[CM] Enable inference for ResNet-50 with TVM - PyTorch backend

[KJlaccHoeUM9l]

Based on this tutorial for TVM-ONNX (MLPerf inference - Python - ResNet50 FP32 - ImageNet - TVM - CPU - Offline), you need to add the ability to run inference for TVM with PyTorch frontend.

To run this inference, you need to change backend parameter as follows:

--backend=tvm-pytorch

When you select this backend, you will be taken to the corresponding branches here and here.

The result of the work should be a prepared PR for the official CK.

[KJlaccHoeUM9l]

CC @Ailurus1

[SennikovAndrey]

Tried to use next script for --backend=tvm-pytorch

-quiet          --clean

And after received next error.

Traceback (most recent call last):
  File "/root/CM/repos/mlcommons@ck/cm-mlops/script/get-ml-model-resnet50-tvm/process.py", line 39, in <module>
    pytorch_model = torch.jit.load(model_path)
  File "/usr/local/lib/python3.8/dist-packages/torch/jit/_serialization.py", line 162, in load
    cpp_module = torch._C.import_ir_module(cu, str(f), map_location, _extra_files)
RuntimeError: PytorchStreamReader failed reading zip archive: failed finding central directory

CM error: Portable CM script failed (name = get-ml-model-resnet50-tvm, return code = 256)

I have one new small idea about solution. Will try it during the next time.

[SennikovAndrey]

Correct code:

cm run script --tags=run,mlperf,inference,generate-run-cmds \
         --adr.python.name=mlperf \
         --adr.python.version_min=3.8 \
         --adr.tvm.tags=_pip-install \
         --submitter="Community" \
         --implementation=python \
         --hw_name=default \
         --model=resnet50 \
         --backend=tvm-pytorch \
         --device=cpu \
         --scenario=Offline \
         --mode=accuracy \
         --test_query_count=5 \
         --quiet \
         --clean

[SennikovAndrey]

It was the problem appeared from different types of files ('.pt' and 'pth'). I fixed it by using if/else construction and considering the two cases separately. But now I have some problems with some next commands.

Now i trying to fix the following error:

Traceback (most recent call last):
  File "/root/CM/repos/Deelvin@ck/cm-mlops/script/get-ml-model-resnet50-tvm/process.py", line 53, in <module>
    mod, params = relay.frontend.from_pytorch(pytorch_model, shape_list)
  File "/usr/local/lib/python3.8/dist-packages/tvm/relay/frontend/pytorch.py", line 4503, in from_pytorch
    graph = script_module.graph.copy()
  File "/usr/local/lib/python3.8/dist-packages/torch/nn/modules/module.py", line 1269, in __getattr__
    raise AttributeError("'{}' object has no attribute '{}'".format(
AttributeError: 'ResNet' object has no attribute 'graph'

CM error: Portable CM script failed (name = get-ml-model-resnet50-tvm, return code = 256)

[Ailurus1]

Solved here