Commit Graph

  • 4349ff1fc2 Starting to fix some tests. Nicolas Patry 2023-11-11 01:02:15 +01:00
  • 03ad494fcd Tweak the basic example to show how to implement sort. sort laurent 2023-11-30 08:01:42 +00:00
  • 7c3cfd1086 Use the llama weight names for the Yi example. (#1381) Laurent Mazare 2023-11-27 20:42:52 +00:00
  • e2eb6590ed Merge pull request #1323 from huggingface/metal3 Nicolas Patry 2023-11-27 13:06:01 +01:00
  • 481c45d78d Add a basic implementation for slice-assign. (#1377) Laurent Mazare 2023-11-26 17:31:22 +00:00
  • 14a2bdc062 Small tweak: remove the macro usage for the range indexing trait. (#1376) Laurent Mazare 2023-11-26 16:30:59 +00:00
  • bfa7c8fc01 Implement the module trait directly for QMatMul. (#1372) Laurent Mazare 2023-11-25 10:09:45 +00:00
  • 762e996ce6 Distibert (#1366) Odunayo 2023-11-24 10:09:14 -05:00
  • ca19a9af62 Fix linspace implementation (#1358) MilkFather 2023-11-23 15:35:13 +08:00
  • ec23427d60 Ensure to copy data to cpu before iterating. (#1360) Marcus Asteborg 2023-11-22 23:24:25 -08:00
  • f83e14f68d Add candle-lora transformers to readme? (#1356) Eric Buehler 2023-11-21 12:54:24 -05:00
  • c7e613ab5e Update the readme. (#1354) Laurent Mazare 2023-11-21 09:38:27 +00:00
  • 8f63f68289 Fix the kalosm link (#1353) Laurent Mazare 2023-11-21 05:18:14 +00:00
  • 1edc3ddf24 Allowing feature metal to compile. Nicolas Patry 2023-11-20 20:17:16 +01:00
  • b380657bfe Merge pull request #1309 from huggingface/metal2 Nicolas Patry 2023-11-20 17:24:01 +01:00
  • 60f624a902 Moving tests around. Nicolas Patry 2023-11-20 16:17:19 +01:00
  • 8d6c6de8e0 Missing new test. Nicolas Patry 2023-11-11 17:42:53 +01:00
  • 7ec345c2eb Adding the test scaffolding. Nicolas Patry 2023-11-10 23:31:09 +01:00
  • 671fc29b36 Fmt. Nicolas Patry 2023-11-20 14:38:20 +01:00
  • dc64adb8e4 Fixing cos_f16 test. Nicolas Patry 2023-11-20 14:17:07 +01:00
  • c66e5d4716 Fix comments. Nicolas Patry 2023-11-20 14:00:39 +01:00
  • bd3b243725 Update candle-metal-kernels/Cargo.toml Nicolas Patry 2023-11-13 18:53:16 +01:00
  • 2813fb5dbc Cleanup fixed a few ops removed debugging scaffolding. Nicolas Patry 2023-11-10 23:00:32 +01:00
  • 7cfffcac10 Debugging rope. Nicolas Patry 2023-11-10 20:25:24 +01:00
  • 38de52bc4b Fixed matmul (display still broken without casting back to CPU first? ) Nicolas Patry 2023-11-10 20:09:25 +01:00
  • d46670f7c0 Tmp state. Nicolas Patry 2023-11-10 15:35:46 +01:00
  • f710fab02e Fixing the kernels + launches to make them faster. Nicolas Patry 2023-11-10 11:14:51 +01:00
  • f82bf2d915 Adding indexing. Nicolas Patry 2023-11-10 01:58:51 +01:00
  • df6814f34e Refactor to simplify our lives for settings the params in the encoder. Nicolas Patry 2023-11-10 01:24:49 +01:00
  • 39406a6721 Adding the actual backend Nicolas Patry 2023-11-09 19:53:14 +01:00
  • 976ad9f9c2 Remove tracing. Nicolas Patry 2023-11-09 19:41:08 +01:00
  • a4c4a56429 Metal part 1 - Scaffolding for metal. Nicolas Patry 2023-11-09 19:30:59 +01:00
  • c93a17694b Speeding up copies using blit. metal4_arc Nicolas Patry 2023-11-19 23:00:10 +01:00
