Commit graph

  • 1a454eb561 mpt : removed hardcoded +178 from convert script in favor of utilizing hparams["vocab_size"] Jan Ploski 2023-10-09 16:48:01 +02:00
  • ad3c2f3b23 comment out n_past instead of marking it unused Cebtenzzre 2023-10-09 10:16:24 -04:00
  • 292363e556 Merge branch 'master' of https://github.com/ggerganov/llama.cpp into pull-3417 Cebtenzzre 2023-10-09 09:25:10 -04:00
  • 7d6a24aad4 mpt : updated convert-mpt-hf-to-gguf.py to reflect changes made to convert-gptneox-hf-to-gguf.py in pr:3252 Jan Ploski 2023-10-06 13:53:32 +02:00
  • 4759bfd64c fix: rm designated initializers M. Yusuf Sarıgöz 2023-10-09 15:54:55 +03:00
  • 49244be7e9
    Merge branch 'ggerganov:master' into context-sensitive-help pudepiedj 2023-10-09 13:51:08 +01:00
  • 3f07ed90a4 Added prompt-file to hep pudepiedj 2023-10-09 13:45:12 +01:00
  • 95bd60a0a6
    ggml-alloc : fix assert in debug builds (#3555) b1357 slaren 2023-10-09 14:44:58 +02:00
  • 51446bf921 Naming convention pudepiedj 2023-10-09 13:35:52 +01:00
  • ee7456926e ggml-alloc : fix assert in debug builds alloc-assert-fix slaren 2023-10-09 14:33:12 +02:00
  • 6a2e064d68 fix unexpected behavior when multiple requests are canceled FSSRepo 2023-10-09 08:14:42 -04:00
  • 8d3681ddbe refactored some issues FSSRepo 2023-10-09 08:00:59 -04:00
  • 59e7c0c51b fixed cancel + removed useless code FSSRepo 2023-10-09 07:53:00 -04:00
  • d78e816365 rm unused import M. Yusuf Sarıgöz 2023-10-09 14:44:35 +03:00
  • fcca0a7004
    refact : fix convert script + zero out KV cache to avoid nans (#3523) b1356 Georgi Gerganov 2023-10-09 14:32:17 +03:00
  • dcc09d2596
    metal : do not use mul_mm kernels when ne00 < 64 (#3542) b1355 Georgi Gerganov 2023-10-09 14:28:27 +03:00
  • 8278a7364a rm unused batch image preprocessing M. Yusuf Sarıgöz 2023-10-09 14:22:18 +03:00
  • f6e92a84a2 Merge branch 'master' into context-sensitive-help pudepiedj 2023-10-09 12:21:31 +01:00
  • 9b0ec4d2cc Are you happy editorconfig? M. Yusuf Sarıgöz 2023-10-09 13:42:04 +03:00
  • 54495c9474 Some cleanup M. Yusuf Sarıgöz 2023-10-09 13:38:48 +03:00
  • 990e8cb329 New comment pudepiedj 2023-10-09 10:36:11 +01:00
  • 3e4de67fdd Update find_implemented_args.py pudepiedj 2023-10-09 10:01:21 +01:00
  • 2e17fcfdba Comment in common.cpp pudepiedj 2023-10-09 09:46:26 +01:00
  • 8af7e2103c Update readme M. Yusuf Sarıgöz 2023-10-09 11:10:09 +03:00
  • 32bdf0ee4b Final reconciliation pudepiedj 2023-10-09 09:10:07 +01:00
  • 444dbce888 Add readme M. Yusuf Sarıgöz 2023-10-09 09:47:56 +03:00
  • d9b89e5421 fix(bloom): fix model size xingchensong 2023-10-09 13:46:50 +08:00
  • d55f155621 feat: Support bloom models xingchensong 2023-10-09 12:06:42 +08:00
  • e4efbdb8c1 update format based on editor-config checker result chenqiny 2023-10-09 11:51:15 +08:00
  • 8640f3bb4e
    Merge branch 'ggerganov:master' into master Qin Yue Chen 2023-10-09 11:14:11 +08:00
  • 1ce890a7c0 support --bigendian option for s390x 1. verified with baichuan7b-chat with float 16 on s390x 2. verified with baichuan7b-chat 3. verified with chinese-alpaca-2-13b-f16 chenqiny 2023-10-09 10:40:41 +08:00
  • c8d7b1b897 remove useless line FSSRepo 2023-10-08 22:36:16 -04:00
  • 8a8535bb6d fix cors + regen + cancel funcs FSSRepo 2023-10-08 22:30:43 -04:00
