Commit Graph

  • a1a5ab8b0a Neon optimized vecdot (#666) Laurent Mazare 2023-08-29 22:28:46 +01:00
  • 59b731de99 Add the powf op. (#664) Laurent Mazare 2023-08-29 20:48:18 +01:00
  • 2d3fcad267 Simplify usage of the pool functions. (#662) Laurent Mazare 2023-08-29 19:12:16 +01:00
  • b31d41e26a Add a convnet training example. (#661) Laurent Mazare 2023-08-29 18:23:01 +01:00
  • 71221559d3 Fix the dilated convolutions. (#659) Laurent Mazare 2023-08-29 16:37:42 +01:00
  • a044907ffc Dilated convolutions (#657) Laurent Mazare 2023-08-29 16:12:11 +01:00
  • 6bd8c2d34b Adding benchmark. remove_wrapper_bench Nicolas Patry 2023-08-29 17:01:40 +02:00
  • ee8bb1bde1 Add avx implemenetations of q2k, q3k and q5k vec-dot functions (#654) Lukas Kreussel 2023-08-29 14:35:56 +02:00
  • 3d2d3c7edb Merge pull request #658 from huggingface/upgrade_hf_hub2 Nicolas Patry 2023-08-29 14:32:15 +02:00
  • 1aca6fa291 Upgrading hf-hub. Nicolas Patry 2023-08-29 14:18:54 +02:00
  • 4ed202447e Upgrading hf-hub. Nicolas Patry 2023-08-29 14:14:26 +02:00
  • 1d6bff53fc Changelog update. (#656) Laurent Mazare 2023-08-29 12:55:56 +01:00
  • 14b4d456e8 Merge pull request #439 from huggingface/training_hub_dataset Nicolas Patry 2023-08-29 13:10:05 +02:00
  • 2d5b7a735d Update the book with new layout of datasets. Nicolas Patry 2023-08-29 12:51:59 +02:00
  • 62ef494dc1 Use multiple transformer layer in the same cross-attn blocks. (#653) Laurent Mazare 2023-08-29 11:13:43 +01:00
  • d0a330448d Backprop support for pooling ops. (#652) Laurent Mazare 2023-08-29 10:17:59 +01:00
  • 4b8d57ba15 AVX version of the q4k vecdot. (#651) Laurent Mazare 2023-08-29 09:41:17 +01:00
  • d5a525f7a7 Fix clippy + save_image. Nicolas Patry 2023-08-29 10:19:44 +02:00
  • 33c23c19b6 Preliminary support for SDXL. (#647) Laurent Mazare 2023-08-29 09:00:04 +01:00
  • 49326fb925 Update .gitignore (#649) Lei 2023-08-29 03:41:33 -04:00
  • fd3131a4ce Fix the debug implementation. (#648) Laurent Mazare 2023-08-28 22:51:39 +01:00
  • 037b41c9dc Cuda conv transpose (#645) Laurent Mazare 2023-08-28 20:58:49 +01:00
  • 72fae3140c Optimize the conv2d transpose cpu kernel. (#644) Laurent Mazare 2023-08-28 20:06:31 +01:00
  • ca26198b95 Fix the cpu kernel for conv-transpose. (#643) Laurent Mazare 2023-08-28 16:45:12 +01:00
  • b292047882 Backprop for conv2d. (#638) Laurent Mazare 2023-08-28 16:08:55 +01:00
  • 09c5bd1881 Rebased Nicolas Patry 2023-08-28 15:47:03 +02:00
  • fe6c88713d Fix waiting upgrade for SSL ? Nicolas Patry 2023-08-16 17:17:18 +02:00
  • 6f3f9285e6 Remove image dep. Nicolas Patry 2023-08-15 19:02:32 +02:00
  • baca3cf69d Fix deps. Nicolas Patry 2023-08-15 18:58:58 +02:00
  • d726484a6d Re-enable local dir for mnist. Nicolas Patry 2023-08-15 15:52:37 +02:00
  • dd06d93d0b Cleanup: Nicolas Patry 2023-08-15 15:42:22 +02:00
  • c109c93db7 Update candle-book/src/SUMMARY.md Nicolas Patry 2023-08-14 18:38:39 +02:00
  • d7a273be51 Training: Nicolas Patry 2023-08-14 17:23:08 +02:00
  • dd02f589c0 Better training+hub Nicolas Patry 2023-08-14 11:09:48 +02:00
  • 7602323667 [Book] Add small error management + start training (with generic dataset inclusion). Nicolas Patry 2023-08-14 10:52:12 +02:00
