-
# 问题
转换完权重之后进行评估验证时出现下述问题
```shell
> number of parameters on (tensor, pipeline) model parallel rank (0, 0): 630167424
loading release checkpoint from /raid/LLM_train/Pai-Megatron-Patch/checkpoint…
-
input长度在6k左右,对于同一个api_server,短的文本(2k)可以生成摘要,但是6k的则直接输出终止符。
相同的6k长度的输入,对于调用**qwen官方的hf**的代码,可以正常输出摘要。
调了很多生成参数,结果不变。
大家有遇到过相似的吗?
-
### Is there an existing issue for the same bug?
- [X] I have checked the existing issues.
### Branch name
main
### Commit ID
1234
### Other environment information
```Markdown
LLM : qwen-1.5 本…
-
当前选用的模型为glm4-chat,对话时报错,详细的错误为:
```
INFO: 127.0.0.1:62988 - "GET /tools HTTP/1.1" 200 OK
2024-07-16 16:39:32,138 httpx 21692 INFO HTTP Request: GET http://127.0.0.1:7861/tools "H…
-
Im not familiar with DynamicCache in huggingface transformers, but I can tell that it's not being passed properly during microbatching checks.
Here's my workaround that enabled Phi-2 and Qwen-1.5 0…
-
I've patched my Ollama to allow the setting of `rope_frequency_base` in the modelfile again, so I can fix this via:
```
PARAMETER rope_frequency_base 1000000
```
but it should also be possible…
-
### Is your feature request related to a problem? Please describe.
Currently, local models and local embeddings are not supported. When will they be supported?
### Describe the solution you'd like
…
-
### Reminder
- [X] I have read the README and searched the existing issues.
### Reproduction
### 我的修改
为保证Reward Model出来的分数是有意义的,并且是归一化到0-1的,训练过程中替换了自己的Loss函数
# loss += -torch.nn…
-
**Is your feature request related to a problem? Please describe.**
I have noticed that some people wish to experiment with open source LLMs (such as llama3 or qwen), but are limited by their computer…
-
请问怎么解决呢