[BUG/Help] <title>验证时报错batch_size不匹配

[laiqinghan]

Is there an existing issue for this?

• [X] I have searched the existing issues

Current Behavior

Traceback (most recent call last): File "/home/cong009/.pycharm_helpers/pydev/pydevd.py", line 1415, in _exec pydev_imports.execfile(file, globals, locals) # execute the script File "/home/cong009/.pycharm_helpers/pydev/_pydev_imps/_pydev_execfile.py", line 18, in execfile exec(compile(contents+"\n", file, 'exec'), glob, loc) File "/mnt/laiqinghan/project/1_FUTU/llmModels/accelerate_chatglm6b/singleGPU_semantic_finetuning.py", line 232, in main() File "/mnt/laiqinghan/project/1_FUTU/llmModels/accelerate_chatglm6b/singleGPU_semantic_finetuning.py", line 217, in main out = model(**input) File "/home/cong009/miniconda3/envs/LLM/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl return forward_call(*input, **kwargs) File "/home/cong009/.cache/huggingface/modules/transformers_modules/THUDM/chatglm-6b/35ca52301fbedee885b0838da5d15b7b47faa37c/modeling_chatglm.py", line 1215, in forward loss = loss_fct(shift_logits.view(-1, shift_logits.size(-1)), shift_labels.view(-1)) File "/home/cong009/miniconda3/envs/LLM/lib/python3.8/site-packages/torch/nn/modules/module.py", line 1194, in _call_impl return forward_call(*input, **kwargs) File "/home/cong009/miniconda3/envs/LLM/lib/python3.8/site-packages/torch/nn/modules/loss.py", line 1174, in forward return F.cross_entropy(input, target, weight=self.weight, File "/home/cong009/miniconda3/envs/LLM/lib/python3.8/site-packages/torch/nn/functional.py", line 3026, in cross_entropy return torch._C._nn.cross_entropy_loss(input, target, weight, _Reduction.get_enum(reduction), ignore_index, label_smoothing) ValueError: Expected input batch_size (224) to match target batch_size (66).

有大佬知道啥原因吗

Expected Behavior

No response

Steps To Reproduce

a

Environment

- OS:
- Python:
- Transformers:
- PyTorch:
- CUDA Support (`python -c "import torch; print(torch.cuda.is_available())"`) :

Anything else?

No response

[jiejue26]

我做验证的时候也有这个问题，请问解决了吗

[zhaoqf123]

https://github.com/THUDM/ChatGLM-6B/issues/620#issuecomment-1537448881

[labixiaoK]

1. eval_dataset.map 时候： 将 preprocess_function_eval 更改为 preprocess_function_train
2. do_eval 时候 ：改为 metrics = trainer.evaluate(metric_key_prefix="eval")
3. 启动脚本时候，确保 predict_with_generate 为 False 就可以了