Skip to content

[megatron] fix: resolve backward propagation error in megatron_actor due to shared logits tensor in-place modification #2484

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

Merged
merged 3 commits into from
Jul 24, 2025
Merged

[megatron] fix: resolve backward propagation error in megatron_actor due to shared logits tensor in-place modification #2484

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
b79e36a
Fix autograd conflict in entropy calculation
HelloWorld686 Jul 11, 2025
97ac79c
fix: optimization to make the tensor clone conditional, preventing un…
HelloWorld686 Jul 11, 2025
7fb2eca
Add a warning to let users know that it is better to enable fused ker…
HelloWorld686 Jul 17, 2025
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
10 changes: 9 additions & 1 deletion verl/workers/actor/megatron_actor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -530,9 +530,17 @@ def logits_processor(logits, label, label_mask):
assert label.shape == label_mask.shape
ret = {}
if calculate_entropy:
logits_bak = logits.clone()
logger.warning_once(
"For memory-efficient computation, enable fused kernels via "
"`actor_rollout_ref.model.use_fused_kernels=True`. "
"The current `clone()` operation ensures correctness but increases memory usage."
)
entropy = vocab_parallel_entropy(logits)
ret["entropy"] = entropy
log_probs = vocab_parallel_log_probs_from_logits(logits, label)
else:
logits_bak = logits
log_probs = vocab_parallel_log_probs_from_logits(logits_bak, label)
log_probs = log_probs.masked_fill(~label_mask, 0.0)
ret["log_probs"] = log_probs
return ret
Expand Down
Loading