-
Notifications
You must be signed in to change notification settings - Fork 1k
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
feat: Parallelise Model Loading #360
Conversation
Almost what I envisioned - only thing I would change is to preload after the preemptive download. We don't want to download all possible model shards, only the relevant one. |
@AlexCheema I think I got it, can you review the changes ?? |
tested on a m3 pro |
exo/main.py
Outdated
current_shard = preemptively_start_download(str(uuid.uuid4()), json.dumps({ | ||
"type": "node_status", | ||
"status": "start_process_prompt", | ||
"shard": shard.to_dict() | ||
})) |
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.
doesn't this need to be awaited?
exo/main.py
Outdated
"shard": shard.to_dict() | ||
})) | ||
if current_shard: | ||
await node.preload_models([current_shard]) |
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.
doesn't preemptively_start_download
already call preload_models
?
closed this one because it has been >month for the last pr, synced my fork and the new pr is at #466 |
test using