Commit Graph

  • e2aa556a99
    whisper : experiments with Flash Attention in the decoder fa-decoder Georgi Gerganov 2022-12-16 21:51:35 +0200
  • f30b5d322c
    ggml : fix bug in new soft max computation Georgi Gerganov 2023-01-07 21:00:07 +0200
  • 44efbf7ff1
    cmake : add -Wno-unused-function + update whisper.js Georgi Gerganov 2023-01-07 20:18:34 +0200
  • d347a59a5f
    ggml : when using BLAS start only 1 CPU thread Georgi Gerganov 2023-01-07 19:48:56 +0200
  • 6394c906af
    ggml : fix running tasks with variable number of threads Georgi Gerganov 2023-01-07 19:20:18 +0200
  • 74ffa14e1d
    ggml : unroll ggml_vec_dot_f16 in ggml_compute_forward_flash_attn_f16 Georgi Gerganov 2023-01-07 17:32:23 +0200
  • 65fdcbbbbb
    whisper : revert accidental MB change Georgi Gerganov 2023-01-07 16:18:21 +0200
  • d61d55cd4b
    ggml : speed-up soft max via Accelerate + unroll Georgi Gerganov 2023-01-07 16:11:41 +0200
  • d51fc3ee0a
    ggml : use vDSP_sve and vDSP_maxv from Accelerate Georgi Gerganov 2023-01-07 16:10:16 +0200
  • f82a7dd019
    ggml : make gcc happy (minor) Georgi Gerganov 2023-01-07 09:34:39 +0200
  • 87dd4a3081
    talk.wasm : bump memory usage + update whisper.js Georgi Gerganov 2023-01-06 21:12:51 +0200
  • 41e05c6b1b
    cmake : support AVX2 in Windows better (#381) m.bell 2023-01-06 12:36:33 -0500
  • fa379cb22a
    Revert "tmp" Georgi Gerganov 2023-01-06 19:32:53 +0200
  • 322f4e6c4e
    go : bindings updated so they can be used in third party packages. (#379) David Thorpe 2023-01-06 17:32:28 +0000
  • 1652965529
    tmp Georgi Gerganov 2023-01-06 19:32:12 +0200
  • 6042c7a3be
    cmake : change min required version to 3.0 (#351) Georgi Gerganov 2023-01-06 19:23:56 +0200
  • 6b351bb669
    command : add "guided-mode" video demo in the README.md Georgi Gerganov 2023-01-06 18:59:26 +0200
  • a62170c656
    ggml : add SSE3 and fp16 conversion lookup table (#368) Abitofevrything 2023-01-06 17:45:59 +0100
  • 1944e7c33e whisper : document POWER VSX support Thomas Fitzsimmons 2023-01-04 23:00:30 -0500
  • 49a8dd6732 ggml : reorganize POWER9 ppc64le SIMD code Thomas Fitzsimmons 2023-01-03 00:45:43 -0500
  • 8c7f642286 ggml : change f16 load and store macro arguments Thomas Fitzsimmons 2023-01-03 00:43:17 -0500
  • ad2a4ffa03
    whisper : do not use F16 tensors when in F32 mode (#369) Georgi Gerganov 2023-01-05 22:56:25 +0200
  • b3c865083e
    ci : add emscripten build Georgi Gerganov 2023-01-05 22:02:58 +0200
  • a0d4f8e65c
    main : make whisper_print_segment_callback() more readable (close #371) Georgi Gerganov 2023-01-05 21:45:05 +0200
  • 4a214d2f07
    cmake : add CMAKE_RUNTIME_OUTPUT_DIRECTORY Georgi Gerganov 2023-01-05 21:40:59 +0200
  • 0a0cfa7985
    ggml : add void to argument-less functions Georgi Gerganov 2023-01-05 21:40:38 +0200
  • 196d738974
    minor : close #370 + Makefile build info print change Georgi Gerganov 2023-01-05 21:35:04 +0200
  • 84c6b42e65
    cmake : update to 3.19 (#351) Andy Maloney 2023-01-05 14:22:48 -0500
  • dd6d582977 whisper : use ranged-based for loops for readability Andy Maloney 2022-12-31 17:58:30 -0500
  • d51c5eb906
    ggml : define MIN / MAX only if not defined (minor) Georgi Gerganov 2023-01-05 21:07:50 +0200
  • 0be6a1afd9
    make : print build information Georgi Gerganov 2023-01-02 13:35:26 +0200
  • a466c3404d
    stream : fix data race on bool + avoid division-by-zero Georgi Gerganov 2023-01-02 10:20:50 +0200