  • 57c3442ea5 make printed param formatting more consistent Cebtenzzre 2023-10-08 18:10:02 -04:00
  • 982c908984 Update contextual help pudepiedj 2023-10-08 22:26:13 +01:00
  • 09c31027db fix uninitialized cparams Cebtenzzre 2023-10-08 16:59:23 -04:00
  • 0050e1ecc6 default n_yarn_orig_ctx to n_ctx_train Cebtenzzre 2023-10-08 16:51:51 -04:00
  • babf0e0c8f fix rope_cuda parameter order Cebtenzzre 2023-10-08 14:31:35 -04:00
  • 01bef02900 Export function to fetch/create default sampler states KerfuffleV2 2023-10-08 11:59:07 -06:00
  • 52def09a31 Try to fix zig build. KerfuffleV2 2023-10-08 09:39:27 -06:00
  • fad923a82d Fix mirostat by completely refactoring sampling! KerfuffleV2 2023-10-08 09:16:59 -06:00
  • 0e6db6fec1 Fix mirostat state when using multiple sequences KerfuffleV2 2023-10-08 03:52:45 -06:00
  • db3abcc114
    sync : ggml (ggml-backend) (#3548) b1354 Georgi Gerganov 2023-10-08 20:19:14 +03:00
  • 2fafc15be6
    zig : add ggml-backend to the build Georgi Gerganov 2023-10-08 19:48:00 +03:00
  • 80e53af236 fixed a bug in lite Concedo 2023-10-09 00:18:03 +08:00
  • 4e5b6293ab adjust streaming timings Concedo 2023-10-08 23:12:45 +08:00
  • e967717385 Merge branch 'master' into concedo_experimental Concedo 2023-10-08 22:55:44 +08:00
  • 840b244c17 update lite Concedo 2023-10-08 22:55:05 +08:00
  • 141329f8db Merge branch 'master' of github.com:ggerganov/llama.cpp vvhg1 2023-10-08 16:45:46 +02:00
  • 20fad2e7d4
    sync : ggml (ggml-backend) Georgi Gerganov 2023-10-08 17:36:03 +03:00
  • eee42c670e
    ci : add Zig CI/CD and fix build (#2996) b1353 Matheus C. França 2023-10-08 10:59:20 -03:00
  • 9872e90696
    ci : remove trailing whitespace Georgi Gerganov 2023-10-08 16:58:39 +03:00
  • 7438728d51
    batched : minor fix table Georgi Gerganov 2023-10-08 16:35:54 +03:00
  • bf06d654de
    batched : add bench tool Georgi Gerganov 2023-10-08 15:57:16 +03:00
  • 2a04d0b5a1 Merge branch 'master' into llava M. Yusuf Sarıgöz 2023-10-08 15:40:39 +03:00
  • 95da79e740 fix: trailing whitespace M. Yusuf Sarıgöz 2023-10-08 15:38:47 +03:00
  • 1b137d5b66 fix build_compiler Matheus Catarino França 2023-10-08 09:30:05 -03:00
  • 204d08be3d fix: new line at EoF M. Yusuf Sarıgöz 2023-10-08 15:24:13 +03:00
  • 52776b0b78 zig CI/CD and fix build Matheus Catarino França 2023-09-03 16:29:09 -03:00
  • 0c2bd79781 fix: crlf -> lf M. Yusuf Sarıgöz 2023-10-08 15:20:39 +03:00
  • 94eeac358a Use ggml_allocr + rm unnecessary code M. Yusuf Sarıgöz 2023-10-08 14:58:47 +03:00
  • 1693fcb713 CLBlast: Fix matrix-vector multiplication shibe2 2023-10-08 15:46:12 +04:00
  • d9dae9318e Merge branch 'master' of github.com:ggerganov/llama.cpp vvhg1 2023-10-08 13:23:15 +02:00
  • 8e6716a102
    api_like_OAI.py : compat with Microsoft Guidance (#2746) Ryder Wishart 2023-10-08 03:55:58 -07:00
  • f861ff916d
    gitignore : server-parallel Georgi Gerganov 2023-10-08 13:54:54 +03:00
  • 528c418478
    Merge branch 'master' into ryderwishart-patch-1 Georgi Gerganov 2023-10-08 13:54:27 +03:00
  • 9c38d181d4
    api_like_OAI.py : simplify function (#2796) arcrank 2023-10-08 06:52:57 -04:00
  • 2f7f634143
    Merge branch 'master' into HEAD Georgi Gerganov 2023-10-08 13:31:33 +03:00
  • a1202a31ed
    k-quants : fix comments about block sizing (#3499) b1350 Johannes Rudolph 2023-10-08 12:21:19 +02:00
