MiniCPM-o icon indicating copy to clipboard operation
MiniCPM-o copied to clipboard

Empty examples in data collator

Open aamir-gmail opened this issue 6 months ago • 2 comments

是否已有关于该错误的issue或讨论? | Is there an existing issue / discussion for this?

  • [X] 我已经搜索过已有的issues和讨论 | I have searched the existing issues / discussions

该问题是否在FAQ中有解答? | Is there an existing answer for this in FAQ?

  • [X] 我已经搜索过FAQ | I have searched FAQ

当前行为 | Current Behavior

While running training I get the following error [example["input_ids"] for example in examples], KeyError: 'input_ids'

I have checked the Dataset is ( SupervisedDataset ) is working properly , I have a batch size of 2 with gradient accumulation of 8 , the examples dictionary passed on to the data collator is empty

期望行为 | Expected Behavior

Training should start.

复现方法 | Steps To Reproduce

train script with full fine tuning , without lora.

运行环境 | Environment

- OS:
- Python:
- Transformers:
- PyTorch:
- CUDA (`python -c 'import torch; print(torch.version.cuda)'`):

备注 | Anything else?

image

aamir-gmail avatar Jul 31 '24 08:07 aamir-gmail