Commit Graph

  • b3c80a841c Built site for gh-pages Quarto GHA Workflow Runner 2024-06-10 00:11:06 +00:00
  • 65b85a1241 Built site for gh-pages Quarto GHA Workflow Runner 2024-06-10 00:10:37 +00:00
  • 5783839c6e download model weights on preprocess step (#1693) Wing Lian 2024-06-09 20:10:17 -04:00
  • cbbf039a46 verbose failure message (#1694) Wing Lian 2024-06-09 20:09:36 -04:00
  • a278084113 Built site for gh-pages Quarto GHA Workflow Runner 2024-06-09 21:14:20 +00:00
  • 851ccb1237 bump deepspeed for fix for grad norm compute putting tensors on different devices (#1699) Wing Lian 2024-06-09 17:13:28 -04:00
  • a42d3af60f Built site for gh-pages Quarto GHA Workflow Runner 2024-06-08 13:49:18 +00:00
  • a0ba3c44bf Built site for gh-pages Quarto GHA Workflow Runner 2024-06-08 13:49:02 +00:00
  • 18cabc0c46 fix for when sample_packing and eval_sample_packing are different (#1695) Wing Lian 2024-06-08 09:48:30 -04:00
  • ed8ef65371 add back packing efficiency estimate so epochs and multi-gpu works properly (#1697) Wing Lian 2024-06-08 09:48:10 -04:00
  • 949859414b Built site for gh-pages Quarto GHA Workflow Runner 2024-06-07 20:39:18 +00:00
  • 00ac3022a1 add qwen2-72b fsdp example (#1696) Wing Lian 2024-06-07 16:38:29 -04:00
  • 2fae3af066 Built site for gh-pages Quarto GHA Workflow Runner 2024-06-07 15:29:37 +00:00
  • 9c1af1a9c0 ensure explicit eval_sample_packing to avoid mismatch issues (#1692) Wing Lian 2024-06-07 11:28:43 -04:00
  • a1e5d790de Built site for gh-pages Quarto GHA Workflow Runner 2024-06-04 20:22:06 +00:00
  • a82a711522 Create phi3-ft-fsdp.yml (#1580) Aaditya Ura (looking for PhD Fall’24) 2024-06-05 01:50:25 +05:30
  • 16ebc4cf88 Built site for gh-pages Quarto GHA Workflow Runner 2024-06-04 20:12:47 +00:00
  • 07ede188ad Built site for gh-pages Quarto GHA Workflow Runner 2024-06-04 20:11:57 +00:00
  • cf64284a04 Phi-3 conversation format, example training script and perplexity metric (#1582) Brian Fitzgerald 2024-06-04 15:11:56 -05:00
  • c996881ec2 add support for rpo_alpha (#1681) Wing Lian 2024-06-04 16:09:51 -04:00
  • 62ee488ae3 Built site for gh-pages Quarto GHA Workflow Runner 2024-06-03 16:51:39 +00:00
  • 1f151c0d52 re-enable DPO for tests in modal ci (#1374) Wing Lian 2024-06-03 12:50:44 -04:00
  • 5cde06587a Fix the broken link in README (#1678) [skip ci] Saeed Esmaili 2024-06-03 15:38:44 +02:00
  • d7ec10e337 add support for MoRA mora Wing Lian 2024-06-01 16:14:56 -04:00
  • 2d1ff23076 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-31 17:14:03 +00:00
  • 05b0bd08d2 need to add back drop_last for sampler (#1676) Wing Lian 2024-05-31 13:13:13 -04:00
  • 3d6a5e473d Built site for gh-pages Quarto GHA Workflow Runner 2024-05-30 17:41:36 +00:00
  • d4f6c65e4c cleanup the deepspeed proxy model at the end of training (#1675) Wing Lian 2024-05-30 13:40:35 -04:00
  • 296955d7e2 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-30 02:28:55 +00:00
  • 1ade2ca48c Built site for gh-pages Quarto GHA Workflow Runner 2024-05-30 02:28:26 +00:00
  • a944f7b32b load explicit splits on datasets (#1652) Wing Lian 2024-05-29 22:27:59 -04:00
  • 4ae5739b4a Built site for gh-pages Quarto GHA Workflow Runner 2024-05-30 02:27:59 +00:00
