv2.gapi.core.cpu', 'mmengine.model.weight_init', 'mmengine.device.utils', 'mmengine.optim.optimizer.zero_optimizer', 'cv2.gapi.core.ocl', 'mmengine.runner.priority', 'mmcv.image.photometric', 'mmengine.fileio', 'mmengine.hooks.naive_visualization_hook', 'cv2.gapi.oak', 'yapf.yapflib.line_joiner', 'rich.errors', 'rich.markup', 'mmengine.structures.instance_data', 'importlib_metadata.compat.py39', 'yapf.yapflib', 'mmengine.utils.version_utils', 'rich._wrap', 'yapf.pytree.blank_line_calculator', 'mmengine.evaluator', 'cv2.gapi.streaming', 'mmcv.transforms.formatting', 'mmengine.hooks.early_stopping_hook', 'mmcv.cnn.bricks.generalized_attention', 'mmcv.arraymisc', 'mmengine.dataset.utils', 'rich.cells', 'yapf_third_party._ylib2to3', 'mmengine.model.wrappers.utils', 'mmengine.fileio.handlers.pickle_handler', 'mmcv.video', 'mmengine.visualization', 'mmcv.cnn.bricks.upsample', 'mmengine.structures', 'torch.distributed.optim.functional_adadelta', 'torch.distributed.optim.named_optimizer', 'rich.region', 'torch.distributed.algorithms.model_averaging', 'mmengine.model.averaged_model', 'rich.measure', 'cv2.segmentation', 'rich._emoji_replace', 'mmcv.cnn.resnet', 'yapf.pytree.subtype_assigner', 'rich.highlighter', 'yapf.pyparser.pyparser', 'mmcv.transforms.loading', 'yapf_third_party._ylib2to3.pgen2.driver', 'mmcv.cnn.bricks.conv_ws', 'importlib_metadata._functools', 'mmengine.hooks.empty_cache_hook', 'tifffile._imagecodecs', 'mmengine.hooks.sampler_seed_hook', 'mmengine.runner.checkpoint', 'addict', 'cv2.flann', 'cv2.gapi.ot', 'mmengine.utils.dl_utils.time_counter', 'rich.default_styles', 'mmcv.cnn.utils.flops_counter', 'rich.themes', 'torch.distributed.algorithms.model_averaging.utils', 'importlib_metadata._text', 'mmcv.cnn.bricks.swish', 'mmengine.fileio.handlers.yaml_handler', 'tifffile.tifffile', 'mmcv.transforms.wrappers', 'mmengine.registry.registry', 'attr.validators', 'cv2.gapi.own', 'mmengine.evaluator.metric', 'mmengine.model.base_model.data_preprocessor', 'mmengine.registry.root', 'attr', 'mmengine.model.efficient_conv_bn_eval', 'mmcv.visualization.optflow', 'cv2.gapi.core', 'mmengine.dataset.sampler', 'cv2.fisheye', 'mmengine.optim', 'rich.spinner', 'mmengine.fileio.backends.local_backend', 'rich.pager', 'mmengine.fileio.handlers.base', 'mmengine.logging.logger', 'cv2.videoio_registry', 'mmengine.dist.utils', 'mmengine', 'mmengine.hooks.profiler_hook', 'mmengine.config.lazy', 'cv2.version', 'mmengine.fileio.backends.http_backend', 'mmengine.model.utils', 'mmcv.image.misc', 'yapf_third_party._ylib2to3.pytree', 'mmcv.version', 'rich.live_render', 'cv2.utils', 'mmengine.utils.timer', 'mmengine.model.base_module', 'mmcv.cnn.bricks.conv2d_adaptive_padding', 'yapf.pyparser', 'cv2.gapi', 'mmengine.fileio.handlers', 'attr._cmp', 'rich.repr', 'mmengine._strategy.base', 'mmengine.logging.message_hub', 'rich.segment', 'cv2.ipp', 'cv2.load_config_py3', 'mmengine.model', 'mmengine.utils.dl_utils.collect_env', 'attr.setters', 'mmengine.model.test_time_aug', 'rich.scope', 'rich._ratio', 'rich.console', 'mmengine.hooks.checkpoint_hook', 'repositories.TokenFusion.semantic_segmentation', 'mmcv.cnn.bricks.scale', 'mmcv.cnn.rfsearch.utils', 'torch.distributed.optim.functional_sgd', 'yapf.pytree.split_penalty', 'cv2.barcode', 'mmcv.cnn.bricks.non_local', 