Commit Graph

  • 3ffc83d90a swift : package no longer use ggml dependency (#1861) Georgi Gerganov 2024-02-12 19:54:11 +02:00
  • e3c5e2cba8 whisper : fix external encoder (#1860) Georgi Gerganov 2024-02-12 19:53:51 +02:00
  • b742f13e70 sync : ggml Georgi Gerganov 2024-02-12 19:07:56 +02:00
  • 52c529eeb1 ggml-alloc : allocate all leafs as if they were inputs (ggml/731) slaren 2024-02-12 18:07:14 +01:00
  • f25edade2b whisper : alternative way to handle the external encoders gg/fix-external-encoder Georgi Gerganov 2024-02-12 16:32:26 +02:00
  • 74c260fe34 whisper : fix usage of extenral encoders (e.g. CoreML) Georgi Gerganov 2024-02-12 15:19:59 +02:00
  • 551529290d talk-llama : sync llama.cpp Georgi Gerganov 2024-02-12 10:39:58 +02:00
  • 25a90ffa38 sync : ggml Georgi Gerganov 2024-02-12 09:32:15 +02:00
  • 866b67ca93 ggml-backend : sync remnant Georgi Gerganov 2024-02-12 09:27:57 +02:00
  • d7e9f58f7f CUDA: mul_mat_vec_q tiling, refactor mul mat logic (llama/5434) Johannes Gäßler 2024-02-11 19:08:39 +01:00
  • 04839bae22 vulkan: only use M-sized matmul on Apple GPUs (llama/5412) Sergio López 2024-02-11 15:12:00 +01:00
  • 3cc6e04a52 ggml : fix compile warnings (unused vars) (llama/4966) Georgi Gerganov 2024-02-11 15:33:01 +02:00
  • b7ef178b9c ggml : add mmla kernels for quantized GEMM (llama/4966) snadampal 2024-02-11 07:22:33 -06:00
  • 47dfe9d4db metal : use autoreleasepool to avoid memory leaks (llama/5437) Ian Bull 2024-02-10 02:53:28 -08:00
  • 1d3270cc8f ggml-alloc : v3 (ggml/727) slaren 2024-02-11 13:37:58 +01:00
  • a6fb6ab597 examples : added audio_ctx argument to main and server (#1857) dscripka 2024-02-12 02:19:07 -05:00
  • 163e74b6c3 metal : option to embed MSL source into compiled binary (#1842) Didzis Gosko 2024-02-11 16:41:41 +02:00
  • f273e66dc6 examples : initialize context params properly (#1852) Georgi Gerganov 2024-02-11 16:39:12 +02:00
  • 02b4c52c12 talk-llama : sync llama.cpp Georgi Gerganov 2024-02-10 10:10:59 +02:00
  • 518199c09e sync : ggml Georgi Gerganov 2024-02-10 09:56:47 +02:00
  • 8b17a2f776 src : relocate new backend sources Georgi Gerganov 2024-02-10 09:50:24 +02:00
  • b6d2827914 ggml : fix error C2078: too many initializers for MSVC ARM64 (llama/5404) Michael Podvitskiy 2024-02-09 10:56:43 +01:00
  • 9711bae0b3 CUDA: more warps for mmvq on NVIDIA (llama/5394) Johannes Gäßler 2024-02-08 21:56:40 +01:00
  • eec38f63bd CUDA: fixed mmvq kernel for bs 2,3,4 and -sm row (llama/5386) Johannes Gäßler 2024-02-07 12:40:26 +01:00
  • ef5e6b746f Basic Vulkan Multi-GPU implementation (llama/5321) 0cc4m 2024-02-07 07:54:50 +01:00
  • 77bf6b5f56 CUDA: mul_mat_vec_q max. batch size 8 -> 4 (llama/5370) Johannes Gäßler 2024-02-06 18:43:06 +01:00
  • b562fff9d0 Slight quantization improvement for Q4_K and Q5_K (llama/5361) Kawrakow 2024-02-06 17:28:02 +02:00
  • b5dec374f4 CUDA: mul_mat_vec_q for batch sizes > 1 (llama/5351) Johannes Gäßler 2024-02-06 14:44:06 +01:00
  • fa0dc6167c ggml : make use of ggml-quants.h possible in C++ code (llama/5338) Kawrakow 2024-02-05 14:09:47 +02:00
  • 55bcd62a4b ggml : avoid duplicating function calls using MIN/MAX macros (llama/5325) Dr. Tom Murphy VII Ph.D 2024-02-05 06:13:57 -05:00
  • 0ed762d691 iq2_xxs: tune quantization (llama/5320) Kawrakow 2024-02-05 10:46:06 +02:00
  • 1b5bb7792e cuda : fix LLAMA_CUDA_F16 (llama/5262) slaren 2024-02-01 18:30:17 +01:00
  • 9b735cea77 metal : add im2col F32 dst support (llama/5132) Georgi Gerganov 2024-01-31 15:35:41 +02:00
