-
Notifications
You must be signed in to change notification settings - Fork 392
update "rollout_max_batch_size" to replace "max_concurrent" for user settings #1225
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
Conversation
7a90af8 to
f0bc0f8
Compare
tests/ray/test_rollout.py
Outdated
| ] | ||
| self.dataloader_cfg = DataloaderConfig( | ||
| pack_max_length=self.max_prompt_length, | ||
| pack_max_length=self.max_prompt_length+self.max_response_length, |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
删掉
2890ff6 to
e7b7f5b
Compare
|
rollout worker的http_concurrency需要能够大于rollout_max_batch_size,现在还是根据rollout_max_batch_size算的 |
max_concurrent和http_concurrency可以绑定,这两个参数是客户端的请求数,而rollout_max_batch_size是推理引擎端的参数,这二者需要分开配置,否则GPU功率打不满 |
xtuner/v1/ray/config/worker.py
Outdated
| api_key (Optional[Union[List[str], str]]): API keys for rollout service. | ||
| Supports single key or list of keys. Defaults to None. | ||
| api_key (Optional[Union[List[str], str]]): API keys for rollout service. åSupports single key or list of keys. Defaults to None. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
fix typo before supports
done, 通过 ttp_concurrency = config.rollout_max_batch_size_per_instance * config.allow_over_concurrency_ratio 来配置 |
用户配置并发度的方式为以rollout_max_batch_per_instance为基准,从而配置数据流等并发度。配置文档见 xtuner/docs/zh_cn/rl/advanced_tutorial/efficiency.md
可能的几种情况如下: