Ask for help!This is the Compiling and Optimizing a Model with TVMC experiment for the TVM website.When conducting the procedure of Automatically Tuning the ResNet Model, the following errors occur:
The error seems clear to me. It says the model file cannot be downloaded, so the rest errors don’t matter.
Thank you for your answer, but I don’t think it’s the.onnx file that doesn’t download because the.onnx file does exist and executing other commands does produce output files (such as resnet50-v2-7-tvm. Tar, Prediction.npz) and the corresponding statement output.In other words, all the steps prior to auto-tuning can be performed, although there are still some caveats… Anyway, thank you for your help and welcome to continue discussing this issue.
Ok if that wasn’t a case, then the error message meant no tuning tasks were extracted from the model. Possible reasons:
- The model input has dynamic shapes (e.g., dynamic batch size).
- The model has no tunable workloads (e.g., conv2d, dense, etc), but it should not be a case for ResNet.
To dive into the root cause, it would be helpful if you could provide a minimal script to reproduce the problem.
I found that the .onnx file size is 0KB, which I think might be the problem.I re-downloaded the file and tuned it again, and the results were normal.It was a bit dramatic, but fortunately the problem was solved.Thank you for your help! ps: Why does this 0KB file not affect the results of previous operations?