  • 12c462d656 llava : add MobileVLM support (llama/5132) JidongZhang-THU 2024-01-31 21:10:15 +08:00
  • fc7b0e2c28 ggml : limit n_threads to the max n_tasks (llama/5238) slaren 2024-01-31 13:43:03 +01:00
  • f850a067ed kompute : llama-bench support and ggml_cpu_has_kompute() (llama/5226) Jared Van Bortel 2024-01-30 19:04:37 -05:00
  • f75e1197f1 ggml : add abort_callback for cpu backend (ggml/725) Michael Podvitskiy 2024-02-09 10:42:27 +01:00
  • aa8a75e287 extra : update sync scripts Georgi Gerganov 2024-02-10 09:55:19 +02:00
  • 80e8a2ea39 server : allow CORS request with authorization headers (#1850) Valentin Gosu 2024-02-09 16:42:41 +01:00
  • 19f8048139 whisper.android : how to build with CLBlast (#1809) Neuman Vong 2024-02-10 02:39:05 +11:00
  • 0f80e5a80a whisper : expose CUDA device setting in public API (#1840) Didzis Gosko 2024-02-09 17:27:47 +02:00
  • b6559333ff make : add macOS deployment target option (#1839) Didzis Gosko 2024-02-09 17:26:29 +02:00
  • 434b8f3b96 talk-llama : stream response (#1121) Georgi Gerganov 2024-02-06 19:56:12 +02:00
  • 7a74e929c8 sync : ggml (#0) Georgi Gerganov 2024-01-30 21:30:26 +02:00
  • 361ecebe90 ggml : fix IQ3_XXS on Metal (llama/5219) Kawrakow 2024-01-30 19:15:28 +02:00
  • 807cbc672e sync : ggml (llama/0) Georgi Gerganov 2024-01-30 16:21:57 +02:00
  • 98ae5276b7 Faster AVX2 dot product for IQ2_XS (llama/5187) Kawrakow 2024-01-30 15:15:07 +02:00
  • 6adb969b09 SOTA 3-bit quants (llama/5196) Kawrakow 2024-01-30 15:14:12 +02:00
  • 8a7d6ff51a ggml alloc: Fix for null dereference on alloc failure (llama/5200) Paul Tsochantaris 2024-01-29 22:19:29 +00:00
  • 25f650a8e8 Nomic Vulkan backend (llama/4456) Jared Van Bortel 2024-01-29 15:50:50 -05:00
  • 44e517f074 ggml : add max buffer sizes to opencl and metal backends (llama/5181) slaren 2024-01-29 09:05:13 +01:00
  • cb9de61659 metal : free metal objects (llama/5161) Paul Tsochantaris 2024-01-28 19:50:16 +00:00
  • a2ef80d66f gguf : fix comparison (ggml/715) Georgi Gerganov 2024-01-29 21:08:18 +02:00
  • baa190446a ggml_cuda_cpy support for 4d tensors and float16->float32 upcasting (ggml/686) John Balis 2024-01-29 06:37:33 -06:00
  • 8f5220d81f gguf : add input validation, prevent integer overflows (ggml/709) Georgi Gerganov 2024-01-29 14:00:10 +02:00
  • 8e391fcf3a ci : fix yolo URLs + fix metal capture (ggml/712) Georgi Gerganov 2024-01-29 13:29:46 +02:00
  • 593657054e metal : add debug capture backend function (ggml/694) Jack Mousseau 2024-01-29 01:22:23 -08:00
  • ae5c4f7340 common : fix wav buffer detection (#1819) JacobLinCool 2024-01-31 01:35:08 +08:00
  • baa30bacdb server : add fields to verbose_json response (#1802) JacobLinCool 2024-01-30 20:15:55 +08:00
  • 3e6fad07aa make : update MSYS_NT (#1813) jwijffels 2024-01-30 13:13:49 +01:00
  • e72e4158de talk-llama : sync llama.cpp Georgi Gerganov 2024-01-28 19:44:10 +02:00
  • bd41733db2 sync : ggml Georgi Gerganov 2024-01-28 19:30:32 +02:00
  • 23c648e98d ggml : add Vulkan backend (llama/2059) 0cc4m 2024-01-28 18:03:59 +01:00
  • 75ab2d06f5 ggml : add unified SYCL backend for Intel GPUs (llama/2690) Abhilash Majumder 2024-01-28 21:26:23 +05:30
  • adc099edee ggml : minor type fix (int64_t -> size_t) Georgi Gerganov 2024-01-28 18:44:58 +02:00
  • 52cce82493 common : fix input buffer check (#1812) Georgi Gerganov 2024-01-27 17:33:09 +02:00
  • ef3c9ed9eb talk-llama : sync llama.cpp Georgi Gerganov 2024-01-27 17:24:53 +02:00
