Skip to content

support loading model with user input params (turbomind)#3204

Closed
irexyc wants to merge 5 commits intoInternLM:mainfrom
irexyc:pass-model-params
Closed

support loading model with user input params (turbomind)#3204
irexyc wants to merge 5 commits intoInternLM:mainfrom
irexyc:pass-model-params

Conversation

@irexyc
Copy link
Copy Markdown
Collaborator

@irexyc irexyc commented Mar 3, 2025

Motivation

Motivated by #3146

A usage demo

Comment thread lmdeploy/messages.py Outdated
num_tokens_per_iter: int = 0
max_prefill_iters: int = 1
communicator: str = 'nccl'
model_params_que: Any = None
Copy link
Copy Markdown
Collaborator

@lvhan028 lvhan028 Mar 4, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I agree with using a queue as a pipeline for data transmission, but I do not agree with making this parameter an engine parameter, as the two are completely unrelated
Can it be an argument in AsyncEngine.init?
@lzhangzz any suggestions?

@lvhan028 lvhan028 added the enhancement New feature or request label Mar 9, 2025
@irexyc irexyc closed this Jul 7, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

enhancement New feature or request

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants