Commit Graph

  • ab1d40ea97 Add more t5 tracing. (#915) Laurent Mazare 2023-09-20 20:20:54 +01:00
  • 3a0d3e05df Add more t5 tracing. (#914) Laurent Mazare 2023-09-20 16:37:51 +01:00
  • 9b24d89d2d Tracing mode for T5. (#913) Laurent Mazare 2023-09-20 15:03:35 +01:00
  • fb1c2ac535 Add flash-attn support. (#912) Laurent Mazare 2023-09-20 14:07:55 +01:00
  • 728e167334 Add details on wuerstchen. (#911) Laurent Mazare 2023-09-20 13:09:35 +01:00
  • 7b1ddcff47 Add clone to various nn layers. (#910) Laurent Mazare 2023-09-20 11:33:51 +01:00
  • f685b2231c Add some missing biases. (#908) Laurent Mazare 2023-09-20 10:14:51 +01:00
  • c0b49d5a50 Wuerstchen parameter tweaks. (#907) Laurent Mazare 2023-09-20 09:26:24 +01:00
  • 098dd0d1e9 fix: add missingtop_p in llama_multiprocess (#905) Mahmoud 2023-09-20 00:54:56 -07:00
  • 05626ef492 Flan T5: Read lm_head when word embeddings are not tied (#903) Juarez Bochi 2023-09-19 14:36:47 -07:00
  • 67a486d18d Line-up the wuerstchen model with the python implementation. (#901) Laurent Mazare 2023-09-19 21:59:44 +01:00
  • 7ad82b87e4 BERT Wasm (#902) Radamés Ajna 2023-09-19 13:31:37 -07:00
  • 8696f64bae Fix T5 kv cache (#899) Juarez Bochi 2023-09-19 12:36:15 -07:00
  • d7e48234d4 Add an erf based gelu op (#900) Laurent Mazare 2023-09-19 19:54:28 +01:00
  • 34f2ecbc3b Fix the leaky relu. (#898) Laurent Mazare 2023-09-19 18:17:17 +01:00
  • 4f91c8e109 Improve the error message on shape mismatch for cat. (#897) Laurent Mazare 2023-09-19 15:09:47 +01:00
  • 06e46d7c3b Only use classifier free guidance for the prior. (#896) Laurent Mazare 2023-09-19 14:13:05 +01:00
  • 4114872aae Make things contiguous. w-uncond laurent 2023-09-19 14:04:56 +01:00
  • f2a648f313 Remove some debugging continue. laurent 2023-09-19 13:43:41 +01:00
  • ec895453cd More shape fixes. laurent 2023-09-19 13:43:19 +01:00
  • 3769d8bf71 Print the prior shape. laurent 2023-09-19 10:26:18 +01:00
  • 5d8e214dfe Fix the latent shape. laurent 2023-09-19 09:21:35 +01:00
  • 576bf7c21f Tweaks. laurent 2023-09-19 09:08:32 +01:00
  • 49a4fa44bb Add another specific layer-norm structure. laurent 2023-09-19 09:06:10 +01:00
  • b936e32e11 Only use classifier free guidance for the prior. laurent 2023-09-19 08:40:02 +01:00
  • 9cf26c5cff Fix typo in error_manage.md (#888) Ikko Eltociear Ashimine 2023-09-19 15:14:15 +09:00
  • aaa9d4ed6c W decoding. (#893) Laurent Mazare 2023-09-19 07:13:44 +01:00
  • 92db8cecd3 Specialized attention module for Wuerstchen. (#890) Laurent Mazare 2023-09-18 21:16:09 +01:00
  • 1542e92629 T5: Add option to override use_cache from config (#892) Juarez Bochi 2023-09-18 12:20:21 -07:00
  • 82a98f6da0 Prior denoising. (#889) Laurent Mazare 2023-09-18 16:51:38 +01:00
  • 5082954c52 Fix the W clip embeddings. (#887) Laurent Mazare 2023-09-18 14:50:14 +01:00
  • 7dd8e12472 Bump the crate versions to v0.2.3. (#886) Laurent Mazare 2023-09-18 12:14:03 +01:00
  • 12696b7b2d Fix typos in SAM WASM example (#884) Aaron Sun 2023-09-18 16:41:50 +08:00
  • ef8cd8fea0 Update the candle-gemm version. (#885) Laurent Mazare 2023-09-18 09:36:20 +01:00