  • d629c034a4
    models : fix HF model URL (close #356) Georgi Gerganov 2023-01-02 09:54:43 +0200
  • f00509d57c
    command : refactor to split command list & general transcription modes (#331) Andy Maloney 2022-12-31 07:08:57 -0500
  • 424c410c42 ggml : improve f16 acceleration for POWER9 ppc64le Thomas Fitzsimmons 2022-12-30 18:57:15 -0500
  • d97e6005e9
    whisper : add whisper_n_audio_ctx and check for invalid audio_ctx Georgi Gerganov 2022-12-31 09:55:33 +0200
  • 3467230a77 models : fix typo in convert-h5-to-ggml.py Ikko Ashimine 2022-12-31 02:51:08 +0900
  • a091581eb3
    cmake : add runtime destination install (#345) Avik Sengupta 2022-12-31 07:48:00 +0000
  • 68daf6e487
    whisper : avoid some memory allocations Georgi Gerganov 2022-12-30 13:42:35 +0200
  • a593b932e4
    main : add -ocsv, aka --output-csv to output a CSV file Niels Mayer 2022-12-29 04:04:00 -0800
  • 9a8ad3db69
    make : add i686 arch (close #329) Georgi Gerganov 2022-12-29 13:58:27 +0200
  • 4e6d2e98ab
    ggml : try to improve threading threads Georgi Gerganov 2022-12-26 18:51:35 +0200
  • 4e0b2069e7
    ggml : barrier refactor + static functions Georgi Gerganov 2022-12-28 18:47:30 +0200
  • ac521a566e
    ggml : simplify the SIMD code (#324) Georgi Gerganov 2022-12-24 10:22:28 +0200
  • 331c0bbddc
    examples : fix memory leak on failure to load gpt2 model (#323) Andy Maloney 2022-12-23 13:19:07 -0500
  • dc90efd504
    examples : small code cleanups (#322) Andy Maloney 2022-12-23 13:18:51 -0500
  • 7282e2109e
    ggml : use vaddvq_f32 for slightly more efficient reduce Georgi Gerganov 2022-12-23 13:48:19 +0200
  • 466ceebb78 ggml : add f16 acceleration for POWER9 ppc64le Thomas Fitzsimmons 2022-12-23 01:19:19 -0500
  • 77226aa89d
    models : fix support for spaces in path (close #315) Georgi Gerganov 2022-12-23 11:11:38 +0200
  • 543bd5627e
    whisper : use emplace_back in place of push_back (#319) Andy Maloney 2022-12-23 04:07:19 -0500
  • 62fee9a9cc
    whisper : fix mem leak on failure to load model (#318) Andy Maloney 2022-12-23 04:06:17 -0500
  • 493d94130d
    ggml : make consts static (#317) Andy Maloney 2022-12-23 04:05:27 -0500
  • 1480a5f1af
    Update README.md Georgi Gerganov 2022-12-23 11:02:46 +0200
  • 0f4227d9ee
    examples : add whisper.swiftui demo app (#308) Digipom 2022-12-23 03:56:18 -0500
  • 4c1fe0c813
    Update README.md Georgi Gerganov 2022-12-22 18:22:58 +0200
  • fa463313ad
    minor : small code cleanups (#302) Andy Maloney 2022-12-22 10:06:19 -0500
  • 501a6b455c minor : flag "ARM FMA" -> "ARM_FMA" Georgi Gerganov 2022-12-22 16:43:57 +0200
  • 91fc08c641 Build a vfpv4 library for armeabi-v7a and do runtime detection to select the right library Kevin Brothaler 2022-12-20 15:15:59 -0500
  • e1432dd91a Check for both __ARM_NEON and __ARM_FEATURE_FMA so that the project can be compiled for armv7a. Kevin Brothaler 2022-12-20 13:33:33 -0500
  • 22193cbfe8 Bump NDK version Kevin Brothaler 2022-12-20 13:33:27 -0500
  • 42c6730732
    whisper : use nullptr (C++11) instead of NULL macro (#299) Andy Maloney 2022-12-22 09:35:18 -0500
  • 76b6211f9b
    cmake : add headers to target (#298) Andy Maloney 2022-12-22 09:34:47 -0500
  • 86a277f78d
    go : run go mod tidy before building examples + fix permissions (#296) Mohit Agarwal 2022-12-22 20:04:20 +0530
  • 231bebca7d
    bindings : initial import of golang bindings (#287) David Thorpe 2022-12-20 07:54:33 +0100