  • 7052b9c884 Interleave committing. Nicolas Patry 2023-11-19 22:50:41 +01:00
  • d9278e3a63 Revert "Tmp gemm." Nicolas Patry 2023-11-19 20:45:06 +01:00
  • c65f68e988 Tmp gemm. tmpgemm Nicolas Patry 2023-11-19 20:43:59 +01:00
  • f49bf6a81d Fix OpenChat 3.5 tokenizer (#1347) Lucas de Ávila Martins 2023-11-19 15:48:04 -03:00
  • 992a788da1 Add OpenChat 3.5 to quantized examples (#1346) Lucas de Ávila Martins 2023-11-19 15:28:52 -03:00
  • 8d8f48c60c feat: add test for individual onnx ops (#1332) drbh 2023-11-19 02:17:09 -05:00
  • eed1631ee2 Reuse buffers on our own reference counts. Nicolas Patry 2023-11-18 23:28:59 +01:00
  • 251c65f9f1 Metal operational. Nicolas Patry 2023-11-18 00:52:38 +01:00
  • d31f11035f Support for CumSum in ONNX models. (#1340) Laurent Mazare 2023-11-17 22:03:40 +00:00
  • a0010898cc Better batched matmul. Nicolas Patry 2023-11-17 10:36:57 +01:00
  • 9ab3f9729f Use the whisper-v3 tokenizer now that it has been added. (#1337) Laurent Mazare 2023-11-16 22:10:31 +00:00
  • a1f41ab37b feat: adds reset_kv_cache (#1335) drbh 2023-11-16 16:17:42 -05:00
  • 92a05b51cf fix: address clippy 0.1.74 issues (#1336) drbh 2023-11-16 16:15:22 -05:00
  • 7e49e0af96 Tmp for allocator. tmpm4 tmp5 Nicolas Patry 2023-11-16 12:50:41 +01:00
  • 181d2299b2 TMp. Nicolas Patry 2023-11-16 11:41:06 +01:00
  • 2801541e5f new_owned -> new()..to_owned(). Nicolas Patry 2023-11-16 11:07:56 +01:00
  • c6763e3b41 Add a simple implementation of cumsum. (#1334) Laurent Mazare 2023-11-15 21:11:15 +00:00
  • 347e31c9ff Add the tril/triu/eye ops. (#1333) Laurent Mazare 2023-11-15 20:34:37 +00:00
  • e8c1c31245 Tmp commit for the heap experiment (heap is indeed decreasing). metal_heap Nicolas Patry 2023-11-14 17:04:23 +01:00
  • 51f05e997d Heap buffers for metal ? Nicolas Patry 2023-11-13 18:56:46 +01:00
  • 4289984d32 Remove some prints. Nicolas Patry 2023-11-13 14:51:40 +01:00
  • 1471f98f0b BF16 metal fix. Nicolas Patry 2023-11-13 14:44:20 +01:00
  • dd4a40f1c0 Fixes + cache compute_pipeline_state. Nicolas Patry 2023-11-13 14:33:16 +01:00
  • 79845bd93b Working version for llama2-c. Nicolas Patry 2023-11-13 12:36:27 +01:00
  • f4fcf60900 Update readme.md (#1322) Ryan Kopf 2023-11-12 03:46:19 -06:00
  • 12561b31d3 Fix pose estimation image path (#1326) Bernardo de Lemos 2023-11-12 09:45:26 +00:00
  • a209ce8ceb Update for 0.3.1. (#1324) Laurent Mazare 2023-11-11 18:48:52 +00:00
  • 6071797450 Add erf. Nicolas Patry 2023-11-11 18:22:16 +01:00
  • b58b247323 Putting back f16 index select. Nicolas Patry 2023-11-11 17:14:06 +01:00
  • 3900091e75 All tests are panicking instead of random failure. Nicolas Patry 2023-11-11 17:06:35 +01:00
  • 54355ff997 Adding some half kernels. Nicolas Patry 2023-11-11 13:30:21 +01:00
  • e02f1912bb Reusing a single buffer (for now) to speed things up. Nicolas Patry 2023-11-11 12:50:25 +01:00
  • a52b71686b Going back on remote metal-rs. Nicolas Patry 2023-11-11 12:05:14 +01:00
  • 7adfb70dff Few fixes. Nicolas Patry 2023-11-11 02:06:48 +01:00