  • 03e194123d Add return types to *.pyi stubs (#880) Lukas Kreussel 2023-09-17 23:11:01 +02:00
  • c2b866172a More Wuerstchen fixes. (#882) Laurent Mazare 2023-09-17 22:08:11 +01:00
  • 06cc329e71 Remove the parameters for the Wuerstchen layer-norm. (#879) Laurent Mazare 2023-09-17 15:59:27 +01:00
  • 5f83c13f17 Add the DDPM scheduler. (#877) Laurent Mazare 2023-09-17 15:03:01 +01:00
  • db3e9dae04 Wuerstchen main (#876) Laurent Mazare 2023-09-17 13:46:38 +02:00
  • 7f65af1f0d Avoid re-encoding the input in the T5 example. (#875) Laurent Mazare 2023-09-17 11:25:54 +02:00
  • eeb54716dd Tweaks for the T5 example. (#874) Laurent Mazare 2023-09-17 11:05:15 +02:00
  • 1a276b5da7 Add a KV cache to T5. (#873) Laurent Mazare 2023-09-17 09:00:45 +02:00
  • 8658df3485 Generate *.pyi stubs for PyO3 wrapper (#870) Lukas Kreussel 2023-09-16 18:23:38 +02:00
  • 7cafca835a readme tweaks. (#867) Laurent Mazare 2023-09-16 08:22:24 +02:00
  • 04ca2b9ebd Update README + SAM (#866) Radamés Ajna 2023-09-15 22:34:13 -07:00
  • 635012d770 Do not backprop through argmin/argmax. (#865) Laurent Mazare 2023-09-15 23:15:40 +02:00
  • 3e49f8fce5 Implement T5 decoding (#864) Juarez Bochi 2023-09-15 13:05:12 -07:00
  • c2007ac88f W fixes. (#862) Laurent Mazare 2023-09-15 16:11:11 +02:00
  • 30be5b6660 Replication pad (#861) Laurent Mazare 2023-09-15 15:06:21 +02:00
  • 107d3d9530 Add the embed mapper convolutions. (#860) Laurent Mazare 2023-09-15 11:38:38 +02:00
  • 2746f2c4be DiffNeXt/unet (#859) Laurent Mazare 2023-09-15 11:14:02 +02:00
  • 81a36b8713 Add link error info (#851) Ssslakter 2023-09-15 13:25:10 +07:00
  • 0633c85514 Add leaky-relu in the activation enum. (#858) Laurent Mazare 2023-09-15 08:05:38 +02:00
  • 39157346cb Add SAM UI Demo (#854) Radamés Ajna 2023-09-14 22:31:58 -07:00
  • 5cefbba757 minor UI fixes (#856) Radamés Ajna 2023-09-14 22:30:50 -07:00
  • 130fe5a087 Add the upblocks. (#853) Laurent Mazare 2023-09-14 23:24:56 +02:00
  • 91ec546feb More DiffNeXt. (#847) Laurent Mazare 2023-09-14 22:16:31 +02:00
  • 0a647875ec Use softmax-last-dim in the quantized example. (#848) Laurent Mazare 2023-09-14 18:29:24 +02:00
  • a0c6d5548c Add the attention block. (#846) Laurent Mazare 2023-09-14 16:40:09 +02:00
  • 286f01db14 Start adding the Wuerstchen diffusion pipeline (#843) Laurent Mazare 2023-09-14 11:56:07 +02:00
  • d6447ad635 Tensor based indexing. (#842) Laurent Mazare 2023-09-14 08:47:07 +02:00
  • 49d3f7f708 Add support to flan-t5 (#840) Juarez Bochi 2023-09-13 10:27:20 -07:00
  • 9a465e1b26 Add 1d upsampling. (#839) Laurent Mazare 2023-09-13 17:50:39 +02:00
  • 31ab2ddaeb Remove the padding. (#838) Laurent Mazare 2023-09-13 14:00:59 +02:00
  • b11a2a7b9d Move the constant to avoid some unused warning. (#837) Laurent Mazare 2023-09-13 12:56:53 +02:00
  • 1c09164021 Add CANDLE_NVCC_CCBIN support for candle-kernels, and eliminate warning. (#836) Charles Lew 2023-09-13 18:39:22 +08:00
  • 3e94324012 Add some sentence similarity part to the t5 example. (#835) Laurent Mazare 2023-09-13 11:44:02 +02:00
  • e6f040d6e3 Readme gallery (#834) Laurent Mazare 2023-09-13 10:05:47 +02:00
  • cbd36157ac Add a gif to the quantized readme. (#833) Laurent Mazare 2023-09-13 09:43:52 +02:00
