奖励模型训练报错'NoneType' object has no attribute 'neftune_noise_alpha'

[ChenBinfighting1]

Reminder

• I have read the README and searched the existing issues.

Reproduction

运行train_web，下页面导入数据后，采用glm2和llama2训练奖励模型，报错：
Traceback (most recent call last):
File "/home/.conda/envs/llama_factory/lib/python3.10/threading.py", line 1016, in _bootstrap_inner
self.run()
File "/home/.conda/envs/llama_factory/lib/python3.10/threading.py", line 953, in run
self._target(*self._args, **self._kwargs)
File "/home/jc/nlp/llm/LLaMA-Factory/src/llmtuner/train/tuner.py", line 28, in run_exp
run_rm(model_args, data_args, training_args, finetuning_args, callbacks)
File "/home/jc/nlp/llm/LLaMA-Factory/src/llmtuner/train/rm/workflow.py", line 50, in run_rm
train_result = trainer.train(resume_from_checkpoint=training_args.resume_from_checkpoint)
File "/home/.conda/envs/llama_factory/lib/python3.10/site-packages/transformers/trainer.py", line 1467, in train
self.model = self._activate_neftune(self.model)
File "/home/.conda/envs/llama_factory/lib/python3.10/site-packages/transformers/trainer.py", line 629, in _activate_neftune
embeddings.neftune_noise_alpha = self.neftune_noise_alpha
AttributeError: 'NoneType' object has no attribute 'neftune_noise_alpha'

请问有碰到过吗？感谢

Expected behavior

No response

System Info

No response

Others

No response

[hiyouga]

代码是最新？

[ChenBinfighting1]

代码是最新？
是的

[hiyouga]

奖励模型不应该使用 NEFTune

[ChenBinfighting1]

发现运行train+web自动带上了这个参数

[ChenBinfighting1]

现在手动运行train_bash，可以训练，是不是web端那边有问题

[hiyouga]

ok，我修复一下