Spaces:
Running
Running
Commit History
whisper : calculate mel spectrogram directly into a ggml_tensor (#2208) 521186a unverified
whisper : add CUDA-specific computation mel spectrograms (#2206) c6894d3 unverified
whisper : remove `speed_up` and `phase_vocoder*` functions (#2198) 7ef0c95 unverified
readme : add conan badge (#2196) f08dc65 unverified
Martin Delille commited on
readme : add install instructions for Conan (#2189) fb4f721 unverified
Carlos Zoido commited on
whisper: use global cache for sin/cos vals and Hann window (#2194) 3a04f56 unverified
release : v1.6.2 3e54141 unverified
Revert "whisper : remove extra backend instance (huh?)" (#2182) b708d81 unverified
server : fix typo (#2181) 18c60fc unverified
Daniel Valdivia commited on
ruby : update bindings (#2154) a2bce18 unverified
Todd commited on
release : v1.6.1 ca6f4b2 unverified
examples : add support for decoding input with ffmpeg (Linux) (#2133) c160b58 unverified
William Tambellini commited on
node : add flash_attn param (#2170) b4d05df unverified
ci: Update build.yml to suppress warnings about node.js versions (#2166) e9954d9 unverified
Tamotsu Takahashi commited on
release : v1.6.0 d823237 unverified
whisper : use flash attention (#2152) 27c0a97 unverified
talk-llama : reject runs without required arguments (#2153) b445508 unverified
sync : ggml aac57a1 unverified
metal : support FA without mask + add asserts (llama/7278) 98ce302 unverified
ggml : add RPC backend (llama/6829) 5838a14 unverified
rm wait() (llama/7233) 328702a unverified
Neo Zhang commited on
CUDA: add FP32 FlashAttention vector kernel (llama/7188) 03d4b22 unverified
scripts : sync ggml-rpc 7b58c58 unverified
whisper : fix model path encoding in windows (#2086) 49f8792 unverified
thewh1teagle commited on
server : return utf-8 (#2138) 2719aa0 unverified
cmake : fix HIP/ROCm build (#2102) a90ae59 unverified
aldorof commited on
node : add additional params (#2000) 933eb40 unverified
valVk commited on
js : remove un-needed request header from fetchRemote (#2119) 6c54394 unverified
Mark Karpelès commited on
cmake : fix metal embed sources path (#2110) 087b1a8 unverified
main : dont print timings with --no-prints (#2108) 685d1c1 unverified
Daniel Ziegenberg commited on
main : add options for temperature control (#2088) 9a3f777 unverified
Daniel Ziegenberg commited on
whisper : switch back to F32 mask (#0) 3b7b90c unverified
whisper.android : update example, add field to print timestamp (#2072) 03fb680 unverified
cmake : fix json INTERFACE library (#2069) 0a1cadb unverified
main : fix double quote escaping in csv output (#2090) 9952a85 unverified
mashizora commited on
metal : tune soft_max number of threads (#0) 99d668a
whisper : remove old flash attn code (#0) fd57e47
ggml : try fix ppc64 (#0) df78c25
ggml : remove oboslete alibi code (skipme) (#0) d25c1e3
talk-llama : sync llama.cpp f5f68d6
sync : ggml 3ea4549
ggml : optimize for ppc64le using VSX intrinsics (ggml/784) 05d3824
metal : fix indent (ggml/0) d4f82d5
ggml : restore sigmoid decl order (ggml/0) 67c5387
ggml : resolve merge (ggml/0) d692b06
ggml : full ALiBi support (llama/7192) 192bda4
metal : fix flash attention kernel requirements (llama/7169) 6cb3028
Minor arithmetic improvement to mmvq wrapper kernel (llama/7172) ae75124
Ouadie EL FAROUKI commited on