Commit Graph

  • 161b51d91a CUDA: faster dequantize kernels for Q4_0 and Q4_1 (llama/4938) Kawrakow 2024-01-15 07:48:06 +02:00
  • f904b31a7d Add ability to use importance matrix for all k-quants (llama/4930) Kawrakow 2024-01-14 16:21:12 +02:00
  • f6614155e4 talk-llama : optional wake-up command and audio confirmation (#1765) Benjamin Heiniger 2024-01-16 14:52:01 +01:00
  • f5f159c320 server : fix building and simplify lib deps on Windows (#1772) Przemysław Pawełczyk 2024-01-15 14:48:13 +01:00
  • 6ebba525f1 talk-llama : sync llama.cpp Georgi Gerganov 2024-01-14 18:08:20 +02:00
  • 2a5874441d talk-llama : llama.cpp Georgi Gerganov 2024-01-14 11:06:28 +02:00
  • d08445c9ad sync : ggml Georgi Gerganov 2024-01-14 10:55:18 +02:00
  • 4a945696cb metal : correctly set SIMD support flags on iOS (llama/4923) Alex Azarov 2024-01-14 09:44:39 +01:00
  • dabc964d83 2-bit quantizations (llama/4897) Kawrakow 2024-01-14 09:45:56 +02:00
  • 654baf693d scripts : sync-ggml-am.sh add option to skip commits Georgi Gerganov 2024-01-14 10:53:19 +02:00
  • f001a3b7b6 talk-llama : sync llama.cpp Georgi Gerganov 2024-01-14 00:13:17 +02:00
  • c615f2c335 sync : ggml Georgi Gerganov 2024-01-14 00:12:17 +02:00
  • d839dd0242 examples : adapt to metal API Georgi Gerganov 2024-01-14 00:09:26 +02:00
  • 435847891c ggml: cache sin/cos for RoPE (llama/4908) Johannes Gäßler 2024-01-13 21:41:37 +01:00
  • 182f290808 metal : remove old API (llama/4919) Georgi Gerganov 2024-01-13 20:45:45 +02:00
  • 447dfc11fc metal : disable log for loaded kernels (llama/4794) Georgi Gerganov 2024-01-13 18:46:37 +02:00
  • 9aa9f3b84e gguf : fix potential infinite for-loop (llama/4600) texmex76 2024-01-13 17:06:20 +01:00
  • 396ebd1e80 metal : refactor kernel loading code (llama/4794) Georgi Gerganov 2024-01-13 18:03:45 +02:00
  • 12490f4398 CUDA: faster q8_0 -> f16 dequantization (llama/4895) Johannes Gäßler 2024-01-12 20:38:54 +01:00
  • db078a9ba8 talk-llama : add optional CLI arg to set the bot name (#1764) RhinoDevel 2024-01-13 19:51:35 +01:00
  • a13a7da5ad examples : add python example for transcription (#1744) james wolf 2024-01-13 12:37:18 -05:00
  • 519f8e8684 whisper : load the model into multiple buffers of max size 1GB (#1763) Georgi Gerganov 2024-01-13 17:47:40 +02:00
  • 40ae0962f4 talk-llama : sync llama.cpp Georgi Gerganov 2024-01-12 22:04:51 +02:00
  • 1560288048 sync : ggml Georgi Gerganov 2024-01-12 21:56:50 +02:00
  • 1ad6fafd91 backend_sched : fix assignments slaren 2024-01-12 20:38:34 +01:00
  • 70840aed5f llama : ggml-backend integration (llama/4766) slaren 2024-01-12 20:07:38 +01:00
  • b24d18feb9 CUDA: fix softmax compile for old CUDA versions (llama/4862) Johannes Gäßler 2024-01-12 12:30:41 +01:00
  • 3fa98f4395 Importance Matrix calculation (llama/4861) Kawrakow 2024-01-12 06:59:57 +01:00
  • d05b7ee90e models : make all scripts to be POSIX Compliant (#1725) Sơn Phan Trung 2024-01-12 19:11:04 +07:00
  • 6dcee35129 ggml : fix 32-bit ARM compat for IQ2_XS (#1758) Georgi Gerganov 2024-01-12 14:02:30 +02:00
  • 5cb345f5e9 go : add SetInitialPrompt method to bindings (#1753) Boris Bliznioukov 2024-01-12 14:44:50 +03:00
  • fbcb52d3cd server : add more parameters to server api (#1754) George Hindle 2024-01-12 11:42:52 +00:00
  • 6b01e3fedd whisper : fix segment length with params.no_timestamps == true Georgi Gerganov 2024-01-12 13:37:38 +02:00
  • f7908f9bb8 params : don't compute timestamps when not printing them (#1755) George Hindle 2024-01-12 11:24:38 +00:00
