Commit Graph

  • 6ae68777aa Rope theta to support Coda Llama (#121) Awni Hannun 2023-12-15 19:51:51 -0800
  • 08e862336a
    Rope theta to support Coda Llama (#121) Awni Hannun 2023-12-15 19:51:51 -0800
  • b225439460 nit Awni Hannun 2023-12-15 19:51:27 -0800
  • 4d6a3669fa llama chat/code Awni Hannun 2023-12-15 19:33:47 -0800
  • 4656b1d5e3 rope theta for llama model Awni Hannun 2023-12-15 19:30:49 -0800
  • 376b273b2f Merge pull request #115 from ml-explore/lora_custom Awni Hannun 2023-12-15 13:54:58 -0800
  • db134d976d
    Merge pull request #115 from ml-explore/lora_custom Awni Hannun 2023-12-15 13:54:58 -0800
  • ad77bc4c3b minimum version Awni Hannun 2023-12-15 13:54:31 -0800
  • 8df211869e minimum version Awni Hannun 2023-12-15 13:54:31 -0800
  • 392b7a2f98
    translate pytorch to mx Juarez Bochi 2023-12-15 16:51:01 -0500
  • 4c88163941 Stable diffusion - check model weights shape and support int for "attention_head_dim" (#85) Pawel Kowalski 2023-12-15 22:01:02 +0100
  • fc1495abaa
    Stable diffusion - check model weights shape and support int for "attention_head_dim" (#85) Pawel Kowalski 2023-12-15 22:01:02 +0100
  • fc13e96e6c Merge pull request #116 from idoru/fix-phi-2-temp-arg Awni Hannun 2023-12-15 12:29:19 -0800
  • 86cae9ba57
    Merge pull request #116 from idoru/fix-phi-2-temp-arg Awni Hannun 2023-12-15 12:29:19 -0800
  • e709b846ff 32 GB example Awni Hannun 2023-12-15 12:20:15 -0800
  • ff0f172363 32 GB example Awni Hannun 2023-12-15 12:20:15 -0800
  • 24a6de2acb 32 GB example Awni Hannun 2023-12-15 12:18:29 -0800
  • ee2ee0f8e5 32 GB example Awni Hannun 2023-12-15 12:18:29 -0800
  • a5bacfd04f Pass along temp argument to generate() Sam Coward 2023-12-15 15:16:41 -0500
  • 877f88dfea Pass along temp argument to generate() Sam Coward 2023-12-15 15:16:41 -0500
  • f019d836ee keep base weights in fp16 Awni Hannun 2023-12-15 10:42:18 -0800
  • 8c8f9d6440 keep base weights in fp16 Awni Hannun 2023-12-15 10:42:18 -0800
  • 6dc067d30c use lower precision base weights Awni Hannun 2023-12-15 10:29:42 -0800
  • 84f02ef58b use lower precision base weights Awni Hannun 2023-12-15 10:29:42 -0800
  • 4444c7a4a5 more nits Awni Hannun 2023-12-15 10:06:14 -0800
  • d108c558fc more nits Awni Hannun 2023-12-15 10:06:14 -0800
  • c889e09773 fix readme Awni Hannun 2023-12-15 09:59:07 -0800
  • fa51553f09 fix readme Awni Hannun 2023-12-15 09:59:07 -0800
  • 9550e1de17 custom data with lora Awni Hannun 2023-12-15 09:56:10 -0800
  • 985f413f99 custom data with lora Awni Hannun 2023-12-15 09:56:10 -0800
  • 4c4317feda
    Merge branch 'ml-explore:main' into main Pawel Kowalski 2023-12-15 18:32:31 +0100
  • 7d4a41ace8 Merge pull request #112 from ml-explore/fix_mixtral Awni Hannun 2023-12-15 08:39:02 -0800
  • 126b99d8be
    Merge pull request #112 from ml-explore/fix_mixtral Awni Hannun 2023-12-15 08:39:02 -0800
  • 330f024d1c
    Move position biases to attention module Juarez Bochi 2023-12-15 11:30:17 -0500
