microsoft / onnxruntime-training-examples

Examples for using ONNX Runtime for model training.
MIT License
310 stars 62 forks source link

no attribute 'deepspeed_plugin' #148

Closed konabuta closed 1 year ago

konabuta commented 1 year ago

T5 sample code doens't work in my Azure ML. Any advice on this?

Traceback (most recent call last):
  File "src/Finetune/train_summarization_deepspeed_optum.py", line 684, in <module>
    main()
  File "src/Finetune/train_summarization_deepspeed_optum.py", line 586, in main
    trainer = ORTSeq2SeqTrainer(
  File "/opt/conda/envs/ptca/lib/python3.8/site-packages/optimum/onnxruntime/trainer.py", line 304, in __init__
    super().__init__(
  File "/opt/conda/envs/ptca/lib/python3.8/site-packages/transformers/trainer.py", line 345, in __init__
    self.create_accelerator_and_postprocess()
  File "/opt/conda/envs/ptca/lib/python3.8/site-packages/transformers/trainer.py", line 3825, in create_accelerator_and_postprocess
    deepspeed_plugin=self.args.deepspeed_plugin,
AttributeError: 'ORTSeq2SeqTrainingArguments' object has no attribute 'deepspeed_plugin'
prathikr commented 1 year ago

Workaround to get demo functioning: https://github.com/microsoft/onnxruntime-training-examples/pull/150

prathikr commented 1 year ago

PR against Optimum for root cause analysis and final fix: https://github.com/huggingface/optimum/issues/1133

konabuta commented 1 year ago

It works now, thanks @prathikr.