  • 18d3c803a8 Scalar support in minimum/maximum. (#832) Laurent Mazare 2023-09-13 09:24:58 +02:00
  • e4553fb355 T5 tweaks (#831) Laurent Mazare 2023-09-13 08:37:04 +02:00
  • d801e1d564 Clippy fix. (#830) Laurent Mazare 2023-09-13 08:16:20 +02:00
  • 9daa6dbe87 Extract T5 module and add main function to use it (#829) Juarez Bochi 2023-09-12 23:14:05 -07:00
  • e82fcf1c59 Add more example readmes. (#828) Laurent Mazare 2023-09-12 18:21:24 +02:00
  • 805bf9ffa7 Implement top_p / nucleus sampling (#819) Juarez Bochi 2023-09-12 09:10:16 -07:00
  • 42da17694a Segment Anything readme (#827) Laurent Mazare 2023-09-12 14:35:55 +01:00
  • 25aacda28e Add useful libraries section (#825) Eric Buehler 2023-09-12 06:06:21 -04:00
  • 7a62aad24a Add a readme for yolo-v8. (#824) Laurent Mazare 2023-09-12 11:01:06 +01:00
  • bb23b90b1d Add a small readme for the quantized example. (#823) Laurent Mazare 2023-09-12 10:17:31 +01:00
  • 2257f4d475 Bump the crate version + update the changelog. (#822) Laurent Mazare 2023-09-12 06:39:24 +01:00
  • 871efc0307 Bugfix for the conv2d cpu kernel. (#820) Laurent Mazare 2023-09-11 23:11:27 +01:00
  • c5a058b169 Use the module trait in stable-diffusion. (#817) Laurent Mazare 2023-09-11 20:40:07 +01:00
  • 59e63d690c Add weight, bias, and hidden_size methods (#816) Eric Buehler 2023-09-11 11:01:11 -04:00
  • dbd4561416 im2col version of the conv1d kernel. (#815) Laurent Mazare 2023-09-11 14:40:09 +01:00
  • 5c35fbbb13 Stable-Diffusion readme (#814) Laurent Mazare 2023-09-11 13:06:29 +01:00
  • 70f38c2069 Proper error on unsupported dtypes when using gemm. (#813) Laurent Mazare 2023-09-11 12:10:51 +01:00
  • d7b9fec849 Move the stable-diffusion modeling code so that it's easier to re-use. (#812) Laurent Mazare 2023-09-11 11:45:57 +01:00
  • 84ee870efd Use softmax-last-dim in whisper. (#810) Laurent Mazare 2023-09-11 11:05:05 +01:00
  • df712ecf64 Handle the case where the kernel is not contiguous in the cuda backend. (#809) Laurent Mazare 2023-09-11 09:48:31 +01:00
  • 6fb665004c Enable im2col on the cpu side. (#805) Laurent Mazare 2023-09-11 09:28:13 +01:00
  • 1cd74129d4 Add Im2Col support on the gpu side. (#808) Laurent Mazare 2023-09-11 08:52:33 +01:00
  • 98d1242b8f im2col based conv2d (#802) Laurent Mazare 2023-09-10 21:02:42 +01:00
  • 18d6db2180 more doc fixes (#804) Ssslakter 2023-09-11 02:36:29 +07:00
  • 4f18180fc7 Bugfix so that im2col produce the same results as conv2d. (#801) Laurent Mazare 2023-09-10 16:59:46 +01:00
  • 559944146f Add an im2col based benchmark. (#800) Laurent Mazare 2023-09-10 16:56:28 +01:00
  • 3dd5804299 Fix typo in readme. (#799) Laurent Mazare 2023-09-10 13:49:47 +01:00
  • 90e077e409 Return the low res mask in the wasm segment-anything module. (#798) Laurent Mazare 2023-09-10 13:03:02 +01:00
  • 584171cae1 Add a wasm module for the segment anything example. (#797) Laurent Mazare 2023-09-10 12:29:37 +01:00
  • 6c58fc59fd Little docs changes (#791) Ssslakter 2023-09-10 18:02:52 +07:00
  • 35f72514f5 Move more models to candle-transformers (#796) Laurent Mazare 2023-09-10 10:20:18 +01:00