Default Branch

master

09e9068007 · whisper.android : support benchmark for Android example. (#542) · Updated 1 year ago

Branches

timing

4f074fb7a8 · tmp : demonstrate how to measure time of ggml ops · Updated 1 year ago

0
1
musl

bf5d4c81b9 · make : fix MUSL Linux build · Updated 1 year ago

2
1
coreml

17a14593de · coreml : simlpify whisper_encode + log messages · Updated 1 year ago

8
2
4-bit

b4ebdb6b57 · bench : add Q4_0 and Q4_1 mul_mat benchmarks · Updated 1 year ago

13
5
guided

a0da7f71a2 · command : wip in progress, improve guided decoding · Updated 1 year ago

16
1
diarization

ec44ad0a75 · diarization : try conv and self-attention embeddings · Updated 1 year ago

17
4
chess

59c997ca2d · wip ignore · Updated 1 year ago

24
1
arghh

7aa1174315 · bench : fix Windows linkage by moving ggml benches in whisper lib .. · Updated 1 year ago

62
1
fa-decoder

e2aa556a99 · whisper : experiments with Flash Attention in the decoder · Updated 1 year ago

84
1
threads

4e6d2e98ab · ggml : try to improve threading · Updated 1 year ago

124
1
nvblas

683f111088 · ggml : initial tests with libnvblas · Updated 1 year ago

203
1
macros-cvt-fp16

e0bd97f41f · ggml : use macros to inline FP16 <-> FP32 conversions · Updated 1 year ago

207
1
stream

0a2621b637 · stream : add "max_tokens" cli arg · Updated 1 year ago

285
5
metal

fa9621e5e9 · mtl : update Makefile to support Metal · Updated 1 year ago

295
2
avx512

5d895d60b6 · Merge branch 'master' into avx512 · Updated 1 year ago

300
7
word-ts-2

210a6fb83c · wip : some unsuccessful experiments using DP · Updated 1 year ago

323
1
experiment/model-compression

4597c9c19b · wip : try to compress just mlp · Updated 2 years ago

437
2