-
### 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.
- [ ] 3. Please note that if the bug-related iss…
-
### Motivation
![image](https://github.com/InternLM/lmdeploy/assets/12838445/3a263f1d-6a60-438e-9251-641de35b7d63)
为啥我采用 训练后量化精度下降的点数比较多
另外,只能节省现存,吞吐提高不明显,在A100 测试的结果
### Related resources
_No re…
-
命令行:
swift deploy --model_type yi-1_5-6b --model_id_or_path 01ai/Yi-1.5-6B-Chat --host 0.0.0.0 --port 8001
错误:
[INFO:swift] Start time of running main: 2024-05-19 09:44:09.371086
[INFO:swift] c…
-
### 📚 The doc issue
![issue](https://github.com/InternLM/lmdeploy/assets/120365110/e96b9d5f-9d1f-4e77-a0e7-c31c7e5c70c3)
AssertionError: 'internlm2_5-7b-chat’is not supported. The supported models…
-
### Prerequisites
- [X] I am running the latest code. Mention the version if possible as well.
- [X] I carefully followed the [README.md](https://github.com/ggerganov/llama.cpp/blob/master/README.md)…
-
您好,感谢SWIFT带来的方便应用。
我在微调InternLM-XComposer2d5时有一个疑惑,就是模型微调过程中的把高分辨率图像切块的hd_num参数在使用SWIFT微调时要如何指定呢?官网的源码里可以直接指定这个参数
-
1. 在单张4090上运行internlm/internlm2_5-7b-chat
2. 运行terninal程序,在加载模型,完成lmdeploy的日志打印后,程序无其他信息输出
3.
![微信截图_20240801110135](https://github.com/user-attachments/assets/b10e8d8b-fe3e-4770-8ba5-e43c23f00222)…
-
pack_to_max_length = True 可以正常运行,pack_to_max_length = False后卡死,log打到"Checkpoints will be saved to...”就卡住了,也不报错,请问是怎么回事啊?
-
I noticed that, for default (sequence classification) models with chat template defined in the tokenizer, `scripts/run_rm.py` formats each conversation by `tokenizer.apply_chat_template` (via the func…
-
Support finetuning LLaVA 1.6