Skip to content

[K3.1-Pro] Map Kimi K3.1 Pro checkpoints on NVIDIA clusters - #4795

Merged
cquil11 merged 3 commits into
mainfrom
cam/kimik3.1-checkpoints
Nov 4, 2026
Merged

cquil11 merged 3 commits into
mainfrom
cam/kimik3.1-checkpoints

Conversation

@cquil11

@cquil11 cquil11 commented Nov 4, 2026 •

Copy link
Copy Markdown
Collaborator

Adds the Kimi-K3.1-Pro checkpoint mapping on every NVIDIA cluster where NVIDIA staged the weights, so moonshotai/Kimi-K3.1-Pro resolves to the pre-staged copy instead of a fresh download, and adds the model's row to MODELS.md.

Cluster Root Path Storage
b200-nscale scratch /scratch/models/Kimi-K3.1-Pro node-local NVMe
b300-dsxe scratch /scratch/models/Kimi-K3.1-Pro node-local NVMe
gb200-nv numa1 /mnt/numa1/models/Kimi-K3.1-Pro node-local NVMe
gb300-nv scratch /scratch/models/Kimi-K3.1-Pro node-local NVMe on racks 1 and 3 (the batch_1 and batch_3 partitions this cluster uses)
h100-dsxe scratch (new node-local volume) /scratch/models/Kimi-K3.1-Pro node-local RAID
h200-dgxc hf-hub-cache /models/gharunners/hf-hub-cache/Kimi-K3.1-Pro shared, the same root Kimi-K3 used

NVIDIA staged the weights per AIAMB-105. Each path was checked on compute nodes and holds the full 97-shard checkpoint.

No benchmark config uses the entries yet, so there is no perf-changelog.yaml entry.

Point b200-nscale, b300-dsxe and gb300-nv at the node-local
/scratch/models/Kimi-K3.1-Pro copy and gb200-nv at /mnt/numa1/models, and
add the model row to MODELS.md.

@claude claude Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

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

Code review found no issues

No high-confidence issues detected in this change.

This review covers commit a2e82f4, which is no longer the latest commit on this pull request; later commits are not covered by it.

h100-dsxe gains its node-local /scratch/models volume for the staged copy;
h200-dgxc reads the copy staged under its hf-hub-cache root.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

Status: Done

Development

Successfully merging this pull request may close these issues.

1 participant