  • 9137c63175 Update README.md (#640) Laurent Mazare 2023-08-28 11:34:54 +01:00
  • 3cca89cc70 Add conv-transpose. (#635) Laurent Mazare 2023-08-28 10:10:12 +01:00
  • 26e1b40992 Repeat-penalty in the falcon example. (#634) Laurent Mazare 2023-08-28 08:13:40 +01:00
  • 1da71a5da1 Neon optimized version of the q4k vecdot product. (#632) Laurent Mazare 2023-08-27 21:30:47 +01:00
  • 24dda44c27 Add wasm support for yolo-v8 pose detection. (#630) Laurent Mazare 2023-08-27 19:49:24 +01:00
  • 72ebb12bca Remove some dead-code annotations. (#629) Laurent Mazare 2023-08-27 18:52:33 +01:00
  • a3f97c143d Bump the crate version + update CHANGELOG. (#628) Laurent Mazare 2023-08-27 18:17:11 +01:00
  • 4c338b0cd9 VarBuilder cleanup (#627) Laurent Mazare 2023-08-27 18:03:26 +01:00
  • be471d50ab Llama quantization. (#625) Laurent Mazare 2023-08-27 14:08:15 +01:00
  • 7151f2cf63 Add the quantize command. (#624) Laurent Mazare 2023-08-27 11:35:19 +01:00
  • 6e485f2deb Add some optional repeat penalty. (#623) Laurent Mazare 2023-08-27 10:48:45 +01:00
  • 5320aa6b7d Move the test-utils bits to a shared place. (#619) Laurent Mazare 2023-08-27 09:42:22 +01:00
  • a8b39dd7b7 Fix for q5_1 quantization. (#617) Laurent Mazare 2023-08-27 08:31:18 +01:00
  • fa0d75b18d Quantization tests + fix some issues. (#616) Laurent Mazare 2023-08-27 08:17:38 +01:00
  • 28658054ff More missing quantized bits. (#615) Laurent Mazare 2023-08-27 07:52:26 +01:00
  • ab36a7f3e3 Fix for when f16c is not available. (#614) Laurent Mazare 2023-08-27 07:19:52 +01:00
  • f704e39761 Missing quants ops (#611) Laurent Mazare 2023-08-26 20:09:04 +01:00
  • fdf15f0e05 Another transmute tweak. (#610) Laurent Mazare 2023-08-26 13:00:24 +01:00
  • 06b37ea7ad Avoid using tmp values. (#609) Laurent Mazare 2023-08-26 12:28:28 +01:00
  • c72eb3d75b Add reference implementation for q4k and q5k (#586) Lukas Kreussel 2023-08-26 13:07:54 +02:00
  • 864227edbf [WIP] Improve Yolo WASM UI example (#591) Radamés Ajna 2023-08-26 03:40:41 -07:00
  • b23b347b35 Merge pull request #601 from huggingface/repair_bf16_f16_cast Nicolas Patry 2023-08-26 12:34:41 +02:00
  • 71518caeee Align tensor device print more with PyTorch (#590) Patrick von Platen 2023-08-26 12:20:22 +02:00
  • 6559eae72c Avoid some transmutes. (#607) Laurent Mazare 2023-08-25 18:21:37 +01:00
  • 46eb225ba5 Add some missing entries to the changelog. (#606) Laurent Mazare 2023-08-25 18:01:38 +01:00
  • aa67e5107d Merge pull request #600 from huggingface/codellama_gpu_support Nicolas Patry 2023-08-25 18:25:26 +02:00
  • c105550405 s/panic/bail/ Nicolas Patry 2023-08-25 18:05:07 +02:00
  • ca6c050b04 Cleanup the pose reporting code. (#605) Laurent Mazare 2023-08-25 16:49:21 +01:00
  • 9c8d6dbc2a Neon intrinsics for the q8_0 vecdot. (#604) Laurent Mazare 2023-08-25 14:42:18 +01:00
  • 0afbc435df Add some configurable legend for yolo detection. (#603) Laurent Mazare 2023-08-25 13:50:31 +01:00
  • d4e75d5825 Let's keep the dirty code on its own. Nicolas Patry 2023-08-25 12:01:12 +00:00
  • be371e827c Intermediary float cast is necessary for cuda 11.8 Nicolas Patry 2023-08-25 11:54:30 +00:00
