ConvolutionBackwardModule2DStrided_basic fails on the current codebase

[chudur-budur]

ConvolutionBackwardModule2DStrided_basic fails on the current codebase.

Similar to: issue #31

(torch-mlir) $USER@$HOST:/localdisk/work/$USER/torch-mlir/projects/pt1|(cpu-proto)> python -m e2e_testing.main -c "cpuproto" -f "ConvolutionBackwardModule2DStrided_basic" -s -v --enable-timer
in shape:  torch.Size([1, 128, 128])
 w shape:  (8192, 16384)
 b shape:  (8192,)
golden_trace: elapsed 0.1507 ms
Compiling ConvolutionBackwardModule2DStrided_basic...
compile: elapsed 0.0023 ms
Running ConvolutionBackwardModule2DStrided_basic...
JIT: elapsed 21.2993 ms
TorchDynamoTestConfig.run(): elapsed 21.3016 ms
run: elapsed 21.3029 ms
FAIL - "ConvolutionBackwardModule2DStrided_basic"

Unexpected outcome summary: (cpuproto)

****** Failed tests - 1 tests
    FAIL - "ConvolutionBackwardModule2DStrided_basic"
        Runtime error: Traceback (most recent call last):
          File "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir_e2e_test/framework.py", line 384, in compile_and_run_test
            trace = config.run(compiled, golden_trace)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
          File "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir_e2e_test/configs/torchdynamo.py", line 157, in run
            module = jit(artifact,
                     ^^^^^^^^^^^^^
          File "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir_e2e_test/configs/torchdynamo.py", line 137, in jit
            return _lower_mlir_module(verbose, output_type, mlir_module, ir_file)
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
          File "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/__init__.py", line 291, in _lower_mlir_module
            run_pipeline_with_repro_report(
          File "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir/compiler_utils.py", line 92, in run_pipeline_with_repro_report
            raise TorchMlirCompilerError(trimmed_message) from None
        torch_mlir.compiler_utils.TorchMlirCompilerError: Lowering Torch Backend IR -> Linalg-on-Tensors Backend IR failed with the following diagnostics:


        python exception: Failure while executing pass pipeline:
        error: "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir_e2e_test/configs/torchdynamo.py":121:0: 'linalg.transpose' op size of permutation 2 does not match the argument rank 4
        note: "/localdisk/work/$USER/torch-mlir/build/tools/torch-mlir/python_packages/torch_mlir/torch_mlir_e2e_test/configs/torchdynamo.py":121:0: see current operation:
        %185 = "linalg.transpose"(%34, %184) <{permutation = array<i64: 1, 0>}> ({
        ^bb0(%arg3: f32, %arg4: f32):
        "linalg.yield"(%arg3) : (f32) -> ()
        }) : (tensor<1x2x8x8xf32>, tensor<2x1x8x8xf32>) -> tensor<2x1x8x8xf32>

        For Torch-MLIR developers, the error can be reproduced with:
        $ torch-mlir-opt -pass-pipeline='builtin.module(torch-backend-to-linalg-on-tensors-backend-pipeline)' /tmp/ConvolutionBackwardModule2DStrided.mlir
        Add '-mlir-print-ir-after-all -mlir-disable-threading' to get the IR dump for debugging purpose.



Summary:
    Failed: 1