'mmcv.cnn.bricks.padding', 'mmengine.hooks.ema_hook', 'rich.terminal_theme', 'mmengine.fileio.file_client', 'cv2.Error', 'mmcv.image.io', 'cv2.gapi.ie', 'mmcv.video.optflow', 'attr._funcs', 'rich.color_triplet', 'mmengine.utils.package_utils', 'rich.screen', 'mmengine.model.wrappers', 'rich._extension', 'importlib_metadata._compat', 'mmcv.video.processing', 'mmengine.fileio.handlers.registry_utils', 'mmengine.version', 'mmengine.optim.scheduler.lr_scheduler', 'mmengine.optim.optimizer.optimizer_wrapper_dict', 'cv2.gapi.ie.detail', 'zipp.glob', 'rich.ansi', 'yapf_third_party._ylib2to3.pygram', 'rich.constrain', 'mmengine.runner._flexible_runner', 'mmcv.image', 'mmengine.evaluator.utils', 'mmengine.dataset', 'yapf.pytree.continuation_splicer', 'cv2.gapi.onnx', 'mmengine.hooks.logger_hook', 'rich.emoji', 'yapf', 'mmengine._strategy.utils', 'cv2.data', 'mmengine.config', 'mmengine.utils.dl_utils', 'mmcv.cnn.bricks.hswish', 'repositories.TokenFusion.semantic_segmentation.models.mix_transformer', 'mmengine.fileio.backends.memcached_backend', 'cv2.gapi.onnx.ep', 'cv2.typing', 'yapf.yapflib.format_decision_state', 'mmengine.utils.dl_utils.torch_ops', 'mmengine.runner.log_processor', 'yapf.yapflib.subtypes', 'yapf.yapflib.yapf_api', 'mmengine._strategy.colossalai', 'rich.palette', 'mmengine.utils.dl_utils.hub', 'mmengine.utils', 'mmengine.model.wrappers.fully_sharded_distributed', 'mmcv.cnn.bricks.hsigmoid', 'cv2.ocl', 'mmengine.utils.dl_utils.misc', 'torch.distributed.optim.post_localSGD_optimizer', 'torch.distributed.optim.functional_adagrad', 'mmengine.fileio.backends.base', 'mmcv.image.geometric', 'timm.models.layers', 'mmcv.cnn.vgg', 'cv2.detail', 'mmengine.visualization.visualizer', 'mmcv.cnn.bricks.conv', 'termcolor.termcolor', 'mmengine.structures.label_data', 'rich._emoji_codes', 'cv2.misc.version', 'mmengine.utils.dl_utils.trace', 'rich.style', 'mmengine.optim.optimizer.apex_optimizer_wrapper', 'rich.padding', 'mmengine.fileio.handlers.json_handler', 'rich.jupyter', 'mmengine.optim.optimizer', 'torch.distributed.optim.functional_adam', 'mmengine.dist.dist', 'torch.distributed.optim', 'rich.progress', 'mmcv.cnn.rfsearch.operator', 'rich.panel', 'yapf.yapflib.errors', 'yapf_third_party._ylib2to3.pgen2.literals', 'mmengine.logging', 'cv2.dnn', 'yapf.pyparser.split_penalty_visitor', 'repositories.TokenFusion', 'mmengine.runner.runner', 'attr._make', 'mmcv.transforms.utils', 'yapf.pytree.pytree_visitor', 'mmengine.utils.progressbar', 'mmcv.cnn.bricks.plugin', 'mmcv.transforms', 'attr.exceptions', 'mmengine.dataset.base_dataset', 'attr._version_info', 'mmengine.logging.history_buffer', 'yapf.yapflib.reformatter', 'torch.distributed.optim.apply_optimizer_in_backward', 'cv2.gapi.ov', 'torch.distributed.algorithms.model_averaging.averagers', 'rich.styled', 'mmengine._strategy.distributed', 'mmcv.video.io', 'zipp.compat.py310', 'mmengine.model.base_model.base_model', 'yapf_third_party._ylib2to3.pgen2.token', 'mmengine.registry.utils', 'torch.distributed.optim.utils', 'importlib_metadata._collections', 'mmengine.config.utils', 'mmengine.utils.manager', 'mmengine.fileio.parse', 'mmcv.image.colorspace', 'mmengine.structures.pixel_data', 'cv2.ogl', 'yapf.yapflib.style', 'mmengine.utils.progressbar_rich', 'termcolor._types', 'rich._spinners', 'cv2.gapi.render.ocv', 