llama.cpp
ggml-webgpu: fix compiler warnings and refactor FlashAttention encoding
#21052
Merged
Merged
ggml-webgpu: fix compiler warnings and refactor FlashAttention encoding #21052
reeselevine
merged
17 commits into
ggml-org:master
from
reeselevine:update-workflows
Update workflows to remove dependence on llvmpipe
48a587ca
Try setting Dawn_DIR
d6ab81fe
remove c++20 initializers
c095e1ec
Move to proper guid
d969d5eb
Try avoiding segfaults on vulkan backend process exit
519eada3
Remove compiler warnings on parameter casting
43685c75
Fix soft_max and update reg_tile accumulation to f32 for better preci…
53ffd67d
Merge remote-tracking branch 'upstream/master' into update-workflows
696da004
Refactor flash_attn a bit
c0aeda1a
remove c++20 initializers and format
ef724e8a
Merge remote-tracking branch 'upstream/master' into update-workflows
19737ea2
Increase div precision for NVIDIA
89b6283e
revert div precision and comment out ggml-ci node for now
c96ae259
Merge remote-tracking branch 'upstream/master' into update-workflows
fe744e03
Formatting
ec783c15
Try debugging on a failing CI node
1971e33c
Revert "Try debugging on a failing CI node"
aaff9d6e
reeselevine
changed the title
reeselevine
changed the title 
ggerganov
approved these changes
on 2026-04-17
CISC
approved these changes
on 2026-04-17
Login to write a write a comment.
Login via GitHub