THUDM
[BUG/Help] <title>bash train.sh出错，AttributeError: 'Seq2SeqTrainer' object has no attribute 'is_deepspeed_enabled'
Is there an existing issue for this?
- [X] I have searched the existing issues
Current Behavior
bash train.sh
PRE_SEQ_LEN=256 LR=2e-2 NUM_GPUS=1
torchrun --standalone --nnodes=1 --nproc-per-node=$NUM_GPUS main.py
--do_train
--train_file ../data/train_data.json
--validation_file ../data/test_data.json
--preprocessing_num_workers 10
--prompt_column instruction
--response_column output
--overwrite_cache
--model_name_or_path ../model
--output_dir output/adgen-chatglm2-6b-pt-$PRE_SEQ_LEN-$LR
--overwrite_output_dir
--max_source_length 64
--max_target_length 128
--per_device_train_batch_size 1
--per_device_eval_batch_size 1
--gradient_accumulation_steps 16
--predict_with_generate
--max_steps 3000
--logging_steps 10
--save_steps 1000
--learning_rate $LR
--pre_seq_len $PRE_SEQ_LEN
Expected Behavior
我搜了一下，说可能是transformers版本问题，但我这不就是4.30.2吗
Steps To Reproduce
bash train.sh
Environment
- OS: Ubuntu
- Python:3.9
- Transformers:4.30.2
- PyTorch:2.0
- CUDA Support (`python -c "import torch; print(torch.cuda.is_available())"`) :
Anything else?
No response
我直接把 trainer_pt_utils.py 841行的 self.is_deepspeed_enabled: 改成了 if False: 反正没用到deepspeed，先跑跑试试，暂时没问题