'mmcv.cnn.bricks', 'mmcv.cnn.utils', 'yapf_third_party._ylib2to3.pgen2.grammar', 'rich._null_file', 'mmengine.dataset.dataset_wrapper', 'rich.progress_bar', 'mmengine.optim.scheduler.param_scheduler', 'mmcv.visualization.color', 'mmengine.runner.activation_checkpointing', 'cv2.utils.nested', 'mmengine.hooks.hook', 'mmengine.hooks.test_time_aug_hook', 'mmcv.cnn.bricks.wrappers', 'cv2', 'mmengine.hooks', 'yapf_third_party', 'mmengine.config.config', 'mmcv.cnn.rfsearch.search', 'torch.distributed.optim.functional_rmsprop', 'yapf.pytree.pytree_unwrapper', 'mmengine.runner.amp', 'repositories.TokenFusion.semantic_segmentation.models.modules', 'mmengine.optim.optimizer.optimizer_wrapper', 'mmcv.cnn.bricks.conv_module', 'mmcv.cnn.bricks.context_block', 'mmengine.fileio.backends.lmdb_backend', 'yapf.yapflib.split_penalty', 'cv2.gapi.render', 'rich.text', 'cv2.aruco', 'rich._log_render', 'mmengine.model.base_model', 'zipp.compat', 'mmengine.visualization.utils', 'cv2.gapi.wip.onevpl', 'mmengine.hooks.iter_timer_hook', 'importlib_metadata._adapters', 'cv2.gapi.own.detail', 'cv2.gapi.wip.draw', 'cv2.ml', 'mmengine.hooks.param_scheduler_hook', 'importlib_metadata._itertools', 'mmcv.cnn.utils.fuse_conv_bn', 'rich.abc', 'torch.distributed.optim.optimizer', 'mmcv.transforms.processing', 'cv2.gapi.imgproc.fluid', 'attr._next_gen', 'mmengine.runner.utils', 'attr._compat', 'mmengine.hooks.sync_buffer_hook', 'importlib_metadata', 'cv2.gapi.video', 'rich.theme', 'yapf.yapflib.logical_line', 'yapf.yapflib.object_state', 'mmcv.cnn.bricks.activation', 'mmengine.optim.optimizer.default_constructor', 'cv2.gapi.ot.cpu', 'torch.distributed.optim.functional_adamw', 'mmcv.arraymisc.quantization', 'yapf_third_party._ylib2to3.pgen2', 'attr.filters', 'rich.control', 'mmcv.cnn.alexnet', 'rich.align', 'mmengine.evaluator.evaluator', 'mmengine.fileio.backends.petrel_backend', 'tifffile', 'rich.color', 'mmengine._strategy.single_device', 'rich._pick', 'rich._cell_widths'}\n",
      "Envs_state_manager restored, new modules deleted! \n"
     ]
    }
   ],
   "source": [
    "sys.path.append(\"repositories/TokenFusion\")\n",
    "from repositories.TokenFusion.semantic_segmentation.models.segformer import WeTr\n",
    "\n",
    "model = decorate_model_for_zip_input(WeTr(\"mit_b2\", 40, pretrained=False).eval())\n",
    "results[\"nyudv2\"][\"TokenFusion-B2\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(WeTr(\"mit_b3\", 40, pretrained=False).eval())\n",
    "results[\"nyudv2\"][\"TokenFusion-B3\"] = cal_flops_params(\n",
    "    model, dataset=\"nyudv2\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(WeTr(\"mit_b2\", 37, pretrained=False).eval())\n",
    "results[\"sunrgbd\"][\"TokenFusion-B2\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "model = decorate_model_for_zip_input(WeTr(\"mit_b3\", 37, pretrained=False).eval())\n",
    "results[\"sunrgbd\"][\"TokenFusion-B3\"] = cal_flops_params(\n",
    "    model, dataset=\"sunrgbd\", single_channel=False, cuda=False\n",
    ")\n",
    "del model\n",
    "\n",
    "envs_helper.restore()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Show the results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'nyudv2': {'ACNet': ('232.021 GFLOPS', '115.597 GMACs', '116.628 M'),\n",
      "            'AsymFormer': ('78.9987 GFLOPS', '39.3641 GMACs', '33.0596 M'),\n",
