🐛 [Bug] RuntimeError: [Error thrown at core/conversion/conversionctx/ConversionCtx.cpp:147] Building serialized network failed in TensorRT

[AllesOderNicht]

[narendasan]

The error gives you an explanation of what happens and how to try to avoid it. It says you are running out of memory so its unable to find a kernel that works and to try increasing the workspace size. Change the value of the setting max_workspace_size to provide tensorrt more memory.

[AllesOderNicht]

Thanks, Now it seems to compile fine.But it still doesn't work properly. I don't know what the reason is, it doesn't show the error.

[narendasan]

You can try increasing logging verbosity to see what is happening at runtime by adding a context

with torch_tensorrt.logging.debug(): 
    model(input_data, input_data)

[github-actions]

This issue has not seen activity for 90 days, Remove stale label or comment or this will be closed in 10 days