423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-12-17 04:10:36 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #1026

Browse Source
This commit is contained in:
hiyouga
2023-09-27 22:57:09 +08:00
parent 35fa94723c
commit b0b0138e1d
2 changed files with 10 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
src/llmtuner/tuner/ppo/trainer.py
View File

@@ -110,9 +110,17 @@ class CustomPPOTrainer(PPOTrainer, Trainer):
# Run PPO step
stats = self.step(queries, responses, rewards)
self.tokenizer.padding_side = "left" # restore padding side
loss_meter.update(stats["ppo/loss/total"], n=len(rewards))
loss_meter.update(float(stats["ppo/loss/total"]), n=len(rewards))
reward_meter.update(torch.stack(rewards).mean().item(), n=len(rewards))
if self.config.log_with is not None:
try:
batch["query"] = self.tokenizer.batch_decode(queries, skip_special_tokens=True)
batch["response"] = self.tokenizer.batch_decode(responses, skip_special_tokens=True)
self.log_stats(stats, batch, rewards)
except:
logger.warning("Failed to save stats due to unknown errors.")
self.state.global_step += 1
self.log_callback.on_step_end(self.args, self.state, self.control)