      "            'CEN_101': ('1.2366 TFLOPS', '617.851 GMACs', '118.202 M'),\n",
      "            'CEN_152': ('1.3277 TFLOPS', '663.306 GMACs', '133.891 M'),\n",
      "            'CMNeXt-B4': ('263.791 GFLOPS', '131.374 GMACs', '116.576 M'),\n",
      "            'CMX-B2': ('134.284 GFLOPS', '66.9018 GMACs', '66.5809 M'),\n",
      "            'CMX-B4': ('268.904 GFLOPS', '134.01 GMACs', '139.874 M'),\n",
      "            'CMX-B5': ('336.078 GFLOPS', '167.522 GMACs', '181.074 M'),\n",
      "            'EMSANet': ('90.551 GFLOPS', '45.125 GMACs', '46.9354 M'),\n",
      "            'ESANet': ('96.6675 GFLOPS', '45.125 GMACs', '46.9558 M'),\n",
      "            'FRNet': ('219.392 GFLOPS', '109.528 GMACs', '87.8068 M'),\n",
      "            'GeminiFusion-B3': ('277.074 GFLOPS', '138.198 GMACs', '74.5513 M'),\n",
      "            'GeminiFusion-B5': ('513.239 GFLOPS', '256.063 GMACs', '137.209 M'),\n",
      "            'MultiMAE': ('808.41 GFLOPS', '403.8 GMACs', '95.3895 M'),\n",
      "            'PGDENet': ('326.558 GFLOPS', '162.911 GMACs', '107.403 M'),\n",
      "            'SA-Gate': ('386.007 GFLOPS', '192.58 GMACs', '110.875 M'),\n",
      "            'ShapeConv': ('337.861 GFLOPS', '168.366 GMACs', '106.792 M'),\n",
      "            'TokenFusion-B2': ('110.422 GFLOPS', '54.9436 GMACs', '26.0184 M'),\n",
      "            'TokenFusion-B3': ('188.751 GFLOPS', '93.9872 GMACs', '45.9173 M'),\n",
      "            'omnivore_swinB': ('199.31 GFLOPS', '99.4782 GMACs', '90.1146 M'),\n",
      "            'omnivore_swinS': ('112.296 GFLOPS', '56.0155 GMACs', '50.6036 M'),\n",
      "            'omnivore_swinT': ('57.4253 GFLOPS', '28.624 GMACs', '28.945 M')},\n",
      " 'sunrgbd': {'ACNet': ('302.24 GFLOPS', '150.581 GMACs', '116.626 M'),\n",
      "             'AsymFormer': ('105.62 GFLOPS', '52.6242 GMACs', '33.0588 M'),\n",
      "             'CEN_101': ('1.5797 TFLOPS', '789.279 GMACs', '118.195 M'),\n",
      "             'CEN_152': ('1.698 TFLOPS', '848.318 GMACs', '133.884 M'),\n",
      "             'CMNeXt-B4': ('340.623 GFLOPS', '169.639 GMACs', '116.575 M'),\n",
      "             'CMX-B2': ('172.347 GFLOPS', '85.8661 GMACs', '66.5794 M'),\n",
      "             'CMX-B4': ('347.444 GFLOPS', '173.152 GMACs', '139.872 M'),\n",
      "             'CMX-B5': ('435.174 GFLOPS', '216.919 GMACs', '181.073 M'),\n",
      "             'EMSANet': ('117.814 GFLOPS', '58.7119 GMACs', '46.9291 M'),\n",
      "             'ESANet': ('96.5128 GFLOPS', '45.0483 GMACs', '46.9496 M'),\n",
      "             'FRNet': ('285.739 GFLOPS', '142.652 GMACs', '87.8053 M'),\n",
      "             'GeminiFusion-B3': ('358.82 GFLOPS', '178.974 GMACs', '74.5505 M'),\n",
      "             'GeminiFusion-B5': ('666.333 GFLOPS',\n",
      "                                 '332.448 GMACs',\n",
      "                                 '137.209 M'),\n",
      "             'MultiMAE': ('808.41 GFLOPS', '403.8 GMACs', '95.3895 M'),\n",
      "             'PGDENet': ('425.406 GFLOPS', '212.223 GMACs', '107.401 M'),\n",
      "             'SA-Gate': ('498.509 GFLOPS', '248.711 GMACs', '110.873 M'),\n",
      "             'ShapeConv': ('436.967 GFLOPS', '217.824 GMACs', '106.791 M'),\n",