  • 00b7a4be02 talk-llama : sync llama.cpp Georgi Gerganov 2024-01-11 22:10:10 +02:00
  • 04b0a768b8 swift : remove local ggml.h reference Georgi Gerganov 2024-01-11 22:00:12 +02:00
  • 87670425f2 swift : track ggml release branch Georgi Gerganov 2024-01-11 21:57:40 +02:00
  • 32e71a1861 sync : ggml Georgi Gerganov 2024-01-11 21:54:17 +02:00
  • 9c857cf280 sync : llama.cpp Georgi Gerganov 2024-01-11 21:49:13 +02:00
  • 97b12212dd ggml : SOTA 2-bit quants (add IQ2_XS) (llama/4856) Kawrakow 2024-01-11 20:39:39 +01:00
  • 9fa34d79ec metal : put encoder debug group behind a define (llama/4873) Paul Tsochantaris 2024-01-11 14:31:52 +00:00
  • a0a64a19dd metal : improve dequantize precision to match CPU (llama/4836) Georgi Gerganov 2024-01-09 19:37:08 +02:00
  • bbc23611fa ggml : fix vld1q_s8_x4 32-bit compat (llama/4828) Georgi Gerganov 2024-01-09 10:42:06 +02:00
  • e9783a1fb4 CUDA: faster softmax via shared memory + fp16 math (llama/4742) Johannes Gäßler 2024-01-09 08:58:55 +01:00
  • 9e0cc28792 metal : fix deprecation warning (ggml/690) Georgi Gerganov 2024-01-11 09:34:59 +02:00
  • 73072a7c73 ggml : remove ggml_cpy_inplace and ggml_cont_inplace (ggml/693) Timothy Cronin 2024-01-11 02:27:48 -05:00
  • a8ba1262ff metal : wrap each operation in debug group (ggml/690) Jack Mousseau 2024-01-10 06:19:19 -08:00
  • e66a9a7806 ggml : change GGML_MAX_NAME at compile time (ggml/682) leejet 2024-01-10 21:13:42 +08:00
  • 338442d773 Fix execlp call (ggml/689) Halalaluyafail3 2024-01-09 11:16:37 -05:00
  • 10651bddf6 SOTA 2-bit quants (llama/4773) Kawrakow 2024-01-08 16:02:32 +01:00
  • 53d4d0b30d CUDA: fixed redundant value dequantization (llama/4809) Johannes Gäßler 2024-01-07 17:24:08 +01:00
  • 2865e4710b ggml : use __builtin_amdgcn_sudot4 in __dp4a for gfx11 (llama/4787) Konstantin Zhuravlyov 2024-01-07 01:52:42 -05:00
  • c46a74a19d ggml : do not sched_yield when calling BLAS (llama/4761) Georgi Gerganov 2024-01-05 15:18:21 +02:00
  • 46dc49a6a1 ggml : include stdlib.h before intrin.h (llama/4736) Georgi Gerganov 2024-01-04 10:12:26 +02:00
  • cc7f872131 swift : checkout ggml commit instead of branch (#1750) Alexandru Mariuti 2024-01-10 17:12:06 +01:00
  • bcc1658cd0 talk-llama : add optional Piper TTS support (#1749) RhinoDevel 2024-01-10 15:15:28 +01:00
  • c46886f599 server : add request path option(#1741) Emmanuel Schmidbauer 2024-01-08 17:39:51 -05:00
  • 29f78392c1 main : add cli option to disable system prints (#1740) Georgi Gerganov 2024-01-08 16:41:28 +02:00
  • 022756a872 server : fix server temperature + add temperature_inc (#1729) Georgi Gerganov 2024-01-07 13:35:14 +02:00
  • 3b8c2dff57 talk-llama : sync latest llama.cpp Georgi Gerganov 2024-01-06 17:22:57 +02:00
  • 0b9af32a8b release : v1.5.4 v1.5.4 Georgi Gerganov 2024-01-05 17:11:27 +02:00
  • 11b1b63b14 fix : cuda order of synchronization when setting a buffer (ggml/679) Erik Scholz 2024-01-05 16:00:00 +01:00
  • 0e26a6c92e metal : switch back to default.metallib (ggml/681) Georgi Gerganov 2024-01-05 16:30:52 +02:00
  • 66d8f0b7f1 ggml : fix q2_k bpw in comments (ggml/680) Georgi Gerganov 2024-01-05 15:36:04 +02:00
  • ba5bcde874 coreml : fix ANE optimized encoder (#1716) Yajing Tang 2024-01-04 06:28:30 -08:00
  • ab0a8593c5 whisper.swiftui : add .gitignore Georgi Gerganov 2024-01-04 14:47:42 +02:00
  • 668ffc9b23 whispser : reset the "batched" timings (#1721) Georgi Gerganov 2024-01-04 13:37:25 +02:00
