llama.cpp
[CANN]: add the basic supports of Flash Attention kernel
#13627

Merged

[CANN]: add the basic supports of Flash Attention kernel #13627

hipudding merged 18 commits into ggml-org:master from shibizhao:flash-attn-cann

cann: add the basic FA support

72df31df

cann: update the readme

3a731825

cann: update the FlashAttention with PSEShift

6a39d638

cann: update the input parameters in FA

8a902b98

cann: update the alibi with max_bias

f5e24a5c

cann: add the constrints of softcap

c8c2908b

cann: update the docs CANN.md

47f2c646

cann: update the docs CANN.md

fb62f015

github-actions added documentation

github-actions added ggml

shibizhao changed the title ~~cann: add the basic supports of Flash Attention kernel~~ [CANN]: add the basic supports of Flash Attention kernel 218 days ago

hipudding requested a review from

hipudding 218 days ago

hipudding added Ascend NPU

noemotiovon commented on 2025-05-21

cann: fix typo of CANN.md

b266beb2

cann: add some comments and update the CANN.md

8a112f0a

cann: update the CANN.md

1779e008

noemotiovon commented on 2025-05-21

cann: update the inner precise for fusedInferAttention

092ccf68

noemotiovon commented on 2025-05-22

cann: update the constraints of flash_attn_ext on ggml-cann.cpp

c380305b

cann: resolve the conflict with laster master branch

89f884e6

Merge branch 'master' into flash-attn-cann

1a3bfecb

cann: clean the whitespace

3b084d5b

cann: clean the whitespace

d23697b8

cann: add a new endline

8a7829b7

hipudding approved these changes on 2025-05-26

hipudding merged 2d38b6e4 into master 212 days ago

shibizhao deleted the flash-attn-cann branch 212 days ago

Reviewers

hipudding

noemotiovon

Assignees

No one assigned

Labels

documentation ggml Ascend NPU

Milestone

No milestone

llama.cpp [CANN]: add the basic supports of Flash Attention kernel #13627 Merged

[CANN]: add the basic supports of Flash Attention kernel #13627

llama.cpp
[CANN]: add the basic supports of Flash Attention kernel
#13627

Merged