transformers
Fix Mamba2 chunked-prefill / speculative decoding for Zamba2, Nemotron-H, Bamba, FalconH1 and GraniteMoeHybrid
#46741
Merged
Overview
Commits
9
Changes
View On GitHub

Fix Mamba2 chunked-prefill / speculative decoding for Zamba2, Nemotron-H, Bamba, FalconH1 and GraniteMoeHybrid #46741

vasqu merged 9 commits into huggingface:main from Sunt-ing:c001-mamba2-chunked-prefill
Sunt-ing
Sunt-ing Fix Mamba2 mixer chunked-prefill/speculative decode for Zamba2 and Ne…
2c8986fe
Ramshankar07
Sunt-ing Prepend the full cached conv_states in the Mamba2 chunked path
1aaaeb62
Sunt-ing
vasqu
vasqu commented on 2026-06-22
Sunt-ing Fix Mamba2 cuda_kernels_forward chunked-prefill for Zamba2 and Nemotr…
60a85f8f
Sunt-ing
vasqu
Sunt-ing Align Zamba2/Nemotron-H chunked-prefill tests with merged mamba2 test
3514e781
Sunt-ing Propagate Mamba2 chunked-prefill cache fix to Bamba, FalconH1, Granit…
91b8f8ce
Sunt-ing Sunt-ing changed the title Fix Mamba2 mixer chunked-prefill / speculative decoding for Zamba2 and Nemotron-H Fix Mamba2 chunked-prefill / speculative decoding for Zamba2, Nemotron-H, Bamba, FalconH1 and GraniteMoeHybrid 99 days ago
Sunt-ing
vasqu
vasqu commented on 2026-06-26
vasqu
Sunt-ing Extract cached conv/recurrent state into locals in the Mamba2 mixers
a0320f55
Sunt-ing Add CPU variant for the nemotron_h chunked-prefill test
e351f092
Sunt-ing
vasqu Merge branch 'main' into c001-mamba2-chunked-prefill
14f547f6
vasqu rename a bit
7bd3cada
vasqu
vasqu approved these changes on 2026-06-29
vasqu
github-actions
github-actions
HuggingFaceDocBuilderDev
github-actions
vasqu vasqu merged fdb6d318 into main 96 days ago

Login to write a write a comment.

Login via GitHub

Reviewers
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Terms Of Use Privacy Policy Refund Policy Impressum