  • 9962371f71 release : v1.5.3 v1.5.3 Georgi Gerganov 2024-01-03 19:36:33 +02:00
  • 993acb5d41 swift : update Package.swift to use ggml as package dependency (#1701) Ashraful Islam 2024-01-03 11:30:26 -06:00
  • a3d0aa73d1 ggml : add error handling to graph_compute (#1714) Finn Voorhees 2024-01-03 08:39:43 -05:00
  • 14c57952f7 cuda : simplify expression Georgi Gerganov 2024-01-03 14:18:46 +02:00
  • 6c369d6788 cuda : mark I16 and I32 ops as unsupported Georgi Gerganov 2024-01-03 13:01:44 +02:00
  • 4cdd9aad9b metal : add kernel_get_rows_i32 Georgi Gerganov 2024-01-03 11:35:46 +02:00
  • f38c057503 metal : optimize ggml_mul_mat_id (faster Mixtral PP) (llama/4725) Georgi Gerganov 2024-01-02 21:07:47 +02:00
  • 1e5544b39b metal : enable shader debugging (cmake option) (llama/4705) Georgi Gerganov 2024-01-02 10:57:44 +02:00
  • d5673af79f ggml : add ggml_vdotq_s32 alias (llama/4715) Georgi Gerganov 2023-12-31 11:43:31 +02:00
  • a28dacec65 CUDA: fixed tensor cores not being used on RDNA3 (llama/4697) Johannes Gäßler 2023-12-30 13:52:01 +01:00
  • dbe29d4e33 ggml : add ggml_cpu_has_avx_vnni() (llama/4589) automaticcat 2023-12-30 15:07:48 +07:00
  • fe3a67c546 CUDA: fix tensor core logic for Pascal and HIP (llama/4682) Johannes Gäßler 2023-12-29 23:12:53 +01:00
  • b138ff2be3 cuda: fix vmm oom issue on NVIDIA AGX Orin (llama/4687) hydai 2023-12-30 00:31:19 +08:00
  • cf6f1e4181 ggml : extend ggml_get_rows, ggml_repeat, ggml_concat (ggml/639) Guillaume Wenzek 2023-12-29 18:07:03 +01:00
  • 620a223814 scripts : fix sync order + metal sed Georgi Gerganov 2024-01-03 11:42:42 +02:00
  • f39f9690ec examples : fix WASM Stack Overflow (#1713) Andreu Huguet 2024-01-02 17:50:04 +01:00
  • f9ca90256b docker : fix the publishing of the CUDA Docker image (#1704) bobqianic 2023-12-30 21:12:31 +00:00
  • 2623640cd6 scripts : do not sync commits from this repo Georgi Gerganov 2023-12-29 15:00:46 +02:00
  • d87de61ae6 ci : build with CLBlast + ggml-opencl use GGML_API (#1576) Tamotsu Takahashi 2023-12-29 19:23:27 +09:00
  • f5f485f899 whisper : replace tensor->n_dims with ggml_n_dims(tensor) (#1694) bobqianic 2023-12-29 09:38:35 +00:00
  • e77b27c331 sync : ggml (VMM, sync-ggml-am, dotprod ARM fixes, CUDA fixes) (#1691) Georgi Gerganov 2023-12-29 11:30:47 +02:00
  • a5cc3dc8a2 download : fix large q5 model name (#1695) Dimo 2023-12-29 10:14:32 +01:00
  • 37a709f655 whisper : Replace WHISPER_PRINT_DEBUG with WHISPER_LOG_DEBUG (#1681) bobqianic 2023-12-23 12:02:58 +00:00
  • 3a5302108d sync : ggml (ggml_scale, ggml_row_size, etc.) (#1677) Georgi Gerganov 2023-12-22 17:53:39 +02:00
  • d2ee117a0a docker : Dockerize whisper.cpp (#1674) Chaoqun 2023-12-22 19:16:02 +08:00
  • db8ccdb850 CI : Add coverage for talk-llama when WHISPER_CUBLAS=1 (#1672) bobqianic 2023-12-21 22:39:46 +00:00
  • d2419030b0 examples : Revert CMakeLists.txt for talk-llama (#1669) bobqianic 2023-12-21 20:48:52 +00:00
  • 8986690c2a cmake : set default CUDA architectures (#1667) bobqianic 2023-12-21 13:44:04 +00:00
  • 9286d3f584 bench.py : add different large models (#1655) Alfredo Montesinos 2023-12-19 04:40:14 -06:00
  • 940de9dbe9 wchess : update README.md Georgi Gerganov 2023-12-14 22:00:47 +02:00
  • 88112c8afb release : v1.5.2 v1.5.2 1.5.2 Georgi Gerganov 2023-12-14 17:56:11 +02:00
  • 375585c07c wchess : update readme Georgi Gerganov 2023-12-14 17:51:14 +02:00
  • fd99ece8e3 wchess : whisper assisted chess (#1595) fraxy-v 2023-12-14 15:58:26 +02:00