llama.cpp
ggml : group all experts in a single ggml_mul_mat_id
#6505
Merged

ggml : group all experts in a single ggml_mul_mat_id #6505

slaren merged 22 commits into master from sl/moe-rework-2
slaren
slaren ggml : group all experts in a single ggml_mul_mat_id
ea2b7953
github-actions
askmyteapot
askmyteapot commented on 2024-04-06
askmyteapot
askmyteapot commented on 2024-04-06
askmyteapot
Dampfinchen
askmyteapot
askmyteapot
askmyteapot
JohannesGaessler
JohannesGaessler commented on 2024-04-07
slaren minor
1b5d78d3
slaren
slaren fix windows build
bc615548
slaren refactor moe ffn to llm_build_moe_ffn
f3f7627b
JohannesGaessler
slaren cleanup
23f7d71a
askmyteapot
slaren update imatrix
9a43e808
slaren
ggerganov
slaren minor
47c3867b
slaren Merge remote-tracking branch 'origin/master' into sl/moe-rework-2
137fbb8f
LostRuins
slaren
slaren add metal impl
fc363e4a
slaren Merge remote-tracking branch 'origin/master' into sl/moe-rework-2
42003fdc
slaren fix merge
fb168ac5
slaren cleanup
bf56fdec
slaren cuda : fix bin bcast with non-cont src0
d68c935c
slaren cleanup
997a9b5b
slaren slaren marked this pull request as ready for review 1 year ago
slaren slaren requested a review from ggerganov ggerganov 1 year ago
slaren cuda : fix binbcast
f7fe79a3
ggerganov
slaren Merge remote-tracking branch 'origin/master' into sl/moe-rework-2
d18b19c8
slaren cuda : fix warnings
0e6963da
slaren metal : enable buffer log prints again
4d8fe076
ggerganov
ggerganov approved these changes on 2024-04-18
ggerganov llama : simplify moe reshapes
2080a97c
slaren ggml-ci
4980e350
slaren
slaren
ggerganov
NeoZhangJianyu
airMeng
slaren
slaren
ggerganov
slaren test-backend-ops : only run all mul mat tests for base types
bd17f27c
slaren llama : disable moe offloading with SYCL
ba5b5467
slaren slaren merged 0d56246f into master 1 year ago
slaren slaren deleted the sl/moe-rework-2 branch 1 year ago
slaren

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone