whisper : add full CUDA and Metal offloading #1472
ggerganov
merged
21 commits into
master
from
ggml-backend-no-sched
whisper : migrate to ggml-backend
65975732
whisper : fix logit reading
7e01486b
whisper : fix tensor allocation during load
3dfbe649
whisper : fix beam-search with CUDA
dcf9511d
whisper : free backends + fix compile warning
12030358
whisper : print when CUDA is enabled
3f5c1b7e
Merge branch 'master' into ggml-backend-no-sched
0ab50253
whisper : fix CoreML
a54d8c9d
make : clean-up
d6dad64f
Merge branch 'master' into ggml-backend-no-sched
728e1785
talk : fix compile warning
c99e290a
whisper : support ggml_conv with CUDA and Metal (#1473)
933c5bef
whisper : clean-up
f53e1388
ggerganov
changed the title 
quantize-all : fix
3bfc43e3
ggml : im2col opts
66bb2e94
whisper : avoid whisper_model_data wrapper
0867e696
whisper : add note that ggml_mul_mat_pad does not work with CUDA
b27726da
ggerganov
commented
on 2023-11-11
whisper : factor out graph compute in common function
b6182293
whisper : fixes
fc8565d0
whisper : fix UB with measure buffers
40c66036
whisper : try to fix the parallel whisper_state functionality (#1479)
5031f547
Reviewers
No reviews
Assignees
No one assigned
Labels
None yet
Milestone
No milestone
Login to write a write a comment.
Login via GitHub