ceos/unsloth
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Update rl_replacements.py

Browse source
This commit is contained in:
Daniel Han 2025-02-16 21:06:47 -08:00
commit 2a26cdc184
1 changed files with 4 additions and 3 deletions
Download patch file Download diff file Expand all files Collapse all files

7
unsloth/models/rl_replacements.py
View file

@ -216,7 +216,8 @@ def grpo_trainer_compute_loss(function_name, function):
# attention_mask = torch.cat([prompt_mask, completion_mask], dim=1)
attention_mask = None
logits_to_keep = completion_ids.size(1) # we only need to compute the logits for the completion tokens
_input_ids = input_ids
_logits_to_keep = logits_to_keep
per_token_logps = self._get_per_token_logps(model, input_ids, attention_mask, logits_to_keep)
# Compute the KL divergence between the model and the reference model
@ -238,8 +239,8 @@ def grpo_trainer_compute_loss(function_name, function):
if RL_REPLACEMENTS["count"] == 10: raise
else: RL_REPLACEMENTS["count"] = 1
RL_REPLACEMENTS["data"] = (
ref_per_token_logps, per_token_logps, input_ids, completion_mask, self.beta, advantages,
loss, completion_length, mean_kl, completion_ids,
ref_per_token_logps, per_token_logps, _input_ids, completion_mask, self.beta, advantages,
loss, completion_length, mean_kl, completion_ids, _logits_to_keep,
)
# Log the metrics
# completion_length = self.accelerator.gather_for_metrics(completion_mask.sum(1)).float().mean().item()