  • 9d4225a058 set chat_template in datasets config automatically (#1664) Wing Lian 2024-05-29 22:27:26 -04:00
  • f7332ac449 use mixins for orpo and kto configs so they work with axolotl customizations (#1674) Wing Lian 2024-05-29 22:27:00 -04:00
  • b41672fd38 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-29 19:42:37 +00:00
  • 16d46b74e4 re-enable phi for tests in modal ci (#1373) Wing Lian 2024-05-29 15:41:46 -04:00
  • 12ea07c470 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-29 15:52:08 +00:00
  • a6b37bdeb4 revert multipack batch sampler changes (#1672) Wing Lian 2024-05-29 11:51:18 -04:00
  • 87457a1a22 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-29 14:22:03 +00:00
  • b7520801a3 handle the system role too for chat templates (#1671) Wing Lian 2024-05-29 10:21:11 -04:00
  • 53febeb970 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-29 14:13:04 +00:00
  • fe650dd326 make sure the CI fails when pytest script fails (#1669) Wing Lian 2024-05-29 10:12:11 -04:00
  • 75ededd815 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-28 22:11:48 +00:00
  • ddb2f39a34 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-28 22:11:28 +00:00
  • 49b967b62f Fix README quick start example usage model dirs (#1668) Abe Voelker 2024-05-28 17:10:40 -05:00
  • 65db903714 Correct name of MixtralBlockSparseTop2MLP (L -> l) (#1667) Seungduk Kim 2024-05-29 07:10:29 +09:00
  • 667c8b0ccf Built site for gh-pages Quarto GHA Workflow Runner 2024-05-28 16:01:24 +00:00
  • 6a5a725f10 Fix: ensure correct handling of val_set_size as float or int (#1655) Davide Caroselli 2024-05-28 18:00:32 +02:00
  • 05c9b6aee8 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-28 15:37:55 +00:00
  • f5febc729a fix lint issue that snuck through (#1665) Wing Lian 2024-05-28 11:36:50 -04:00
  • c6437eae72 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-28 15:26:15 +00:00
  • 230e0ac363 Fix Lora config error for Llama3 (#1659) Faria Huq 2024-05-28 11:25:08 -04:00
  • cc11c6bce2 Generalizing the chat_template prompt strategy (#1660) [skip ci] Keith Stevens 2024-05-29 00:24:13 +09:00
  • 5f91064040 Fix Google Colab notebook 2024-05 (#1662) [skip ci] Maciek 2024-05-28 17:23:52 +02:00
  • ef223519c9 update deps (#1663) [skip ci] Wing Lian 2024-05-28 11:23:34 -04:00
  • 8a20a7b711 document how to use share_strategy="no" (#1653) [skip ci] Charles Frye 2024-05-24 14:15:44 -04:00
  • 2aa3f63bcd Built site for gh-pages Quarto GHA Workflow Runner 2024-05-23 21:33:08 +00:00
  • 367b2e879b Switch to parallel FFD bin packing algorithm. (#1619) Wing Lian 2024-05-23 17:32:14 -04:00
  • 3039af581d Built site for gh-pages Quarto GHA Workflow Runner 2024-05-23 17:04:13 +00:00
  • bbfed318bc support for custom messages field in sharegpt (#1651) Wing Lian 2024-05-23 13:03:22 -04:00
  • 317ffa8bc0 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-22 12:35:13 +00:00
  • 84bb8061ba Update tiny-llama qlora.yml addressing eval packing error (#1638) Jaydeep Thik 2024-05-22 08:34:06 -04:00
  • dc5dc6ca69 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-22 12:30:02 +00:00
  • a27d5e1f4e enable loraplus setting for dpo trainer (#1646) George Grigorev 2024-05-22 13:29:06 +01:00
  • 8f1a4fa0af Built site for gh-pages Quarto GHA Workflow Runner 2024-05-22 12:29:03 +00:00
