-
Notifications
You must be signed in to change notification settings - Fork 56
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Release/2.4] Increase the precision from float32 to float64 for test related to linear algebra #1748
base: release/2.4
Are you sure you want to change the base?
Conversation
…s the precision to pass the test. Reference - https://pytorch.org/docs/stable/notes/numerical_accuracy.html#linear-algebra-stability
Jenkins build for 95706699a3f7bd12c72bb7abd29d9237f48dcb93 commit finished as FAILURE Detected error during Pytorch building:
|
Jenkins build for 95706699a3f7bd12c72bb7abd29d9237f48dcb93 commit finished as FAILURE Detected error during Pytorch building:
|
Jenkins build for 95706699a3f7bd12c72bb7abd29d9237f48dcb93 commit finished as FAILURE Detected error during Pytorch building:
|
This is skipped on Nvidia, fails when made to force run on Nvidia.
Both amd and nvidia gpus -
/opt/conda/lib/python3.11/site-packages/torch/autograd/graph.py:769: UserWarning: There is a performance drop because we have not yet implemented the batching rule for aten::tril_. Please file us an issue on GitHub so that we can prioritize its implementation. (Triggered internally at ../aten/src/ATen/functorch/BatchedFallback.cpp:81.)
return Variable._execution_engine.run_backward( # Calls into the C++ engine to run the backward pass
To make it more precise:
dtype = torch.float64
This is a temporary fix and the test should be revisited when batching rule is implemented for aten::tril_.
Running the computation in float64 (as NumPy does by default) improves the precision to pass the test. Reference - https://pytorch.org/docs/stable/notes/numerical_accuracy.html#linear-algebra-stability