red Zhao

sorry,I will repaste the error.

> The title shows `inference error while serving MiniCPM-V-2.6` However, in the content, the model is `ogvlm2-llama3-chinese-chat-19B` > > So, which one is correct? I have edited the content of...

> The `--debug` flag will result in the disabling of parallel execution. It is advisable to attempt removing this flag. SO important info......

下载数据集参考此issue [leval数据集问题](https://github.com/open-compass/opencompass/issues/1825)

How do you deploy the api server [http://10.24.9.4:8000/v1]? Can any-curl-command get the correct answer?

试试参照这个issue: [1590](https://github.com/open-compass/opencompass/issues/1590)，不过我用的是旧版本的compass，我测试仍然是可以执行的 ```python from mmengine.config import read_base with read_base(): # from opencompass.configs.models.qwen.hf_qwen2_1_5b_instruct import models # from opencompass.configs.models.qwen2_5.lmdeploy_qwen2_5_1_5b_instruct import models as lmdeploy_qwen2_5_1_5b_instruct # from opencompass.configs.models.qwen2_5.vllm_qwen2_5_1_5b_instruct import models as vllm_qwen2_5_1_5b_instruct # from...

接上条，目前两种评估方式 ```bash python run.py configs/eval_flame_demo.py \ -w outputs/demo \ --debug ``` 会触发 opencompass/tasks/subjective_eval.py 中的 _score() 方法，这里面后处理得到issue报错里的 'judgements'，所以流程是成功的 另一种评估方式，貌似在被**弃用**，即 ```bash python run.py \ --hf-type base --hf-path /space/llms/Qwen2-1.5B-Instruct \ --datasets flames_gen \...

> 你好，请问如何测评本地的微调大模型。我在相关文档里没有找了类似的教程 和评测本地开源模型一样，替换模型文件路径就行了

> ### Prerequisite > * [x] I have searched [Issues](https://github.com/open-compass/opencompass/issues/) and [Discussions](https://github.com/open-compass/opencompass/discussions) but cannot get the expected help. > * [x] The bug has not been fixed in the [latest...

> 你好， qwen系列的模型评估生成的内容全部是感叹号，目前使用了官方的Qwen2-7B-Instruct、Qwen2.5-7B-Instruct，生成全部是感叹号，能看一下是什么原因么？  而llama3-8B、glm-4-9b、Yi-1.5-9B可以正常运行 我这边用这些模型都是可以的，可能是你的评估数据集存在bug，你多用几种不同数据集确实是否一样，另外，还是一样的话，同样建议你用vllm/lmdeploy来替换默认的推理引擎，一方面可以加速，一方面兼容性会好点