使用deepspeed run run_rm.sh

[yangzhipeng1108]

Could not estimate the number of tokens of the input, floating-point operations will not be computed Traceback (most recent call last): File "/root/nas-share/chat/MedicalGPT-main/reward_modeling.py", line 643, in main() File "/root/nas-share/chat/MedicalGPT-main/reward_modeling.py", line 615, in main train_result = trainer.train(resume_from_checkpoint=checkpoint) File "/usr/local/conda/lib/python3.9/site-packages/transformers/trainer.py", line 1537, in train return inner_training_loop( File "/usr/local/conda/lib/python3.9/site-packages/transformers/trainer.py", line 1881, in _inner_training_loop self.optimizer.step() AttributeError: 'DummyOptim' object has no attribute 'step'

RewardTrainer是否支持deepspeed

[xxm1668]

deepspeed版本号问题

[yangzhipeng1108]

请问大神，什么版本比较合适

[xxm1668]

0.9.2

[yinheng-gh]

我是在运行pretrain的时候遇到这个问题，把deepspeed改成0.9.2会在accelerate检查依赖版本时提示最低需要0.9.3，然后改成0.9.3，还是会遇到上面那个问题