  • 3ad02147e4 Starting to fix some tests. Nicolas Patry 2023-11-11 01:02:15 +01:00
  • 4f39695465 Missing new test. Nicolas Patry 2023-11-11 17:42:53 +01:00
  • 4cf4844c9d Adding the test scaffolding. Nicolas Patry 2023-11-10 23:31:09 +01:00
  • d840838e95 Cleanup fixed a few ops removed debugging scaffolding. Nicolas Patry 2023-11-10 23:00:32 +01:00
  • 61a070fdd1 Debugging rope. Nicolas Patry 2023-11-10 20:25:24 +01:00
  • e35669647d Fixed matmul (display still broken without casting back to CPU first? ) Nicolas Patry 2023-11-10 20:09:25 +01:00
  • 53e8b7ee3e Tmp state. Nicolas Patry 2023-11-10 15:35:46 +01:00
  • cc26cce23c Fixing the kernels + launches to make them faster. Nicolas Patry 2023-11-10 11:14:51 +01:00
  • 02c2ec2c71 Adding indexing. Nicolas Patry 2023-11-10 01:58:51 +01:00
  • 9a2784b8ab Refactor to simplify our lives for settings the params in the encoder. Nicolas Patry 2023-11-10 01:24:49 +01:00
  • 0f652f0e3d Adding the actual backend Nicolas Patry 2023-11-09 19:53:14 +01:00
  • ddee9dc1dd Remove tracing. Nicolas Patry 2023-11-09 19:41:08 +01:00
  • fc9bb7784a Metal part 1 - Scaffolding for metal. Nicolas Patry 2023-11-09 19:30:59 +01:00
  • f1e678b39c Mention the Yi-6b/Yi-34b models in the readme. (#1321) Laurent Mazare 2023-11-11 12:39:11 +01:00
  • a007f8fdb4 Add the Yi-6b and Yi-34b models. (#1320) Laurent Mazare 2023-11-11 12:00:48 +01:00
  • 2341aa079e Fix quantized zephyr chat prompt (#1314) (#1317) Michael Leandersson 2023-11-11 09:14:12 +01:00
  • 9e666d4229 Add the var method. (#1315) Laurent Mazare 2023-11-10 22:47:57 +01:00
  • d9c1f7e201 Fixed matmul (display still broken without casting back to CPU first? ) metal2-tmp Nicolas Patry 2023-11-10 20:09:25 +01:00
  • 315ba4cf0c Tmp state. Nicolas Patry 2023-11-10 15:35:46 +01:00
  • 1b12142a02 Add min to buckets in relative_position_bucket (#1312) Andy Braga 2023-11-10 13:57:25 +03:00
  • 915f0e5b69 Fixing the kernels + launches to make them faster. Nicolas Patry 2023-11-10 11:14:51 +01:00
  • d2c3f14773 Fix for flash-attn. (#1310) Laurent Mazare 2023-11-10 10:27:27 +01:00
  • 26c4e5bf1d Metal part 1 - Scaffolding for metal. (#1308) Nicolas Patry 2023-11-10 08:35:48 +01:00
  • 9975f2b239 Adding indexing. Nicolas Patry 2023-11-10 01:58:51 +01:00
  • d7cc660c68 Refactor to simplify our lives for settings the params in the encoder. Nicolas Patry 2023-11-10 01:24:49 +01:00
  • c54ed0ab48 Adding the actual backend Nicolas Patry 2023-11-09 19:53:14 +01:00
  • af5e77f409 Remove tracing. Nicolas Patry 2023-11-09 19:41:08 +01:00
  • 8cf39d27ce Metal part 1 - Scaffolding for metal. Nicolas Patry 2023-11-09 19:30:59 +01:00
  • 18d30005c5 Add support to UL2 model family (#1300) Juarez Bochi 2023-11-09 12:55:09 -05:00
  • 6958384327 Add support for TrOCR Model (#1303) Ogundepo Odunayo 2023-11-09 12:49:17 -05:00
  • e6697471bb Add weight and bias functions to LayerNorm (#1306) jwnz 2023-11-10 00:09:01 +09:00
  • 73d02f4f57 fix: negative axis (#1296) YangNianYi 2023-11-09 06:28:21 +08:00
  • f772213e84 Fix bug introduced in madlad PR (#1298) Juarez Bochi 2023-11-08 11:55:46 -05:00