ggerganov/whisper.cpp

Pull Requests Commits

alloc : fix allocation data of pre-allocated leafs

slaren committed 2 years ago

66df44b0

whisper : allocate encoder results in dedicated buffer

ggerganov committed 2 years ago

Verified c6c94de4

talk-llama : sync llama.cpp

ggerganov committed 2 years ago

Verified de4d067f

ggerganov committed 2 years ago

Verified e715f6a6

update examples and tests

slaren committed 2 years ago

Verified f60ccfd8

ggml : add ggml-common.h

ggerganov committed 2 years ago

Verified 3753a2b2

ggml : designate enum vals for integer types (llama/6050)

ggerganov committed 2 years ago

Verified 592dd256

metal : build metallib + fix embed path (llama/6015)

ggerganov committed 2 years ago

Verified c8709d46

llama : add pipeline parallelism support (llama/6017)

slaren committed 2 years ago

Verified 8932c2d6

Update get version (llama/6025)

AidanBeltonS committed 2 years ago

Verified 2bddfdd7

ggml : reuse quantum structs across backends (llama/5943)

ggerganov committed 2 years ago

Verified 46e3c3f1

ggml : fix UB in IQ2_S and IQ3_S (llama/6012)

ggerganov committed 2 years ago

Verified ef24ae0c

sycl : update IQ1_S kernels (WIP - not working!) (llama/5995)

ggerganov committed 2 years ago

Verified a753926f

1.5 bit: we can do even better (llama/5999)

ikawrakow committed 2 years ago

Verified 9dc60fc0

ggml, ci : Windows ARM runner and build fixes (llama/5979)

Xarbirus committed 2 years ago

Verified d73a6362

Better 1.5 bit quantization (llama/5971)

ikawrakow committed 2 years ago

Verified f79d0d4f

Add q3_s and q1_s (llama/5886)

abhilash1910 committed 2 years ago

Verified 4f88940f

metal : move mm_id indices to shared mem (llama/5982)

ggerganov committed 2 years ago

Verified 7bdb1de9

ggml : fix unnecessary f32 -> f16 -> f32 casts (mmla) (llama/5951)

ggerganov committed 2 years ago

Verified 653d2e8f

ggml : remove old quantization functions (llama/5942)

ggerganov committed 2 years ago

Verified 2fef660d

ggml : add ggml-common.h to deduplicate shared code (llama/5940)

ggerganov committed 2 years ago

Verified 24eba5a2

llama : support Mamba Selective State Space Models (llama/5328)

compilade committed 2 years ago

Verified 6e9d3aa3

extra : update sync scripts after ggml-common.h

ggerganov committed 2 years ago

Verified 9ae0d188

whisper : document whisper_batch.n_seq_id (#1942)

josharian committed 2 years ago

Verified a56f435f

whisper : improve beam search candidate diversity (#1947)

josharian committed 2 years ago

Verified ec166499

bindings/go : add linker flags to make metal work (#1944)

josharian committed 2 years ago

Verified ccf022f9

whisper : make beam candidate sort more stable (#1943)

josharian committed 2 years ago

Verified 2852e1af

ggml : try fix 32-bit arm compat (#1938)

ggerganov committed 2 years ago

Verified ce945b50

talk-llama : use llama_decode instead of llama_eval

ggerganov committed 2 years ago

Verified 2f5a5a66

talk-llama : sync llama.cpp

ggerganov committed 2 years ago

Verified 8e409d11

Older