Commit graph

  • 084dd216cd Cleanups for gguf-py KerfuffleV2 2023-08-27 10:58:13 -06:00
  • 1793f25cfa convert: Fix permute calls and method/func definitions KerfuffleV2 2023-08-27 02:56:47 -06:00
  • 849a31f17d Remove comment that no longer applies Iwan Kawrakow 2023-08-29 19:06:32 +03:00
  • 5a2ba37a3a
    Stop generation at multiple linebreaks, bind to <F2> chaihahaha 2023-08-29 23:57:00 +08:00
  • 380fa0f0ca fixed broken typical sampler issues Concedo 2023-08-29 23:50:59 +08:00
  • ba5590f44b review: simplify ifs and str comparisons staviq 2023-08-29 17:44:54 +02:00
  • 3b72afb960
    Merge branch 'ggerganov:master' into master chaihahaha 2023-08-29 23:43:21 +08:00
  • aef36d0dc2 baby-llama : remove commented opt_params_adam Cebtenzzre 2023-08-29 11:38:20 -04:00
  • a9ed865205 make : move -Werror=implicit-int to CFLAGS Cebtenzzre 2023-08-29 11:28:22 -04:00
  • 3a10f5aa34 review: macro f lowercase, str append to sstream staviq 2023-08-29 17:29:37 +02:00
  • f82db06ab5 Merge remote-tracking branch 'upstream/master' into check-c-compliance Cebtenzzre 2023-08-29 11:26:51 -04:00
  • 45bdb9af95 10X faster BPE tokenizer Iwan Kawrakow 2023-08-29 18:06:14 +03:00
  • f7ac43155b
    Merge branch 'ggerganov:master' into betterlogs staviq 2023-08-29 17:02:11 +02:00
  • 5cae6ba54a remove gcc-8 from macos build test Alon Faraj 2023-08-29 17:19:19 +03:00
  • b5cd70df9e exclude other ops that set an inplace flag from automatic inplace slaren 2023-08-29 15:54:23 +02:00
  • 5460aec056 update ggml_diag_mask to work correctly with automatic inplace slaren 2023-08-29 15:51:42 +02:00
  • 53885d7256
    py : fix "usage" messages (#2873) maddes8cht 2023-08-29 15:51:02 +02:00
  • 28c6e324d2 update ggml-alloc to use view_src slaren 2023-08-29 15:51:01 +02:00
  • 1f5a0f09a8 ggml : add view_src and view_offs slaren 2023-08-29 13:36:56 +02:00
  • 5cb2b0c1a9 fix path to model Alon Faraj 2023-08-29 16:17:59 +03:00
  • c2fd80a0e8 enable tests/test-tokenizer-0-llama Alon Faraj 2023-08-29 16:09:19 +03:00
  • 89cc7a3338 makefile: - add all - add test Alon Faraj 2023-08-29 16:04:29 +03:00
  • f0a13e2add build ci: run make test Alon Faraj 2023-08-29 16:04:10 +03:00
  • 522da032a1 Fix "usage" message in convert-to-gguf python scripts Mathias Bachmann 2023-08-29 15:03:37 +02:00
  • ecdf113c69
    main : add trace log Georgi Gerganov 2023-08-29 15:12:37 +03:00
  • c72d344c1a
    main : wip logs Georgi Gerganov 2023-08-29 14:23:39 +03:00
  • 5c978f48a4
    log : try to fix Windows build Georgi Gerganov 2023-08-29 14:21:14 +03:00
  • 891ac40668
    Merge branch 'master' into HEAD Georgi Gerganov 2023-08-29 14:01:24 +03:00
  • 6b4c65be83
    log : do not append to existing log + disable file line func by default Georgi Gerganov 2023-08-29 13:59:55 +03:00
  • b97958a511
    log : fix compile warnings Georgi Gerganov 2023-08-29 13:46:43 +03:00
  • 2c1930d338
    Merge branch 'master' into HEAD Georgi Gerganov 2023-08-29 13:45:49 +03:00
  • bcce96ba4d
    convert.py : fix baichuan7B support (#2870) jameswu2014 2023-08-29 17:48:41 +08:00
  • 2b148cde8a
    convert.py : fix trailing whitespaces Georgi Gerganov 2023-08-29 12:48:01 +03:00
  • 74e0caeb82
    readme : add react-native binding (#2869) Jhen-Jie Hong 2023-08-29 17:30:10 +08:00
  • d4b5e16c32
    make : fix clang tests build, add missing examples (#2859) b1112 Cebtenzzre 2023-08-29 04:42:41 -04:00
