Skip to content

Fix training gradient underflow in quantization tests #13539

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.

Sign up for GitHub

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

Open
jiqing-feng wants to merge 3 commits into
base: main
Choose a base branch
from
+2 −1
Open

Fix training gradient underflow in quantization tests #13539

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
3152585
Fix training gradient underflow in quantization tests
jiqing-feng Apr 22, 2026
04fc264
Merge branch 'main' into torchao-fix-training-underflow
jiqing-feng Apr 22, 2026
f08c764
Merge branch 'main' into torchao-fix-training-underflow
jiqing-feng Apr 22, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 2 additions & 1 deletion tests/models/testing_utils/quantization.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -407,7 +407,8 @@ def _test_quantization_training(self, config_kwargs):
# Step 3: run forward and backward pass
inputs = self.get_dummy_inputs()

with torch.amp.autocast(torch_device, dtype=torch.float16):
# Use bfloat16 instead of float16 to avoid gradient underflow with quantized layers
with torch.amp.autocast(torch_device, dtype=torch.bfloat16):
out = model(**inputs, return_dict=False)[0]
out.norm().backward()

Expand Down
Loading