  • 97909e5068 Move the yolo model bits in a separate file. (#602) Laurent Mazare 2023-08-25 12:47:55 +01:00
  • 1c1e34735e static_cast ? Nicolas Patry 2023-08-25 11:40:36 +00:00
  • db8bab8b7a Different casting ? Nicolas Patry 2023-08-25 10:49:22 +00:00
  • bc131b402b Repairing cast bf16/f16 Nicolas Patry 2023-08-25 10:38:19 +00:00
  • 8bc5fffa45 More support for pose estimation in yolo-v8. (#599) Laurent Mazare 2023-08-25 11:21:11 +01:00
  • 4826a4212e Adding support for codellama in examples. Nicolas Patry 2023-08-25 09:56:11 +00:00
  • afc10a3232 AVX version for the q8-0 multiplications. (#598) Laurent Mazare 2023-08-25 10:14:49 +01:00
  • d728e646c2 Use resolver 2 explicitely. (#597) Laurent Mazare 2023-08-25 09:35:40 +01:00
  • c093b03d51 Generic implementation of vecdot for q80. (#596) Laurent Mazare 2023-08-25 09:04:05 +01:00
  • d8ba0452dc Fail on bf16. (#594) Laurent Mazare 2023-08-25 06:10:38 +01:00
  • 189442a0fa Add the pose estimation head for yolo. (#589) Laurent Mazare 2023-08-24 22:12:34 +01:00
  • 2cde0cb74b More pickle support. (#588) Laurent Mazare 2023-08-24 18:45:10 +01:00
  • e21c686cdc Fixes for clippy 1.72. (#587) Laurent Mazare 2023-08-24 17:46:17 +01:00
  • c265ac50fa Add a function to write gguf files. (#585) Laurent Mazare 2023-08-24 17:03:06 +01:00
  • a87c6f7652 Merge pull request #561 from patrickvonplaten/add_installation Nicolas Patry 2023-08-24 16:25:52 +02:00
  • afd965f77c More non square testing (#582) Laurent Mazare 2023-08-24 13:01:04 +01:00
  • d2f42ab086 Referenze implementations of q2k and q3k vec-dot functions (#580) Lukas Kreussel 2023-08-24 13:35:54 +02:00
  • ca318a6ec7 Add to the cuda example a reproduction of the issue. (#579) Laurent Mazare 2023-08-24 12:07:31 +01:00
  • dd64465899 Add a test for conv2d with padding + bugfix the random number generation on cuda. (#578) Laurent Mazare 2023-08-24 10:16:37 +01:00
  • 79916c2edb Use the hub weights for efficientnet. (#573) Laurent Mazare 2023-08-23 18:20:21 +01:00
  • 431051cc32 Add Efficientnet (#572) Laurent Mazare 2023-08-23 18:02:58 +01:00
  • eedd85ffa7 Move the imagenet specific bits to a separate file. (#571) Laurent Mazare 2023-08-23 16:42:09 +01:00
  • 7478dda255 Cosmetic tweaks. (#570) Laurent Mazare 2023-08-23 15:45:40 +01:00
  • 329f661d9b Trace softmax (#568) Laurent Mazare 2023-08-23 15:25:50 +01:00
  • 075b505480 Mirror GGML's unit tests (#569) Lukas Kreussel 2023-08-23 16:25:17 +02:00
  • aba1e90797 Add some group parameter to convolutions. (#566) Laurent Mazare 2023-08-23 12:58:55 +01:00
  • 1f58bdbb1d Apply suggestions from code review Patrick von Platen 2023-08-23 13:33:45 +02:00
  • c98d3cfd8b Update candle-book/src/guide/installation.md Patrick von Platen 2023-08-23 13:31:54 +02:00
  • c5e43ad0ab Apply suggestions from code review Patrick von Platen 2023-08-23 13:27:29 +02:00
  • 2c280007e8 Apply suggestions from code review Patrick von Platen 2023-08-23 13:26:21 +02:00
  • 4ee1cf038a Get the rms epsilon from GGUF. (#565) Laurent Mazare 2023-08-23 11:40:20 +01:00
  • 0f4ff8a739 Fix the quantized example. (#564) Laurent Mazare 2023-08-23 11:09:55 +01:00
  • 89a00b56cc add chat models in quantized example (#551) cksac 2023-08-23 18:05:33 +08:00