LianghuiGuo comments

Results 14 comments of


                                            LianghuiGuo

Compiling on macOS fails due to ffmpeg 5

amazing!

[Bug] I downloaded the Llama-7b model on the huggingface to the local location, and then modified the path to the file location. After running the code, the following error occurred. How can I solve it?

同问，我的报错是/bin/sh: torchrun: command not found，查了下说torch1.9.1之后就支持torchrun了，我是torch2.0.1，还是报错

Hello, I used Chinese data for training and found that all the results I got have the first character as "的", such as "的里有一的男士" and "的山小上一的女人和一手子的人". I would like to ask if there is anything else I need to modify besides the data when training Chinese language.

兄弟，你训中文模型，用的啥预训练模型呀？vocab_size，词表什么的是不是要改成中文？

Can this model output Chinese image captionings?

同问

LMDeploy support accelerating DeepSeek VL models now!!! :rocket:

请问DeepSeek-VL-chat-7B如果使用LMDploy加速，和DeepSeek-VL-chat-1.3B相比，推理速度如何呢？

LMDeploy support accelerating DeepSeek VL models now!!! :rocket:

DeepSeek-VL-chat-1.3B也可以支持量化么？

> > 参考LLAVA的方式构建自己的数据集，数据量14w，按照官方给的方式进行finetune，发现进入trainer.train()之后就hang住了，strout和strerr也没有信息，没有报错，请问是哪里不对吗 > > strout中的输出 > > ```python > > [2023-12-04 15:00:43,681] [WARNING] [partition_parameters.py:823:_post_init_method] param `cls_token` in MplugOwlVisionEmbeddings not on GPU so was not broadcasted from rank 0 >...

mPLUG-Owl2，finetune，训练卡住没有输出

解决了，我这边原因是数据集中图片路径的问题。路径改好了就OK

RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

手动改了一下device，可以跑通。 ```python def apply_rotary_pos_emb(q, k, cos, sin, position_ids): # print(q.device, k.device, cos.device, sin.device, position_ids.device) # cuda:0 cuda:0 cpu cpu cuda:0 # The first two dimensions of cos and sin are...