PR #21747 [js/webgpu] optimize MatmulNBits

[js/webgpu] optimize MatmulNBits #21747

guschmue merged 11 commits into microsoft:main from qjia7:opt_matmulnbits

opt matmulnbits

79a4ac15

add outputNumber > 1

6bd5417f

qjia7 force pushed from 8d5c18ea to bba62597 1 year ago

clean code

d28ac0a2

qjia7 force pushed from bba62597 to d28ac0a2 1 year ago

qjia7 changed the title ~~[WIP] Opt matmulnbits~~ Opt matmulnbits 1 year ago

qjia7 marked this pull request as ready for review 1 year ago

fs-eire changed the title ~~Opt matmulnbits~~ [js/webgpu] optimize MatmulNBits 1 year ago

formant and add missing shaderCache hints

dab4542d

use global_idx

e86150a2

tune outputNumber

e0072102

add limitations

2bf70efe

satyajandhyala added ep:WebGPU

Merge branch 'main' into opt_matmulnbits

f68a2da9

fix workgroupSize to reduce shader recompilation

cbacc4a4

support zeroPoints as input

cb775ed0

replace the old algorithm

2ab2c4ed

satyajandhyala commented on 2024-08-22

qjia7 commented on 2024-08-23

qjia7 requested a review from

satyajandhyala 1 year ago

satyajandhyala approved these changes on 2024-08-23

guschmue approved these changes on 2024-08-23

guschmue merged 87165b92 into main 1 year ago

Reviewers

guschmue

satyajandhyala

Assignees

No one assigned

Labels

ep:WebGPU

Milestone

No milestone

onnxruntime [js/webgpu] optimize MatmulNBits #21747 Merged

[js/webgpu] optimize MatmulNBits #21747

onnxruntime
[js/webgpu] optimize MatmulNBits
#21747

Merged