Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[bug] OpenFunctions-v2: <Issue>CUDA_VISIBLE_DEVICES=1 bfcl evaluate --model Qwen/Qwen2.5-7B-Instruct --test-category simple 执行这条命令,score目录生成的指标文件没有具体的数据。 #784

Closed
belief888 opened this issue Nov 25, 2024 · 17 comments
Labels
BFCL-General General BFCL Issue

Comments

@belief888
Copy link

执行CUDA_VISIBLE_DEVICES=1 bfcl evaluate --model Qwen/Qwen2.5-7B-Instruct --test-category simple 命令,目录:/bfcl/gorilla/berkeley-function-call-leaderboard/score/ 下的data_overall.csv里面没有详细的指标数据,请问这是啥原因造成的?谢谢!

@belief888 belief888 added the hosted-openfunctions-v2 Issues with OpenFunctions-v2 label Nov 25, 2024
@HuanzhiMao HuanzhiMao added BFCL-General General BFCL Issue and removed hosted-openfunctions-v2 Issues with OpenFunctions-v2 labels Nov 25, 2024
@HuanzhiMao
Copy link
Collaborator

HuanzhiMao commented Nov 25, 2024

You need to run bfcl generate xxx to obtain the model result first, before running the evaluation command. See here.

你需要先跑bfcl generate xxx拿到模型的输出,然后才能跑bfcl evaluate xxx

@HuanzhiMao
Copy link
Collaborator

HuanzhiMao commented Nov 25, 2024

score那个文件夹下有东西吗?应该有BFCL_v3_simple_score.json

@belief888
Copy link
Author

score目录下有data_live.csv,data_non_live.csv,data_multi_turn.csv,data_overall.csv这四个文件,没有BFCL_v3_simple_score.json

@belief888
Copy link
Author

BFCL_v3_simple_result.json 这个文件在另外一个文件夹result下

@HuanzhiMao
Copy link
Collaborator

我看你最一开始是用的 --model Qwen/Qwen2.5-7B-Instruct,为啥后来要换成--model /data/models/Qwen2___5-7B-Instruct

@belief888
Copy link
Author

我本地电脑的模型是从huggingface下载来的,存入在/data/models/Qwen2___5-7B-Instruct目录下,想直接加载就行了,不想再下载

@belief888
Copy link
Author

请问这个是啥回事?

@HuanzhiMao
Copy link
Collaborator

HuanzhiMao commented Nov 25, 2024

我本地电脑的模型是从huggingface下载来的,存入在/data/models/Qwen2___5-7B-Instruct目录下,想直接加载就行了,不想再下载

你应该改huggingface那边的路径,export HF_HOME=xxx--model这个argument不管你本地存储的路径,它是用来决定使用哪个model handler的

@HuanzhiMao
Copy link
Collaborator

不是,export HF_HOME=xxx是输在terminal里的...
--model Qwen/Qwen2.5-7B-Instruct,然后model_metadata那边你应该是不用改的

@belief888
Copy link
Author

export HF_HOME=/data/models/Qwen2___5-7B-Instruct 输在terminal ?

@HuanzhiMao
Copy link
Collaborator

terminal里应该输的是export HF_HOME=/data/models

@belief888
Copy link
Author

gorilla-openfunctions-v2 这个模型我己经下载好在/data/gcr/models里面了

@HuanzhiMao
Copy link
Collaborator

这个目前没法解决。gorilla-openfunctions-v2的hosted API endpoint最近有点问题;如果要换成local inference的话需要写新的model handler。可以考虑换个模型

@belief888
Copy link
Author

我现在想评估eval一下gorilla-openfunctions-v2这个模型的指标,看输出结果是啥样的?

@belief888
Copy link
Author

请问有好的解决方案吗?

@HuanzhiMao
Copy link
Collaborator

我可以发给你我们在 #748 里面跑出来的gorilla-openfunctions-v2的输出结果(不是最新的commit)。如果你想在你那边自己跑的话,现在解决不了(API掉线了)

@HuanzhiMao
Copy link
Collaborator

HTTPSConnectionPool(host='cdn-lfs-us-1.hf-mirror.com', port=443): Read timed out. 这个说的是你网络的问题吧?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
BFCL-General General BFCL Issue
Projects
None yet
Development

No branches or pull requests

2 participants