  • dad1cdb1ef Simplify q6_k fp16 fix 0cc4m 2023-10-08 11:56:07 +02:00
  • 85c1a63a15 Fix q6_k for GPUs without fp16 support 0cc4m 2023-10-08 11:52:44 +02:00
  • fdd5ad9a54
    metal : do not use mul_mm kernels when ne00 < 64 Georgi Gerganov 2023-10-08 12:05:15 +03:00
  • ee268b5446
    llama : no longer perform uninitialized access to the KV cache fix-kv-cache-access Georgi Gerganov 2023-10-08 11:49:38 +03:00
  • acead654d2
    Merge branch 'master' into fix-refact fix-refact Georgi Gerganov 2023-10-08 11:25:16 +03:00
  • 94e502dfb7
    ci : enable on obj-c changes + fix metal build (#3540) b1349 Georgi Gerganov 2023-10-08 11:24:50 +03:00
  • 7d8b24932f
    zig : fix build by introducing train.cpp (#3539) Luo Tian 2023-10-08 16:24:01 +08:00
  • 7812ea0e02
    ci : enable on obj-c changes + fix metal build Georgi Gerganov 2023-10-08 11:20:40 +03:00
  • 0f8df395ce
    metal : assert various kernel requirements Georgi Gerganov 2023-10-08 11:04:20 +03:00
  • d8fa5ca230 Merge branch 'master' into concedo_experimental Concedo 2023-10-08 15:51:42 +08:00
  • 80dfe2ba49 blasthreads should apply for any thread count below 32 Concedo 2023-10-08 15:51:18 +08:00
  • a2b8473354 force flush sse Concedo 2023-10-08 15:12:07 +08:00
  • b0ec5218c3
    metal : support MTLGPUFamily < Apple7, formatting, style (#3524) Georgi Gerganov 2023-10-08 10:01:53 +03:00
  • 6b9554a740
    metal : print more GPU info + disable mul_mm for MTLGPUFamiliy < Apple7 metal-improve-batching Georgi Gerganov 2023-10-08 09:53:38 +03:00
  • 63d3b06a43
    llama : fix missing break in Persimmon arch case statements (#3535) b1346 Kerfuffle 2023-10-07 23:22:17 -06:00
  • 133897a558 updated lite (+1 squashed commits) Concedo 2023-10-08 11:11:08 +08:00
  • fa62c8c73a support s390x big endian chenqiny 2023-10-08 11:47:39 +08:00
  • 8c94145f78 fix zig build by introducing train.cpp Luo Tian 2023-10-08 11:40:50 +08:00
  • f797cba377 Merge branch 'master' into concedo_experimental Concedo 2023-10-08 10:43:34 +08:00
  • b592c70deb Rewrite special token handling from #1931 staviq 2023-10-08 02:43:23 +02:00
  • aa4e11dfe5 Fix missing break in Persimmon arch case statements KerfuffleV2 2023-10-07 17:01:46 -06:00
  • 22daada82e
    Merge ea73dace98 into a16e89cec8 Elsa Granger 2023-10-07 15:30:50 -07:00
  • 8690f425ec LLaVA is working e2e, needs to optimize memory allocation + cleanup M. Yusuf Sarıgöz 2023-10-08 01:15:13 +03:00
  • a16e89cec8
    Fix trying to strip newline from empty prompt and cfg prompt file content (#3534) b1345 Kerfuffle 2023-10-07 15:31:41 -06:00
  • 9c5d6f0ef6 Update helper dev pudepiedj 2023-10-07 21:40:45 +01:00
  • 9bd050f133 Merge branch 'master' of https://github.com/ggerganov/llama.cpp into ntkv2 Cebtenzzre 2023-10-07 15:54:05 -04:00
  • 34f0fccef9 Fix trying to strip newline from empty prompt and cfg prompt file content KerfuffleV2 2023-10-07 14:26:20 -06:00
  • 4d03833211
    gguf.py : fix CI for publishing GGUF package (#3532) b1344 M. Yusuf Sarıgöz 2023-10-07 22:14:10 +03:00
  • ba44776dc2 bump version gguf-v0.4.4 gguf-fix-publish M. Yusuf Sarıgöz 2023-10-07 21:47:48 +03:00
  • 5ad84f0ba4 bump version gguf-v0.4.3 M. Yusuf Sarıgöz 2023-10-07 21:43:59 +03:00
  • 6dd3e8ea6a bump version gguf-v0.4.2 M. Yusuf Sarıgöz 2023-10-07 21:29:29 +03:00