  • d0497ddc0b
    Load decoder weights Juarez Bochi 2023-12-15 10:50:04 -0500
  • 737db11152 Merge pull request #108 from devonthomas35/phi2_eos Awni Hannun 2023-12-15 07:34:11 -0800
  • 71b40c474c
    Merge pull request #108 from devonthomas35/phi2_eos Awni Hannun 2023-12-15 07:34:11 -0800
  • 009ed0179c
    Load position bias embeddings Juarez Bochi 2023-12-15 10:16:11 -0500
  • 001b5803ce fix RoPE bug + minor updates Awni Hannun 2023-12-14 21:45:25 -0800
  • ec11763527 fix RoPE bug + minor updates Awni Hannun 2023-12-14 21:45:25 -0800
  • f6ac70c736 Refactor EOS check devonthomas35 2023-12-14 21:11:23 -0800
  • d74d9453dd
    Refactor EOS check devonthomas35 2023-12-14 21:11:23 -0800
  • 12a5597ac3 Merge pull request #107 from ml-explore/hf_mixtral Awni Hannun 2023-12-14 16:57:19 -0800
  • a3ecda22fe
    Merge pull request #107 from ml-explore/hf_mixtral Awni Hannun 2023-12-14 16:57:19 -0800
  • 7cf66dc88c format Awni Hannun 2023-12-14 16:56:50 -0800
  • b863e7cca0 format Awni Hannun 2023-12-14 16:56:50 -0800
  • f37ecda996 feat: stop phi-2 output at <|endoftext|> Cody 2023-12-14 19:37:33 -0500
  • 7f992db5bc Remove unnecessary return devonthomas35 2023-12-14 15:52:22 -0800
  • d7d7aabded
    Remove unnecessary return devonthomas35 2023-12-14 15:52:22 -0800
  • 8d496ba61a Stop generating at eos token devonthomas35 2023-12-14 15:50:59 -0800
  • 4549dcbbd0
    Stop generating at eos token devonthomas35 2023-12-14 15:50:59 -0800
  • 6249f46215 incude instruct option Awni Hannun 2023-12-14 15:40:38 -0800
  • e434e7e5c2 incude instruct option Awni Hannun 2023-12-14 15:40:38 -0800
  • 449f7a694b use official HF for mixtral Awni Hannun 2023-12-14 15:30:32 -0800
  • 078fed3d8d use official HF for mixtral Awni Hannun 2023-12-14 15:30:32 -0800
  • 95a1d50318 Merge pull request #106 from fahnub/main Awni Hannun 2023-12-14 14:15:19 -0800
  • 67a208b13e
    Merge pull request #106 from fahnub/main Awni Hannun 2023-12-14 14:15:19 -0800
  • 330e8e8bc9 minor dep fix in phi Fahad Nadeem 2023-12-15 03:09:33 +0500
  • 9b887cef08 minor dep fix in phi Fahad Nadeem 2023-12-15 03:09:33 +0500
  • 62924d8135
    Pass config to all modules, fix ln Juarez Bochi 2023-12-14 15:51:03 -0500
  • c0001a94f2
    Load all encoder weights Juarez Bochi 2023-12-14 15:38:41 -0500
  • bca5ca4f98
    Add skeleton Juarez Bochi 2023-12-14 15:21:36 -0500
  • 53e58795c2 Merge pull request #77 from SarthakYadav/main Awni Hannun 2023-12-14 12:19:40 -0800
  • 09fff84a85
    Merge pull request #77 from SarthakYadav/main Awni Hannun 2023-12-14 12:19:40 -0800
  • 016c45fb8e update format Awni Hannun 2023-12-14 12:18:44 -0800
  • e12e4d5825 typo / nits Awni Hannun 2023-12-14 12:14:01 -0800
  • b9439ce74e typo / nits Awni Hannun 2023-12-14 12:14:01 -0800
  • 5673716daa updates + format Awni Hannun 2023-12-14 12:09:10 -0800
  • b1b9b11801 updates + format Awni Hannun 2023-12-14 12:09:10 -0800