  • 6299eb5919 allow report_to for multiple providers (#1647) Wing Lian 2024-05-22 08:27:44 -04:00
  • f4b63cb079 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-21 13:09:46 +00:00
  • 7c2bf3091f Fix llama3 chat_template (extra <|eot_id|> on last turn) (#1635) Leonard 2024-05-21 22:08:53 +09:00
  • 89c9821839 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-20 20:06:45 +00:00
  • 22ae21a6c2 Add KTO support (#1640) Ben Redmond 2024-05-20 16:05:16 -04:00
  • a401fdb6a8 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-20 18:25:34 +00:00
  • ba45531802 fixes to save on fractional save_steps (#1643) Wing Lian 2024-05-20 14:24:45 -04:00
  • fe05f114d8 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-20 13:55:58 +00:00
  • 8a1572a831 Unsloth optims for Llama (#1609) Wing Lian 2024-05-20 09:55:06 -04:00
  • 82a6febc68 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-17 04:24:29 +00:00
  • 702a669cad add save_only_model option (#1634) Jeffrey Quesnelle 2024-05-16 21:23:18 -07:00
  • d4ee1fcd8d Built site for gh-pages Quarto GHA Workflow Runner 2024-05-16 05:26:31 +00:00
  • 891ae8aa13 fix ray install (#1630) Wing Lian 2024-05-16 01:25:42 -04:00
  • 1b6b2cfc88 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-16 04:06:54 +00:00
  • 0c49ecc429 more fixes to work with runpod + skypilot (#1629) Wing Lian 2024-05-16 00:05:56 -04:00
  • ddccbf4fb5 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-16 02:28:54 +00:00
  • 60113437e4 cloud image w/o tmux (#1628) Wing Lian 2024-05-15 22:27:40 -04:00
  • 30f1dddeae Built site for gh-pages Quarto GHA Workflow Runner 2024-05-16 01:36:55 +00:00
  • 419b2a6a98 install rsync too (#1627) Wing Lian 2024-05-15 21:36:00 -04:00
  • c2ae1ebb37 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-16 00:49:53 +00:00
  • 2501a371c6 fix setting the authorized keys when there are more than one in the env var (#1626) Wing Lian 2024-05-15 20:48:56 -04:00
  • b8faf3d3e4 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-15 23:42:42 +00:00
  • e6937e884b fix symlinks for axolotl outputs (#1625) Wing Lian 2024-05-15 19:41:45 -04:00
  • e5940539e3 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-15 17:28:34 +00:00
  • 039e2a0370 bump versions of deps (#1621) Wing Lian 2024-05-15 13:27:44 -04:00
  • a2bf5e9929 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-15 16:45:03 +00:00
  • 4fde300e5f update outputs path so that we can mount workspace to /workspace/data (#1623) Wing Lian 2024-05-15 12:44:13 -04:00
  • 4ff8abd33e Built site for gh-pages Quarto GHA Workflow Runner 2024-05-15 13:46:37 +00:00
  • 3319780300 update torch 2.2.1 -> 2.2.2 (#1622) Wing Lian 2024-05-15 09:45:27 -04:00
  • 3a97b393b2 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-15 00:11:41 +00:00
  • 81da7d2531 Fix total_num_steps (#1566) bofeng huang 2024-05-15 02:10:37 +02:00
  • e9a1f288cf support for custom trainer_cls from config custom-trainer-cls Wing Lian 2024-05-14 18:57:53 -04:00
  • ce661b448d Built site for gh-pages Quarto GHA Workflow Runner 2024-05-14 12:52:11 +00:00
  • 1e1921b794 FIX: max_length and max_prompt_length was not being sent to ORPOTrainer (#1584) Ali Mosavian 2024-05-14 14:51:17 +02:00
  • 1e2a2d0f83 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-14 12:49:30 +00:00