coffezhou opened a new issue, #18601: URL: https://github.com/apache/tvm/issues/18601
### Expected behavior The ConvTranspose operator in TVM should produce right shape. ### Actual behavior For the following model, <img width="580" height="475" alt="Image" src="https://github.com/user-attachments/assets/ca12b78e-77de-43a4-9395-5490e7fd973f"; /> it can be executed by onnxruntime and onnx's ReferenceEvaluator, the shapes of results are as follows: ``` onnxruntime: (1, 6, 56, 56) ReferenceEvaluator: (1, 6, 56, 56) ``` However, the shape of results produced by TVM is: ``` TVM: (1, 6, 55, 55) ``` which is different from those of onnxruntime and onnx's ReferenceEvaluator. According to the documents of [ConvTranspose](https://onnx.ai/onnx/operators/onnx__ConvTranspose.html), the shape of the output is calculated via the following equation: ``` output_shape[i] = stride[i] * (input_size[i] - 1) + output_padding[i] + ((kernel_shape[i] - 1) * dilations[i] + 1) - pads[start_i] - pads[end_i] ``` For the last dim, the shape should be: 2*(28-1) + 1 + ((3-1) + 1) - 1 - 1 = 54 + 1 + 3 - 1 - 1 = 56 This is confusing that why the shape of TVM's results is (1, 6, 55, 55). ### Environment OS: Ubuntu 20.04 TVM: 0.23.dev0 (https://github.com/apache/tvm/commit/f4e28d3153323ad97a7e74740c9fb22300fd6cd0) onnxruntime: 1.23.2 ### Steps to reproduce This bug can be reproduced by the following code with the model in the attachment. ``` from typing import Dict, List, Literal, Optional import sys import os import numpy as np import onnx from onnx.reference import ReferenceEvaluator import onnxruntime from onnx import ModelProto, TensorProto, helper import tvm import tvm.testing from tvm import relax from tvm.relax.frontend.onnx import from_onnx import argparse import pickle def test() -> None: onnx_model = onnx.load("11.onnx") # Configure model format. onnx_model.ir_version = 8 onnx_model.opset_import[0].version = 14 with open("inputs.pkl", 'rb') as fp: inputs = pickle.load(fp) # Run the model through onnx to get the expected result. try: ort_session = onnxruntime.InferenceSession( onnx_model.SerializeToString(), providers=["CPUExecutionProvider"] ) ort_output = ort_session.run([], inputs) except Exception as e: print(e) print("This model cannot be executed by onnxruntime!") sys.exit(1) print("onnxruntime:", ort_output[0].shape) # ReferenceEvaluator sess = ReferenceEvaluator("11.onnx") re_output = sess.run(None, inputs) print("ReferenceEvaluator:", re_output[0].shape) tvm.testing.assert_allclose(re_output[0], ort_output[0], rtol=0.1, atol=0.1) # TVM # Convert the onnx model into relax through the onnx importer. tvm_model = from_onnx(onnx_model, opset=14, keep_params_in_input=True) # Convert operators for inference mode. tvm_model = relax.transform.DecomposeOpsForInference()(tvm_model) # Legalize any relax ops into tensorir. tvm_model = relax.transform.LegalizeOps()(tvm_model) # Separate model from parameters. tvm_model, params = relax.frontend.detach_params(tvm_model) # Compile the relax graph into a VM then run. with tvm.transform.PassContext(opt_level=3): ex = tvm.compile(tvm_model, target="llvm") vm = relax.VirtualMachine(ex, tvm.cpu()) # Prepare inputs. input_list = [ inputs[key.name_hint] for key in tvm_model["main"].params if key.name_hint in inputs ] if params: input_list += params["main"] # Run model and check outputs. vm.set_input("main", *input_list) vm.invoke_stateful("main") tvm_output = vm.get_outputs("main") print("TVM:", tvm_output.shape) if __name__ == "__main__": test() ``` [testcase.zip](https://github.com/user-attachments/files/24305443/testcase.zip) ### Triage Please refer to the list of label tags [here](https://github.com/apache/tvm/wiki/Issue-Triage-Labels) to find the relevant tags and add them below in a bullet format (example below). * needs-triage -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
