PR #7628 Vulkan Mixture of Experts (MoE) support

Vulkan Mixture of Experts (MoE) support #7628

0cc4m merged 8 commits into master from 0cc4m/vulkan-moe

Finish Vulkan mul_mat_id implementation

579f059a

Add Vulkan sum_rows and div ops

b4abdbb8

Fix MUL_MAT_ID matrix matrix shader

45928e8d

Merge remote-tracking branch 'origin/master' into 0cc4m/vulkan-moe

8ecdda1e

github-actions added Vulkan

github-actions added python

Fix MUL_MAT_ID matrix vector shader dispatch size

c8f93774

mofosyne added Review Complexity : High

Fix MUL_MAT_ID matrix vector shader and dispatch code

2c3d0b42

Update Vulkan CPU offload for MUL_MAT_ID

6e0e0beb

Fix crash when using split mode none and setting a main GPU

fe3f6958

slaren approved these changes on 2024-06-02

0cc4m merged 3d7ebf63 into master 1 year ago

0cc4m deleted the 0cc4m/vulkan-moe branch 1 year ago

Reviewers

slaren

Assignees

No one assigned

Labels

Vulkan python Review Complexity : High

Milestone

No milestone