8000 [Parallelization] - How to compile and run on multiple gpus? · Issue #361 · mirage-project/mirage · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
[Parallelization] - How to compile and run on multiple gpus? #361
Open
@zhaotf16

Description

@zhaotf16

Description

Hi, mirage team!
Awesome work! I am already done with your Qwen3 demo. Now I'm on parallel inference. Could u tell me how I can load my compiled mirage kernel to multiple GPUs? Is it compatible with torch.distributed? And what should I, or what will u do if a model is too large to compile into a mega-kernel if my cuda memory is limited.

Metadata

Metadata

Assignees

Labels

questionFurther information is requested

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions

    0