      "             'TokenFusion-B2': ('142.041 GFLOPS', '70.678 GMACs', '26.0177 M'),\n",
      "             'TokenFusion-B3': ('244.13 GFLOPS', '121.565 GMACs', '45.9165 M')}}\n"
     ]
    }
   ],
   "source": [
    "from pprint import pprint\n",
    "\n",
    "pprint(results)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "| Model           | Parameters   | FLOPs (NYUDepth v2)   | FLOPs (SUNRGBD)   |\n",
      "|:----------------|:-------------|:----------------------|:------------------|\n",
      "| ACNet           | 116.628 M    | 232.021 GFLOPS        | 302.24 GFLOPS     |\n",
      "| AsymFormer      | 33.0596 M    | 78.9987 GFLOPS        | 105.62 GFLOPS     |\n",
      "| CEN_101         | 118.202 M    | 1.2366 TFLOPS         | 1.5797 TFLOPS     |\n",
      "| CEN_152         | 133.891 M    | 1.3277 TFLOPS         | 1.698 TFLOPS      |\n",
      "| CMNeXt-B4       | 116.576 M    | 263.791 GFLOPS        | 340.623 GFLOPS    |\n",
      "| CMX-B2          | 66.5809 M    | 134.284 GFLOPS        | 172.347 GFLOPS    |\n",
      "| CMX-B4          | 139.874 M    | 268.904 GFLOPS        | 347.444 GFLOPS    |\n",
      "| CMX-B5          | 181.074 M    | 336.078 GFLOPS        | 435.174 GFLOPS    |\n",
      "| EMSANet         | 46.9354 M    | 90.551 GFLOPS         | 117.814 GFLOPS    |\n",
      "| ESANet          | 46.9558 M    | 96.6675 GFLOPS        | 96.5128 GFLOPS    |\n",
      "| FRNet           | 87.8068 M    | 219.392 GFLOPS        | 285.739 GFLOPS    |\n",
      "| GeminiFusion-B3 | 74.5513 M    | 277.074 GFLOPS        | 358.82 GFLOPS     |\n",
      "| GeminiFusion-B5 | 137.209 M    | 513.239 GFLOPS        | 666.333 GFLOPS    |\n",
      "| MultiMAE        | 95.3895 M    | 808.41 GFLOPS         | 808.41 GFLOPS     |\n",
      "| omnivore_swinB  | 90.1146 M    | 199.31 GFLOPS         |                   |\n",
      "| omnivore_swinS  | 50.6036 M    | 112.296 GFLOPS        |                   |\n",
      "| omnivore_swinT  | 28.945 M     | 57.4253 GFLOPS        |                   |\n",
      "| PGDENet         | 107.403 M    | 326.558 GFLOPS        | 425.406 GFLOPS    |\n",
      "| SA-Gate         | 110.875 M    | 386.007 GFLOPS        | 498.509 GFLOPS    |\n",
      "| ShapeConv       | 106.792 M    | 337.861 GFLOPS        | 436.967 GFLOPS    |\n",
      "| TokenFusion-B2  | 26.0184 M    | 110.422 GFLOPS        | 142.041 GFLOPS    |\n",
      "| TokenFusion-B3  | 45.9173 M    | 188.751 GFLOPS        | 244.13 GFLOPS     |\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "table = {\n",
    "    \"Model\": list(results[\"nyudv2\"].keys()),\n",
    "    \"Parameters\": [results[\"nyudv2\"][key][2] for key in results[\"nyudv2\"].keys()],\n",
    "    \"FLOPs (NYUDepth v2)\": [\n",
    "        results[\"nyudv2\"][key][0] for key in results[\"nyudv2\"].keys()\n",
    "    ],\n",
    "    \"FLOPs (SUNRGBD)\": [\n",
    "        (results[\"sunrgbd\"][key][0] if key in results[\"sunrgbd\"].keys() else None)\n",
    "        for key in results[\"nyudv2\"].keys()\n",
    "    ],\n",
    "}\n",
    "df = pd.DataFrame(table)\n",
    "# This is where the table in readme comes from\n",
    "print(df.to_markdown(index=False))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "dformer_new",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}