  • 3aedf97ca0 [Fix]: convert.py support baichuan7B jameswu2014 2023-08-29 16:37:59 +08:00
  • 3a007648f2
    metal : add option to disable debug logs (close #2764) b1111 Georgi Gerganov 2023-08-29 11:33:46 +03:00
  • 611363ac79 scripts : add pipefail Georgi Gerganov 2023-08-29 10:50:30 +03:00
  • 8d067125b2 readme : add react-native binding Jhen 2023-08-29 15:46:50 +08:00
  • 95b6e5212f
    added struct to llama_dump_timing_info_yaml's llama_context (#2857) b1109 Marcus Dunn 2023-08-28 23:33:27 -07:00
  • ecb1b20c85
    add gguf constants and load/save functions from train-text-from-scratch xaedes 2023-08-29 01:40:02 +02:00
  • e030f7b2c5
    add LLM_KV_TRAINING_TYPE to train-text-from-scratch checkpoints xaedes 2023-08-29 01:27:28 +02:00
  • ca97583f0b
    remove vocab related code as it is unnecessary xaedes 2023-08-29 01:19:45 +02:00
  • a3b45298f1
    remove unused code xaedes 2023-08-29 01:12:51 +02:00
  • 1faee64db9
    handle rms_norm and rope parameters the same as in train-text-from-scratch xaedes 2023-08-29 01:09:35 +02:00
  • 007280c82f
    make default value of float member a float literal xaedes 2023-08-29 01:00:58 +02:00
  • 49af7fbe12
    add comment explaining why finetune checkpoints are allocated in one block xaedes 2023-08-29 00:57:28 +02:00
  • 9a28bce29a
    reduce large memory overhead in train-text-from-scratch xaedes 2023-08-29 00:56:44 +02:00
  • 271c0300de
    remove prediction related code to reduce duplicated code with main xaedes 2023-08-29 00:50:59 +02:00
  • b44138c87d fix most gcc and clang warnings Cebtenzzre 2023-08-25 16:46:12 -04:00
  • 5ce92aed37
    finetune bug fixes to compile with merged in code from master xaedes 2023-08-29 00:41:19 +02:00
  • daedc6f419
    replace llama_n_mult by llama_n_ff xaedes 2023-08-29 00:40:53 +02:00
  • aa8016e95d
    bug fix: replace GGML_TYPE_SIZE[t] by ggml_type_size(t) xaedes 2023-08-29 00:40:30 +02:00
  • aecc3b3890
    fix dump_non_result_info_yaml to output multiple lora adapters xaedes 2023-08-29 00:39:59 +02:00
  • 2cbca64b4b make : fix build-info.h dependencies Cebtenzzre 2023-08-28 18:23:43 -04:00
  • 281a823eb4 make : add missing examples Cebtenzzre 2023-08-28 17:27:20 -04:00
  • b04263c1fe
    Merge branch 'master' into finetune-lora xaedes 2023-08-29 00:17:04 +02:00
  • d79af7d868 Resolve conflicts Ravindra Marella 2023-08-29 02:58:09 +05:30
  • 59a8ec15ab make : do not pass headers to the compiler Cebtenzzre 2023-08-28 17:18:16 -04:00
  • 8411b02fc7
    Merge branch 'ggerganov:master' into master m3ndax 2023-08-28 22:56:18 +02:00
  • 44c117f41e
    train : mem usage and other improvements (#2439) b1108 xaedes 2023-08-28 21:51:47 +02:00
  • 011c2585e8
    added struct to llama_dump_timing_info_yaml's llama_context Marcus Dunn 2023-08-28 12:35:57 -07:00
  • 93535a460a
    train : fix compile warnings Georgi Gerganov 2023-08-28 22:26:10 +03:00
  • f6828cba9e
    remove GGML_ALIGNED_REALLOC and use normal malloc/realloc/free for gguf ctx->kv & ctx->infos xaedes 2023-08-28 20:21:03 +02:00
  • fa6efc998c
    Merge branch 'ggerganov:master' into master m3ndax 2023-08-28 19:34:18 +02:00
  • 78bc8538d5
    Update convert-llama2c-to-ggml.cpp m3ndax 2023-08-28 19:21:00 +02:00
  • 43033b7bb4
    llama-bench : set locale to utf8 (#2832) b1107 slaren 2023-08-28 19:19:18 +02:00
  • 440d221c62
    add missing blank line at end of file xaedes 2023-08-28 19:17:47 +02:00