  • 7fe3ed5e00 sync : ggml Georgi Gerganov 2024-01-27 17:23:25 +02:00
  • 6061241292 Add OpenCL add kernel (llama/5151) 0cc4m 2024-01-26 23:07:32 +01:00
  • 0878ab7c15 cuda : fix tensor size calculation for non-split buffer (llama/5145) slaren 2024-01-26 18:59:43 +01:00
  • c65edd5b64 ggml-alloc : add 10% margin to the buffer sizes (llama/5149) slaren 2024-01-26 18:18:26 +01:00
  • 3c8d14e9c5 ggml : update softmax n_task calculation (llama/5126) snadampal 2024-01-26 11:17:59 -06:00
  • c3977cb2ce metal : remove unused n_buffers and buffers (llama/5129) Paul Tsochantaris 2024-01-26 12:16:07 +00:00
  • 6da1661bc2 metal : show compile log messages Georgi Gerganov 2024-01-25 11:26:17 +02:00
  • cc56540661 cuda : fix 2-bit quants on amd hip (llama/5105) Engininja2 2024-01-24 16:18:15 -06:00
  • 94c1ae8668 llama : pre-allocate input tensors in a separate buffer (llama/5100) slaren 2024-01-24 12:48:14 +01:00
  • 55d54359e0 metal : disable support for MUL_MAT F32 x F16 Georgi Gerganov 2024-01-23 15:50:56 +02:00
  • d33c2ad354 CUDA: more info when no device code (llama/5088) Johannes Gäßler 2024-01-23 13:31:56 +01:00
  • 9afa7ff624 minor : clean-up some warnings and style (llama/5094) Georgi Gerganov 2024-01-23 14:12:57 +02:00
  • 0649289f02 ggml : parallelize FP32 conversion when using BLAS (llama/5045) Reinforce-II 2024-01-22 21:15:08 +08:00
  • aaeaa43878 llava : MobileVLM support (llama/4954) XiaotaoChen 2024-01-22 21:09:35 +08:00
  • 078b8e23bf llama : run all KQV ops on the CPU with no KV offload (llama/5049) slaren 2024-01-20 16:05:49 +01:00
  • 74da3e1757 cuda : fix compile error in jetson platform (llama/4975) Kylin 2024-01-20 15:01:46 +08:00
  • 2d2c93a798 ggml : check ggml_add src1 type (ggml/708) Judd 2024-01-26 21:04:01 +08:00
  • 4bbb60efce docs : make model options / model install methods clearer (#1806) Michael Rienstra 2024-01-26 07:39:54 -08:00
  • 1cf679dec4 cmake : make libwhisper.so position independent (#1792) trixirt 2024-01-22 05:02:35 -08:00
  • 41026c1e4b cmake : temporary remove VLA check (#1795) Georgi Gerganov 2024-01-22 14:51:42 +02:00
  • d6b9be21d7 whisper.android : return output from benchmarks (#1785) Neuman Vong 2024-01-20 01:17:38 +11:00
  • c0329acde8 server : implement "verbose_json" format with token details (#1781) Ryan Hitchman 2024-01-18 13:58:42 -07:00
  • fb466b3417 ggml : sync ggml-metal.m Georgi Gerganov 2024-01-18 11:03:13 +02:00
  • 1f50a7d29f sync : llama.cpp Georgi Gerganov 2024-01-17 21:23:33 +02:00
  • 1de21b913d sync : ggml Georgi Gerganov 2024-01-17 21:22:38 +02:00
  • 4aea058e5a ggml : add IQ2 to test-backend-ops + refactoring (llama/4990) Georgi Gerganov 2024-01-17 18:54:56 +02:00
  • fd10234363 imatrix : offload to GPU support (llama/4957) Georgi Gerganov 2024-01-17 18:46:30 +02:00
  • 8fb5c6a409 backend : add eval callback (llama/4935) Georgi Gerganov 2024-01-17 18:39:41 +02:00
  • 2fe5fbfcc2 metal : create autorelease pool during library build (llama/4970) Georgi Gerganov 2024-01-17 18:38:39 +02:00
  • 01637e1a4c ggml : importance matrix support for legacy quants (llama/4969) Kawrakow 2024-01-16 19:51:26 +02:00
  • 1b349eb1f9 metal : log recommendedMaxWorkingSetSize on iOS 16+ (llama/4936) Alex Azarov 2024-01-16 14:33:02 +01:00
  • 138eaebead ggml : introduce GGML_CALL function annotation (llama/4850) Justine Tunney 2024-01-16 03:16:33 -08:00
  • 61b9192f27 cuda : fix dequantize kernel names (llama/4938) Georgi Gerganov 2024-01-15 13:27:00 +02:00