PR #9438 IBM Granite MoE Architecture

IBM Granite MoE Architecture #9438

ggerganov merged 10 commits into ggml-org:master from gabe-l-hart:GraniteMoE

gabe-l-hart force pushed 1 year ago

github-actions added python

compilade commented on 2024-09-14

gabe-l-hart force pushed 1 year ago

gabe-l-hart marked this pull request as ready for review 1 year ago

gabe-l-hart force pushed 1 year ago

ggerganov approved these changes on 2024-09-23

compilade commented on 2024-09-23

gabe-l-hart force pushed 1 year ago

compilade commented on 2024-09-23

compilade approved these changes on 2024-09-23

feat(gguf-py): Add granitemoe architecture

8a4ca231

feat(convert_hf_to_gguf): Add GraniteMoeModel

e0b72290

fix(granitemoe convert): Split the double-sized input layer into gate…

014e59d3

feat(granitemoe): Implement granitemoe

eca37cd4

Typo fix in docstring

71bc4c1f

fix(conversion): Simplify tensor name mapping in conversion

5eb28c47

fix(convert): Remove unused tensor name mappings

f2360996

fix(convert): Sanity check on merged FFN tensor sizes

317b15bb

fix: Allow "output" layer in granite moe architecture (convert and cpp)

1c8b3e4c

gabe-l-hart force pushed to 1c8b3e4c 1 year ago

fix(granite): Add missing 'output' tensor for Granite

a843f1fa

compilade added merge ready

ggerganov merged 3d6bf691 into master 1 year ago

gabe-l-hart deleted the GraniteMoE branch 1 year ago

Reviewers

ggerganov

compilade

Assignees

No one assigned

Labels

python merge ready

Milestone

No milestone

llama.cpp IBM Granite MoE Architecture #9438 Merged

IBM Granite MoE Architecture #9438

llama.cpp
IBM Granite MoE Architecture
#9438

Merged