-
我查阅了几个LLM部署推理框架,目前似乎都不支持InternVL,请问项目组有推荐使用的框架吗?
-
LLaVA的MME指标是不是不太对,我自己测出来的llava-internlm2-7b是1407
![image](https://github.com/InternLM/xtuner/assets/34935911/a23460d7-2732-4e5f-8bde-76175a153c68)
![image](https://github.com/InternLM/xtuner/assets/…
-
如题。。如果pretrain就把图片切那么多份,训练成本是不是有些cover不住
-
### Checklist
- [X] 1. I have searched related issues but cannot get the expected help.
- [X] 2. The bug has not been fixed in the latest version.
### Describe the bug
1.session length长度不一致,…
-
### Checklist
- [ ] 1. I have searched related issues but cannot get the expected help.
- [X] 2. The bug has not been fixed in the latest version.
### Describe the bug
Internvl2 api 使用没法正常返回结果,用tra…
-
### Reminder
- [X] I have read the README and searched the existing issues.
### Reproduction
同样的模型,同样的超参,同样的数据,为什么sft结果不一致,且loss有细微差别。训练参数如下:
cuda=0,1,2,3
stage=sft
model_path=internlm2-chat-7b
…
-
### Motivation
Hello,
There is a client program which calls different types of opanai models, like "gpt-3.5-turbo" and "gpt-4-turbo". Now I want to use local llm instead of openai models and I *…
-
使用modelscope上下载的llava-internlm2-7b推理MMBench_DEV_EN
gpu:4*V100 32g
耗时约12小时,请问是否正常?
推理脚本:
```
export CUDA_VISIBLE_DEVICES=0,1,2,3
model_path=./xtuner/internlm/internlm2-chat-7b
ve_path=./xtuner/o…
-
### Describe the bug
使用最新的llama.cpp代码(b1874),转化模型时报错:
python3 convert.py ../internlm2-chat-20b --outtype f16
/Users/pom/AIGC/llama.cpp-3/gguf-py
Loading model file ../internlm2-chat-20b/pytorch_mo…
gaord updated
7 months ago
-
Add dll path C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.4\bin, please note cuda version should >= 11.3 when compiled with cuda 11
Special tokens have been added in the vocabulary, make su…