BrettRyland
BrettRyland
Additionally, replacing the ```python x = x + 1 if x.shape[0] > 1 else x # - 1 # Uncommenting this results in the compilation succeeding. ``` line in `MyMod.forward`...
Here's the same stacktrace, but with `torch-tensorrt` compiled in debug mode with `pip3 install -e ./py` so that it shows line numbers and extra debugging: ```gdb brett@br-workhorse:~/repos/Autosensor/NN$ gdb --ex=r --args...
Note: I've been able to partially work around this (partially because I'm getting a different issue afterwards) by modifying `torchvision/ops` with ```diff diff -x __pycache__ /home/brett/github/torchvision/torchvision/ops/ps_roi_align.py /home/brett/.local/lib/python3.10/site-packages/torchvision/ops/ps_roi_align.py 2a3 > from...
As a temporary work-around while this and other issues are getting fixed, is it possible to tell `torch_tensorrt.compile` to ignore certain sections of the model? E.g., to compile the backbone...
> As a temporary work-around while this and other issues are getting fixed, is it possible to tell `torch_tensorrt.compile` to ignore certain sections of the model? E.g., to compile the...
Sorry to re-raise this issue, but I'm still getting the same runtime error for deformable convolutions on the latest build of master (commit 91a92ca4), which includes PR #1024. Expanding upon...
> @BrettRyland Could you please try this PR: #1067 We refactored this part recently, and I tried your model with this PR, didn't find the issue you had. I still...
> @BrettRyland Did you clear your cache? I could get your model work after I used that PR. You could do: `pip3 uninstall torch_tensorrt` to uninstall the previously installed torch_tensorrt....
OK, I've reduced my model to a smaller repro script [trt_bug.py](https://github.com/pytorch/TensorRT/files/8861004/trt_bug.py.txt) which still gives ``` RuntimeError: [Error thrown at core/partitioning/shape_analysis.cpp:68] Expected ivalues_maps.count(input) to be true but got false Could not...
I managed to bypass the issue by using ```charp anchor_always_score = scores[:, self.anchor_always_index, :] # Avoid indexed assignment, which quantisation doesn't like. https://discuss.pytorch.org/t/how-to-get-around-proxy-object-does-not-support-item-assignment/122655 anchor_always_score *= 0 anchor_always_score += self.false_anchor_score ```...