support adaptive_max_pooling, fix adpative_avg_pooling bug

[inocsin]

Description

Please include a summary of the change and which issue is fixed. Please also include relevant motivation and context. List any dependencies that are required for this change.

Type of change

• Bug fix (non-breaking change which fixes an issue)
  Fix aten::adaptive_avg_pool2d bug when input size is not integer multiple of output size, update the test case
• New feature (non-breaking change which adds functionality)
  Support aten::adaptive_max_pool2d with test case
  -Bug fix
  Remove "#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1)" part.
  Since the TensorRT dependency is updated, the original GPU part will never be used, all the operator will be forced to run on cpu. So I remove the TensorRT flag. But the GPU part failed with unknow reason after the flag was removed, so I refactored the gpu part, now it can run on GPU correctly.

Checklist:

• My code follows the style guidelines of this project (You can use the linters)
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas and hacks
• I have made corresponding changes to the documentation
• I have added tests to verify my fix or my feature
• New and existing unit tests pass locally with my changes

[github-actions]

There are some changes that do not conform to C++ style guidelines:

diff --git a/workspace/core/conversion/converters/impl/plugins/adaptive_max_pool2d_plugin.cpp b/tmp/changes.txt
index 84a2e1f..597c30e 100644
--- a/workspace/core/conversion/converters/impl/plugins/adaptive_max_pool2d_plugin.cpp
+++ b/tmp/changes.txt
@@ -194,10 +194,8 @@ int AdaptiveMaxPool2dPlugin::enqueue(
    cudaStream_t stream) {
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1)
  at::Tensor input = at::from_blob((void*)inputs[0], util::toVec(inputDesc->dims), [](void*) {}, tensor_options_);
-  at::Tensor output = at::from_blob(
-      outputs[0], util::volume(outputDesc->dims), [](void*) {}, tensor_options_);
-  at::Tensor indices = at::from_blob(
-      outputs[1], util::volume(outputDesc->dims), [](void*) {}, index_tensor_options_);
+  at::Tensor output = at::from_blob(outputs[0], util::volume(outputDesc->dims), [](void*) {}, tensor_options_);
+  at::Tensor indices = at::from_blob(outputs[1], util::volume(outputDesc->dims), [](void*) {}, index_tensor_options_);

  at::cuda::CUDAStream torch_stream = at::cuda::getStreamFromPool();
  at::cuda::CUDAStreamGuard torch_guard(torch_stream);
diff --git a/workspace/core/conversion/converters/impl/plugins/interpolate_plugin.cpp b/tmp/changes.txt
index ff9ba86..5689d88 100644
--- a/workspace/core/conversion/converters/impl/plugins/interpolate_plugin.cpp
+++ b/tmp/changes.txt
@@ -254,8 +254,7 @@ int InterpolatePlugin::enqueue(
    cudaStream_t stream) {
#if NV_TENSORRT_MAJOR < 7 || (NV_TENSORRT_MAJOR == 7 && NV_TENSORRT_MINOR < 1)
  at::Tensor input = at::from_blob((void*)inputs[0], util::toVec(inputDesc->dims), [](void*) {}, tensor_options_);
-  at::Tensor output = at::from_blob(
-      outputs[0], util::volume(outputDesc->dims), [](void*) {}, tensor_options_);
+  at::Tensor output = at::from_blob(outputs[0], util::volume(outputDesc->dims), [](void*) {}, tensor_options_);

  at::cuda::CUDAStream torch_stream = at::cuda::getStreamFromPool();
  at::cuda::CUDAStreamGuard torch_guard(torch_stream);
ERROR: Some files do not conform to style guidelines

[github-actions]

Code conforms to Python style guidelines

[narendasan]

Can you apply the linting?

[inocsin]

Can you apply the linting?

applied linting

[peri044]

The tests are running fine. It looks like the existing Interpolate plugin already calls at::adaptive_avg_pool2d by using the mode adaptive_pool2d https://github.com/NVIDIA/TRTorch/blob/master/core/conversion/converters/impl/plugins/interpolate_plugin.cpp#L329
Can we make use of the mode flag to support adaptive_max_pool2d as well ? Does this require a separate plugin in this PR?

[inocsin]

The tests are running fine. It looks like the existing Interpolate plugin already calls at::adaptive_avg_pool2d by using the mode adaptive_pool2d https://github.com/NVIDIA/TRTorch/blob/master/core/conversion/converters/impl/plugins/interpolate_plugin.cpp#L329
Can we make use of the mode flag to support adaptive_max_pool2d as well ? Does this require a separate plugin in this PR?

The output number, datatype and parameter number of adaptive_max_pool are different from interpolate_plugin, in that case should we use several if-else to merge these two plugin into one?

[inocsin]

@narendasan @peri044 please review the update, I have fixed some bugs of original interpolate plugin

[peri044]

Thanks @inocsin Merged these changes into the new plugins interface. Tried to cherry pick your commits but there are lot of changes in the new API which were limiting it.