HIPT_4K/HIPT_4K Inference + Attention Visualization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of Patches: 196\n",
      "# of Patches: 196\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "HIPT_4K(\n",
       "  (model256): VisionTransformer(\n",
       "    (patch_embed): PatchEmbed(\n",
       "      (proj): Conv2d(3, 384, kernel_size=(16, 16), stride=(16, 16))\n",
       "    )\n",
       "    (pos_drop): Dropout(p=0.0, inplace=False)\n",
       "    (blocks): ModuleList(\n",
       "      (0): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (1): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (2): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (3): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (4): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (5): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (6): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (7): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (8): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (9): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (10): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (11): Block(\n",
       "        (norm1): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=384, out_features=1152, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=384, out_features=384, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=384, out_features=1536, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=1536, out_features=384, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (norm): LayerNorm((384,), eps=1e-06, elementwise_affine=True)\n",
       "    (head): Identity()\n",
       "  )\n",
       "  (model4k): VisionTransformer4K(\n",
       "    (phi): Sequential(\n",
       "      (0): Linear(in_features=384, out_features=192, bias=True)\n",
       "      (1): GELU(approximate='none')\n",
       "      (2): Dropout(p=0.0, inplace=False)\n",
       "    )\n",
       "    (pos_drop): Dropout(p=0.0, inplace=False)\n",
       "    (blocks): ModuleList(\n",
       "      (0): Block(\n",
       "        (norm1): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=192, out_features=576, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=192, out_features=192, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=192, out_features=768, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=768, out_features=192, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (1): Block(\n",
       "        (norm1): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=192, out_features=576, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=192, out_features=192, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=192, out_features=768, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=768, out_features=192, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (2): Block(\n",
       "        (norm1): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=192, out_features=576, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=192, out_features=192, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=192, out_features=768, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=768, out_features=192, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (3): Block(\n",
       "        (norm1): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=192, out_features=576, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=192, out_features=192, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=192, out_features=768, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=768, out_features=192, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (4): Block(\n",
       "        (norm1): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=192, out_features=576, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=192, out_features=192, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=192, out_features=768, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=768, out_features=192, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "      (5): Block(\n",
       "        (norm1): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (attn): Attention(\n",
       "          (qkv): Linear(in_features=192, out_features=576, bias=True)\n",
       "          (attn_drop): Dropout(p=0.0, inplace=False)\n",
       "          (proj): Linear(in_features=192, out_features=192, bias=True)\n",
       "          (proj_drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "        (drop_path): Identity()\n",
       "        (norm2): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "        (mlp): Mlp(\n",
       "          (fc1): Linear(in_features=192, out_features=768, bias=True)\n",
       "          (act): GELU(approximate='none')\n",
       "          (fc2): Linear(in_features=768, out_features=192, bias=True)\n",
       "          (drop): Dropout(p=0.0, inplace=False)\n",
       "        )\n",
       "      )\n",
       "    )\n",
       "    (norm): LayerNorm((192,), eps=1e-06, elementwise_affine=True)\n",
       "    (head): Identity()\n",
       "  )\n",
       ")"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from hipt_4k import HIPT_4K\n",
    "from hipt_model_utils import get_vit256, get_vit4k, eval_transforms\n",
    "from hipt_heatmap_utils import *\n",
    "light_jet = cmap_map(lambda x: x/2 + 0.5, matplotlib.cm.jet)\n",
    "\n",
    "pretrained_weights256 = '../Checkpoints/vit256_small_dino.pth'\n",
    "pretrained_weights4k = '../Checkpoints/vit4k_xs_dino.pth'\n",
    "device256 = torch.device('cpu')\n",
    "device4k = torch.device('cpu')\n",
    "\n",
    "### ViT_256 + ViT_4K loaded independently (used for Attention Heatmaps)\n",
    "model256 = get_vit256(pretrained_weights=pretrained_weights256, device=device256)\n",
    "model4k = get_vit4k(pretrained_weights=pretrained_weights4k, device=device4k)\n",
    "\n",
    "### ViT_256 + ViT_4K loaded into HIPT_4K API\n",
    "model = HIPT_4K(pretrained_weights256, pretrained_weights4k, device256, device4k)\n",
    "model.eval()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Standalone HIPT_4K Model Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Input Shape: torch.Size([1, 3, 4096, 4096])\n",
      "Output Shape: torch.Size([1, 192])\n"
     ]
    }
   ],
   "source": [
    "region = Image.open('./image_demo/image_4k.png')\n",
    "x = eval_transforms()(region).unsqueeze(dim=0)\n",
    "print('Input Shape:', x.shape)\n",
    "print('Output Shape:', model.forward(x).shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# HIPT_4K Attention Heatmaps\n",
    "Code for producing attention results (for [256 x 256], [4096 x 4096], and hierarchical [4096 x 4096]) can be run (as-is) below. There are several ways these results can be run:\n",
    "1. **hipt_4k.py** Class (Preferred): This class blends inference and heatmap creation in a seamless and more object-oriented manner, and is where I am focusing my future code development around.\n",
    "2. Helper Functions in **hipt_heatmap_utils.py** (Soon-to-be-deprecated): Heatmap creation was originally written as helper functions. May be more useful and easier from research perspective.\n",
    "\n",
    "Please use whatever is most helpful for your use case :) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 256 x 256 Demo (Saving Attention Maps Individually)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "patch = Image.open('./image_demo/image_256.png')\n",
    "output_dir = './attention_demo/256_output_indiv/'\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "create_patch_heatmaps_indiv(patch=patch, model256=model256, \n",
    "                            output_dir=output_dir, fname='patch',\n",
    "                            cmap=light_jet, device256=device256)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 256 x 256 Demo (Concatenating + Saving Attention Maps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "patch = Image.open('./image_demo/image_256.png')\n",
    "output_dir = './attention_demo/256_output_concat/'\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "create_patch_heatmaps_concat(patch=patch, model256=model256, \n",
    "                            output_dir=output_dir, fname='patch',\n",
    "                            cmap=light_jet, device256=device256)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4096 x 4096 Demo (Saving Attention Maps Individually)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "region = Image.open('./image_demo/image_4k.png')\n",
    "output_dir = './attention_demo/4k_output_indiv/'\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "create_hierarchical_heatmaps_indiv(region, model256, model4k, \n",
    "                                   output_dir, fname='region', \n",
    "                                   scale=2, threshold=0.5, cmap=light_jet, alpha=0.5,\n",
    "                                   device256=device256, device4k=device4k)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4096 x 4096 Demo (Concatenating + Saving Attention Maps)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "region = Image.open('./image_demo/image_4k.png')\n",
    "output_dir = './attention_demo/4k_output_concat/'\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "create_hierarchical_heatmaps_concat(region, model256, model4k, \n",
    "                                   output_dir, fname='region', \n",
    "                                   scale=2, cmap=light_jet, alpha=0.5,\n",
    "                                   device256=device256, device4k=device4k)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python HIPT_env",
   "language": "python",
   "name": "hipt_env"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}