Skip to content

DeepSeek V3/R1 inference recipe on Trillium with JetStream + MaxText + Pathways #55

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

Merged
merged 8 commits into from
May 2, 2025

Conversation

RajeshThallam
Copy link
Contributor

This is DeepSeek V3/R1 inference recipe on Trillium with JetStream + MaxText + Pathways

@RajeshThallam RajeshThallam marked this pull request as ready for review May 2, 2025 20:41
Copy link
Collaborator

@mitalisi mitalisi left a comment

Choose a reason for hiding this comment

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

One nit - can we rename the folder as DeepSeek-R1-671B instead of DeepSeek-V3-671B.

@RajeshThallam
Copy link
Contributor Author

One nit - can we rename the folder as DeepSeek-R1-671B instead of DeepSeek-V3-671B.

Done!

@vipannalla vipannalla merged commit 52b69ee into AI-Hypercomputer:main May 2, 2025
1 check passed
gcloud compute networks create ${NETWORK_NAME_1} --mtu=8896 --project=${PROJECT_ID} --subnet-mode=auto --bgp-routing-mode=regional
gcloud compute firewall-rules create ${NETWORK_FW_NAME_1} --network ${NETWORK_NAME_1} --allow tcp,icmp,udp --project=${PROJECT_ID}

# Secondary subnet for multinic experience. Need custom ip routing to be different from first network’s subnet.
Copy link
Collaborator

Choose a reason for hiding this comment

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

The secondary network seems to be only needed for multislice / multi-nic setups. However the cluster created here uses NUM_SLICES=1. So this can be skipped?

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

Successfully merging this pull request may close these issues.

4 participants