Skip to content

[Feat.] Support on policy distillation in fsdp #420

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

Draft
PopSoda2002 wants to merge 1 commit into
base: main
Choose a base branch
from
Draft

[Feat.] Support on policy distillation in fsdp #420

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
14 changes: 14 additions & 0 deletions miles/backends/fsdp_utils/actor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -629,6 +629,20 @@ def _train_step(self, packed_batch, reported_accum, mbs_id, grad_accum):
response_lengths = [batch["response_lengths"] for batch in unpacked_batches]

advantages = advantages.to(device=log_probs.device)

if self.args.advantage_estimator == "on_policy_distillation":
student_log_probs = log_probs
teacher_log_probs = rollout_data.get("teacher_log_probs")
response_lengths = rollout_data.get("response_lengths")
device = student_log_probs[0].device
teacher_log_probs = [t_log_prob.to(device=device) for t_log_prob in teacher_log_probs]
teacher_log_probs = [
t_log_prob[-response_length:] for t_log_prob, response_length in zip(teacher_log_probs, response_lengths)
]
advantages = [
teacher_log_prob - student_log_prob
for teacher_log_prob, student_log_prob in zip(teacher_log_probs, student_log_probs)
]
old_log_probs = old_log_probs.to(device=log_probs.device)
ppo_kl = old_log_probs - log_probs

Expand Down
Loading