  • 4cac181917 Merge pull request #103 from arpitingle/patch-1 Awni Hannun 2023-12-14 10:19:40 -0800
  • 0e88a6afa1
    Merge pull request #103 from arpitingle/patch-1 Awni Hannun 2023-12-14 10:19:40 -0800
  • 541265b74d Update README.md arpit 2023-12-14 23:40:50 +0530
  • 5b08da2395
    Update README.md arpit 2023-12-14 23:40:50 +0530
  • f4745d8576 Merge pull request #97 from jbarrow/main Awni Hannun 2023-12-14 09:21:26 -0800
  • 92efa32060
    Merge pull request #97 from jbarrow/main Awni Hannun 2023-12-14 09:21:26 -0800
  • fa9e34b041 cleanup conversion to use single qkv matrix Awni Hannun 2023-12-14 09:19:44 -0800
  • 8f60d60814 cleanup conversion to use single qkv matrix Awni Hannun 2023-12-14 09:19:44 -0800
  • 45c1800fc6 update readme Awni Hannun 2023-12-14 08:37:34 -0800
  • 0c1c500714 update readme Awni Hannun 2023-12-14 08:37:34 -0800
  • c2eb435697 change file name for consistency, update readme. Awni Hannun 2023-12-14 08:34:24 -0800
  • 3d2a23184a change file name for consistency, update readme. Awni Hannun 2023-12-14 08:34:24 -0800
  • 5822639f23 don't drop last tokens Awni Hannun 2023-12-14 08:27:44 -0800
  • 840c0c36c2 don't drop last tokens Awni Hannun 2023-12-14 08:27:44 -0800
  • c26eafc125 fix args, update README, remove extra files Awni Hannun 2023-12-14 08:18:01 -0800
  • 1613e608a9 fix args, update README, remove extra files Awni Hannun 2023-12-14 08:18:01 -0800
  • 05c82ddf5f fix fp16 + nits Awni Hannun 2023-12-14 08:08:28 -0800
  • a8d4149147 fix fp16 + nits Awni Hannun 2023-12-14 08:08:28 -0800
  • 879a576fb6 updated header Sarthak Yadav 2023-12-14 16:28:00 +0100
  • 29b7a97342 updated header Sarthak Yadav 2023-12-14 16:28:00 +0100
  • bb44222a86 Merge pull request #98 from finnless/patch-1 Awni Hannun 2023-12-14 07:13:19 -0800
  • b11997122d
    Merge pull request #98 from finnless/patch-1 Awni Hannun 2023-12-14 07:13:19 -0800
  • a2aadb24bd Merge pull request #102 from burakbudanur/main Awni Hannun 2023-12-14 07:12:20 -0800
  • 363108d7b3
    Merge pull request #102 from burakbudanur/main Awni Hannun 2023-12-14 07:12:20 -0800
  • f603d53bef Corrected the typo in 'ffn_dim_multiplier' in and added 'rope_theta' to the list unused. Without these, llama examples did not run. Burak Budanur 2023-12-14 14:02:11 +0100
  • f691e00e5a Corrected the typo in 'ffn_dim_multiplier' in and added 'rope_theta' to the list unused. Without these, llama examples did not run. Burak Budanur 2023-12-14 14:02:11 +0100
  • 7deeff35bc Reverting to a single model Pawel Kowalski 2023-12-14 09:57:55 +0100
  • a3c0343b31 simplified ResNet, expanded README with throughput and performance Sarthak Yadav 2023-12-14 09:05:04 +0100
  • 15a6c155a8 simplified ResNet, expanded README with throughput and performance Sarthak Yadav 2023-12-14 09:05:04 +0100
  • 0301cbd88b add cache + generation, clean up some stuff Awni Hannun 2023-12-13 22:26:33 -0800
  • 88d7b67e6e add cache + generation, clean up some stuff Awni Hannun 2023-12-13 22:26:33 -0800