Richard Li comments

Results 15 comments of


                                            Richard Li

hf转换后的13B model无法达到公布性能，求公开模型hf格式的模型权重哈希值以及transformer版本

我也遇到了同样的问题，13B的效果比7B的要差不少。我是自己转换的模型，不知道问题出在哪里。模型用的是下面两个： https://huggingface.co/minlik/chinese-alpaca-7b-merged https://huggingface.co/minlik/chinese-alpaca-13b-merged

hf转换后的13B model无法达到公布性能，求公开模型hf格式的模型权重哈希值以及transformer版本

> > 我也遇到了同样的问题，13B的效果比7B的要差不少。我是自己转换的模型，不知道问题出在哪里。模型用的是下面两个： https://huggingface.co/minlik/chinese-alpaca-7b-merged https://huggingface.co/minlik/chinese-alpaca-13b-merged > > 请问您测试的效果如何？我看你用的是llama-7b，问答用alpaca-7b

有人量化了13B模型可以提供下载链接的吗

我上传了一份 https://huggingface.co/minlik/chinese-alpaca-13b-quantized

有人量化了13B模型可以提供下载链接的吗

> @minlik 请问这个 13B 是普通版还是 plus 版？普通版

我合并+量化了 7B 和 13B 的模型提供给大家下载，并写了使用方法

> @mcu13321 你测试的输出似乎不是很对劲，有可能模型不完整。😂 > > 这个13B的模型跟7B的相比，效果比较差。是merge的时候出了问题吗？有办法验证最终合成的模型是否有问题吗？我可以再重新合一下模型试试效果。

Question about whether Yi-VL-6B can be fine-tuned using its own dataset

I added the finetuning scripts. See #368

Question about whether Yi-VL-6B can be fine-tuned using its own dataset

> WARNING: tokenization mismatch I didn't encounter the same issue. Could you please share your training scripts? After reviewing the code, I noticed that the WARNING might be caused by...

Question about whether Yi-VL-6B can be fine-tuned using its own dataset

@a2382625920 I thought of another possibility. The training code is modified from LLaVA. If you have installed llava locally, you can uninstall it and try again.

Question about whether Yi-VL-6B can be fine-tuned using its own dataset

> > 我想到了另一种可能性。训练代码是从 LLaVA 修改而来的。如果您已在本地安装了 llava，则可以将其卸载并重试。 > > I did use llava's virtual environment to run the code, and after I uninstalled it and installed Yi's installer environment, the following...

Features will release training code of VL?

I added the finetuning scripts. See #368