github/ColossalAI
1
0
Fork 0
mirror of https://github.com/hpcaitech/ColossalAI.git synced 2025-07-10 13:53:19 +00:00
Code Issues Packages Projects Releases Wiki Activity

remove unused code

Browse Source
This commit is contained in:
YeAnbang 2025-04-04 10:11:11 +08:00
parent 09a3173a49
commit 061d8cb3b6
1 changed files with 1 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
applications/ColossalChat/coati/distributed/grpo_consumer.py
View File

@ -252,10 +252,6 @@ class GRPOConsumer(BaseConsumer):
- (inputs["reference_action_log_probs"] - action_log_probs)
- 1
)
decode_tokens_100 = self.tokenizer.batch_decode(
input_ids_forward_micro_batch[:, -num_action:],
skip_special_tokens=False,
)
loss, skip_update, _ = self.policy_loss_fn(
action_log_probs,
action_log_probs,
@ -277,7 +273,7 @@ class GRPOConsumer(BaseConsumer):
loss = policy_model_outputs["loss"]
if self.booster.plugin.stage_manager.is_last_stage():
# calculate kl
# calculate kl, as we cannot do this inside callback, kl needs be calculate again
action_logits = policy_model_outputs["outputs"]["logits"]
action_log_probs = calc_action_log_probs(
action_logits / self.generate_config["temperature"],