Skip to content
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

[E2E] Torchbench float16 training timm_efficientnet accuracy regression #904

Open
mengfei25 opened this issue Sep 12, 2024 · 0 comments
Open
Assignees
Milestone

Comments

@mengfei25
Copy link
Contributor

🐛 Describe the bug

xpu train timm_efficientnet
[WARNING] Failed to create Level Zero tracer: 2013265921
(I): Detected 2048 spills, recompiling the kernel using large GRF mode
(I): Kernel has now 0 spills
(I): Detected 1024 spills, recompiling the kernel using large GRF mode
(I): Kernel has now 0 spills
(I): Detected 1024 spills, recompiling the kernel using large GRF mode
(I): Kernel has now 0 spills
(I): Detected 1024 spills, recompiling the kernel using large GRF mode
(I): Kernel has now 0 spills
E0907 00:54:25.073000 2172206 site-packages/torch/_dynamo/utils.py:1798] RMSE (res-fp64): 0.00040, (ref-fp64): 0.00010 and shape=torch.Size([4, 96, 1, 1]). res.dtype: torch.float16, multiplier: 3.000000, tol: 0.001000, use_larger_multiplier_for_smaller_tensor: 0
E0907 00:54:25.074000 2172206 site-packages/torch/_dynamo/utils.py:1670] Accuracy failed for key name blocks.1.0.se.conv_reduce.weight.grad
fail_accuracy

Versions

last known good:
pytorch: 351509000650b477e3a09cbf487288de5d8af616
torch-xpu-ops: aae765a

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Projects
None yet
Development

No branches or pull requests

3 participants