  • a925e9304a
    fix non-windows GGML_ALIGNED_REALLOC xaedes 2023-08-28 19:16:27 +02:00
  • 12c4e5b50f
    Merge branch 'master' into pr-train-mem-usage-improvements xaedes 2023-08-28 19:14:18 +02:00
  • 17ab46dffc
    update train-text-from-scratch README.md xaedes 2023-08-28 19:13:20 +02:00
  • 3e7dfd08c4
    remove prediction related code xaedes 2023-08-28 19:08:11 +02:00
  • 3155019b53
    remove trailing whitespace xaedes 2023-08-28 18:39:50 +02:00
  • 63bf200b87
    remove code used to verify correctness of checkpoint file conversion xaedes 2023-08-28 18:38:52 +02:00
  • 31c093c2cc
    bug fixes for convert-train-checkpoint-to-gguf.py loading checkpoints with opt_version=0 xaedes 2023-08-28 18:33:00 +02:00
  • e8df9e6815
    temporarily add code to write old checkpoint files xaedes 2023-08-28 18:17:51 +02:00
  • 6b214d1451
    Merge branch 'ggerganov:master' into master m3ndax 2023-08-28 18:15:52 +02:00
  • 6b73ef1201
    YAML result logging + preset script (#2657) b1106 Johannes Gäßler 2023-08-28 17:59:39 +02:00
  • 5f27ade48e
    bug fixes for convert-train-checkpoint-to-gguf xaedes 2023-08-28 17:57:10 +02:00
  • e9e8ac4c82 Fix multiple offloading JohannesGaessler 2023-08-28 17:44:10 +02:00
  • 75fafcbccc
    make : fix tests build (#2855) b1105 alonfaraj 2023-08-28 18:38:35 +03:00
  • be475f60af
    llama.cpp : fix wrong vsnprintf call in MS compiler (#2856) b1104 grahameth 2023-08-28 17:38:12 +02:00
  • 84a68186da
    editorconfig : fixes Georgi Gerganov 2023-08-28 18:36:13 +03:00
  • 6a30dffebb YAML result logging + preset script JohannesGaessler 2023-08-15 15:46:10 +02:00
  • 9f5b7813c6 skip-unused: fix -ngl=1 case by ensure input & of view are offloaded consistently ochafik 2023-08-28 15:44:55 +01:00
  • 7734df8937 llama.cpp: Fix wrong vsnprintf call in MS compiler grahameth 2023-08-28 16:34:01 +02:00
  • c690c20362
    print data checksums before saving and after loading to verify correctness xaedes 2023-08-28 16:09:53 +02:00
  • cb3d35b22f makefile: - fix test name - add missing tests build Alon Faraj 2023-08-28 17:07:09 +03:00
  • f6a446ec01 skip-unused: revert extra spaces ochafik 2023-08-28 14:30:29 +01:00
  • f97f92bce5
    remove trailing whitespace xaedes 2023-08-28 15:28:19 +02:00
  • 2eaeb7e872 skip-unused: fix brackets & tabs ochafik 2023-08-28 14:27:54 +01:00
  • daa0b6c6a4
    set name of tensors with empty name from what was read from gguf xaedes 2023-08-28 15:27:26 +02:00
  • e86b3e3257
    avoid printing lots of spaced on the unusual case that loss gets nan xaedes 2023-08-28 15:26:44 +02:00
  • 3d8d884049
    bug fix in load_opt_context_gguf xaedes 2023-08-28 15:07:00 +02:00
  • 3af6b86301
    ggml : tiny ggml_vec_dot_q4_K_q8_K AVX2 improvement (#2819) b1103 Ronny Brendel 2023-08-28 14:51:08 +02:00
  • cec628e7fd
    temporarily disable broken 512 build ci_cublas_linux-b1104-cec628e Green Sky 2023-08-26 01:54:14 +02:00
  • 0e1730a90e
    ci: add linux binaries to release build Green Sky 2023-05-05 00:01:30 +02:00
  • 35feac6560
    ggml : sync (mem align to header + conv_transpose_2d fixes + ggml_alloc) (#2852) b1102 Georgi Gerganov 2023-08-28 14:24:53 +03:00
  • 92b1bbd2ec
    CUDA: fix RoPE asserts, block sizes (#2833) b1101 Johannes Gäßler 2023-08-28 13:23:55 +02:00
  • 8af50b3c81
    ggml-alloc : sync more fixes Georgi Gerganov 2023-08-28 14:14:08 +03:00