PR #7192 ggml : full ALiBi support

ggml : full ALiBi support #7192

ggerganov merged 10 commits into master from gg/refactor-alibi-2

ggml : full ALiBi support

7fdca334

ggml : update ggml_soft_max_ext() CUDA, SYCL

d0592d49

ggerganov force pushed from 922a5b3d to d0592d49 1 year ago

ggml : ggml_flash_attn_ext() support ALiBi (CPU)

166e60bf

ggerganov force pushed from a4c7cf7e to 166e60bf 1 year ago

mofosyne added Review Complexity : High

mofosyne added enhancement

mofosyne added model

ggml : ggml_flash_attn_ext() support ALiBi (Metal)

97c27f59

ggerganov force pushed from ba4d12ab to 97c27f59 1 year ago

ggerganov commented on 2024-05-10

ggml : fix warning

f7055d31

JoanFM commented on 2024-05-10

ggml : ggml_flash_attn_ext() support ALiBi (CUDA)

865af990

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

ggerganov requested a review from

slaren 1 year ago

JoanFM requested changes on 2024-05-10

ggml : fix assert message

536983b1

JoanFM commented on 2024-05-10

vulkan : add dev notes

397b1f8f

NeoZhangJianyu approved these changes on 2024-05-10

slaren approved these changes on 2024-05-10

ggml : require mask when using ALiBi

0faf92e7

ggerganov force pushed from a6166051 to 0faf92e7 1 year ago

mofosyne added refactoring

convert : fix convert for refact models

03e940cd

ggerganov merged 9cb317f7 into master 1 year ago

Reviewers

slaren

NeoZhangJianyu

JoanFM

Assignees

No one assigned

Labels

enhancement model refactoring Review Complexity : High

Milestone

No milestone

llama.cpp ggml : full ALiBi support #7192 Merged

ggml : full ALiBi support #7192

llama.cpp
ggml : full ALiBi support
#7192

Merged