  • 90564f85f9
    Update README.md Georgi Gerganov 2022-12-19 22:09:21 +0200
  • 99da1e5cc8
    cmake : enable and fix -Wall -Wextra -Wpedantic C++ warnings Georgi Gerganov 2022-12-19 20:45:08 +0200
  • 8e3f129b4d
    minor : resolves some of warnings when compiling with clang/clang++ (#294) Matheus de Sousa 2022-12-19 15:19:01 -0300
  • 1d716d6e34
    release : v1.0.4 v1.0.4 1.0.4 Georgi Gerganov 2022-12-17 19:52:42 +0200
  • 419b8a6402 Add AVX,AVX2 support for ggml_vec_scale_f32 katsu560 2022-12-17 08:42:30 +0900
  • 1eb81f863f
    make : revert accidental change of optimization flags Georgi Gerganov 2022-12-17 18:57:42 +0200
  • fba10a4c68 whisper : language auto-detect (#59) Georgi Gerganov 2022-12-17 17:58:08 +0200
  • afe2db0fe2
    Add Roadmap Georgi Gerganov 2022-12-16 23:41:57 +0200
  • a7047b2a28
    ggml : implement ggml_compute_forward_dup_f16() special cases Georgi Gerganov 2022-12-16 21:50:41 +0200
  • 32fbc8cd04
    main : add option to print the progress (#276) Georgi Gerganov 2022-12-16 20:20:43 +0200
  • b8065d90f5
    main : add "--prompt" command line argument (#90) Georgi Gerganov 2022-12-16 19:43:16 +0200
  • 4312995974 command : better indentation Georgi Gerganov 2022-12-13 21:46:42 +0200
  • 5eeeb3412d command : update README, show how to use guided mode Georgi Gerganov 2022-12-13 21:36:29 +0200
  • 6a69e3ae27 command : adding guided mode Georgi Gerganov 2022-12-13 19:21:32 +0200
  • bf69b669a0 whisper : add whisper_tokenize() Georgi Gerganov 2022-12-13 19:21:07 +0200
  • ea19ed33f1
    Update README.md (#46) Georgi Gerganov 2022-12-16 19:28:51 +0200
  • 675e787171
    Add Android sample (#277) Digipom 2022-12-16 12:20:13 -0500
  • c6c3ad5a98
    ci : add Windows build without OpenBLAS + change to Release (#85) (#282) Georgi Gerganov 2022-12-16 18:51:46 +0200
  • 6a7c82501e
    whisper : improve decoding strategy (#244) Georgi Gerganov 2022-12-16 18:31:17 +0200
  • a82d331034
    stream : update README.md + comments Georgi Gerganov 2022-12-16 18:04:19 +0200
  • c37c2443c1
    Update README.md (#56) Georgi Gerganov 2022-12-16 18:01:05 +0200
  • 0f11759406
    ggml : make more compatible with c99 (#262) Georgi Gerganov 2022-12-16 18:00:12 +0200
  • 5a5c5ddcca
    Update README.md Georgi Gerganov 2022-12-15 20:38:08 +0200
  • 34e0b4b9ef
    stream : fix build Georgi Gerganov 2022-12-15 20:15:36 +0200
  • b0f8013eb9
    stream : add sliding window mode Georgi Gerganov 2022-12-15 18:28:22 +0200
  • 124c718c73
    whisper : fix UB when reading buffer of length 0 bytes (#265) Georgi Gerganov 2022-12-13 23:13:55 +0200
  • f66ac6dc4f
    ggml : fix indentation Georgi Gerganov 2022-12-13 23:09:01 +0200
  • 9955fa4ed7
    ggml : make compatible with c99 (#262) Georgi Gerganov 2022-12-13 23:07:49 +0200
  • a613f16aec
    talk : improve prompting Georgi Gerganov 2022-12-12 23:44:36 +0200
  • 930c693989
    release : v1.0.3 1.0.3 Georgi Gerganov 2022-12-12 20:36:52 +0200
  • d8a0dde31a
    Update README.md Georgi Gerganov 2022-12-12 20:33:09 +0200
  • 9e3e6f253a
    release : v1.0.2 Georgi Gerganov 2022-12-12 20:25:56 +0200
  • 57ccd7cc4f
    Update README.md Georgi Gerganov 2022-12-12 20:23:10 +0200
  • 812ae3ffbd
    Update README.md Georgi Gerganov 2022-12-12 20:20:51 +0200
  • f309f97df6
    Node.js package (#260) Georgi Gerganov 2022-12-12 20:17:27 +0200
  • aa6adda26e
    talk : make compatible with c++11 (part 2) Georgi Gerganov 2022-12-11 20:34:04 +0200