transformers
POC: Experts offload
#50717
Open

POC: Experts offload #50717

tarekziade wants to merge 5 commits into main from experts-offload
tarekziade
tarekziade Add experts_offload: stream MoE experts from disk through a bounded d…
412192f6
tarekziade tweaks
cc6b2e28
tarekziade tarekziade marked this pull request as draft 2 days ago
HuggingFaceDocBuilderDev
tarekziade experts_offload: faster reads, bounded prefill memory, MetalConfig fo…
ac255713
tarekziade experts_offload: commit queued MPS work early; install only finished …
4cacb570
tarekziade experts_offload: optional host tier of quantized experts (host_cache_…
2dd2dd68

Login to write a write a comment.

Login via GitHub

Reviewers
No reviews
Assignees
No one assigned
Labels
Milestone