EeyoreLee

BCELoss无效，DiceLoss可以

2

当我选择BCELoss时，loss确实也可以收敛，但是F1 score都是0 。我确定F1的计算方式是没问题的，因为我看了输出的logits，都是负值，意味着预测的都是O标签。 但是我只是将loss functional 改成 DiceLoss的时候，一切正常work了。我不知道问题出在哪里，我没有直接使用该仓库的代码，如果有需要，我可以贴上我的脚本。

### Is there an existing issue for this? - [X] I have searched the existing issues ### Current Behavior 在训练chat的时候，通过什么指标来评估对话内容的质量或者说在训练的时候什么情况认为可以中断训练 ### Expected Behavior _No response_ ### Steps To Reproduce null...

[Bug] Some space out of range strip when using `-l` argument.

2

``` def do_something(): xs = [] for x in xs: y = 'y' if y: pass def do_something_else(): return '-path 1' ``` the line 5 (one-based) has 8 spaces or...

1. add a arg named `add_eot_token` to avoid to add eot token when `args.add_eot_token` is False. (fix step2 llama2) 2. remove some redundant code. like `chosen_token["input_ids"] = chosen_token["input_ids"]` 3. avoid...

| Model | Dataset | Peft | Step | Acc | PS | |:---:|:---:|:---:|:---:|:---:|:---:| | llama2-7b | Dahoas/full-hh-rlhf | LoRA | 2 | 0.64 | \ | | llama2-7b |...

The new version format: {year}.{month}.1{day}{hour}{minute}{second}

1

Just worked on prerelease version and will also work on next release version. So it's better if you attach the version to issue here that I can easily know which...