Commit Graph

  • 1634ac82e0 make sure to save on the last step (#1615) Wing Lian 2024-05-14 08:48:39 -04:00
  • 155db32084 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-14 12:18:27 +00:00
  • 02982733ec fix attention mask collation (#1603) Wing Lian 2024-05-14 08:17:30 -04:00
  • 5d97e65f95 add dstack section (#1612) [skip ci] Chansung Park 2024-05-14 21:13:45 +09:00
  • e5677a13f7 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-11 22:29:55 +00:00
  • 2147cf6837 Llama3 dpo (#1610) Wing Lian 2024-05-11 18:29:03 -04:00
  • 9a85554798 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-11 04:09:18 +00:00
  • 50421c8b1d feat: Add LLaMA-3 instruct prompt strategies for fine-tuning (#1553) Ram 2024-05-11 09:38:04 +05:30
  • ab79b48c5a Built site for gh-pages Quarto GHA Workflow Runner 2024-05-10 14:41:06 +00:00
  • b32c08f8cc adding llama3 fastchat conversation monkeypatch (#1539) Antoni-Joan Solergibert 2024-05-10 16:40:05 +02:00
  • fff06af8d0 ignore the fsdp_config section too (#1606) [skip ci] Wing Lian 2024-05-09 13:30:39 -04:00
  • 7c5aa4791f drop position_ids for olmo model olmo-no-position_ids Wing Lian 2024-05-09 00:25:15 -04:00
  • 3ef59c5503 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-08 14:40:28 +00:00
  • 796a085b2f make sure to save the lora adapter at the end of RL/dpo training (#1573) Wing Lian 2024-05-08 10:39:33 -04:00
  • 5212b6278f Built site for gh-pages Quarto GHA Workflow Runner 2024-05-07 15:31:34 +00:00
  • cb78a36374 improve tool handling roles (#1587) Wing Lian 2024-05-07 11:30:40 -04:00
  • 5e32276d7c Built site for gh-pages Quarto GHA Workflow Runner 2024-05-07 13:53:47 +00:00
  • 8b9c15b17f feat: exclude mamba blocks for jamba (#1578) NanoCode012 2024-05-07 22:52:57 +09:00
  • 07a642bdb1 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-07 13:48:44 +00:00
  • 9e1480e9ca Pass deepspeed and fsdp as None explicitly when merging adapters to allow custom device_map (#1575) Chirag Jain 2024-05-07 19:17:55 +05:30
  • 317761406e add support for NCA nca-pair Wing Lian 2024-05-06 17:01:14 -04:00
  • 6a9ac4ad27 consistency w sppo -> sppo_hard sppo Wing Lian 2024-05-06 16:58:58 -04:00
  • 027f7d54f0 update for sppo Wing Lian 2024-05-03 08:41:59 -04:00
  • 0554105baa add mistral instruct strategy and fix dpo_loss input Wing Lian 2024-05-02 23:02:03 -04:00
  • f58fcd09ec use DPOConfig Wing Lian 2024-05-02 13:44:26 -04:00
  • 60fecac367 bump trl Wing Lian 2024-05-02 12:12:53 -04:00
  • b301068098 remove override Wing Lian 2024-05-02 11:01:20 -04:00
  • df645906eb invert check Wing Lian 2024-05-02 09:31:43 -04:00
  • 7fea5822f0 add support for SPPO Wing Lian 2024-05-02 08:56:15 -04:00
  • 3574f8e320 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-06 19:44:37 +00:00
  • 3367fca732 Gradio configuration parameters (#1591) marijnfs 2024-05-06 21:43:42 +02:00
  • c597bfa313 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-06 05:06:24 +00:00
  • 1ac899800b docs(config.qmd): add loraplus example (#1577) tpoisonooo 2024-05-06 13:05:28 +08:00
  • cbe5dc18aa Built site for gh-pages Quarto GHA Workflow Runner 2024-05-05 22:46:39 +00:00
  • 70185763f6 add torch 2.3.0 to builds (#1593) Wing Lian 2024-05-05 18:45:45 -04:00
  • 2d6e3d408e Built site for gh-pages Quarto GHA Workflow Runner 2024-05-05 14:09:37 +00:00
  • 120b809465 fix for jupyterlab on cloud start (#1594) Wing Lian 2024-05-05 10:08:43 -04:00
  • f225b12c83 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-05 03:20:11 +00:00
  • 29cf15a28c improve save callbacks (#1592) Wing Lian 2024-05-04 23:19:18 -04:00
  • b0e93849c6 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-03 15:06:20 +00:00
  • dde02fcb94 Pass weakref to model in the SIGINT handler to free up model post train function (#1581) Chirag Jain 2024-05-03 20:35:28 +05:30
  • e4eb305e10 Built site for gh-pages Quarto GHA Workflow Runner 2024-05-03 15:03:52 +00:00
  • b9bb169602 FIX: TRL trainer preprocessing step was running in one process (#1583) Ali Mosavian 2024-05-03 17:02:59 +02:00
  • db261e2ce8 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-30 16:06:07 +00:00
  • 601c08b4c2 ADD: warning hub model (#1301) JohanWork 2024-04-30 18:05:12 +02:00
  • 924e64e7be Built site for gh-pages Quarto GHA Workflow Runner 2024-04-30 15:37:04 +00:00
  • cc5d31e0d9 Add debug option for RL dataset preprocessing (#1404) Abhinand 2024-04-30 21:06:04 +05:30
  • 1aeece6e24 chore(doc): clarify micro_batch_size (#1579) [skip ci] NanoCode012 2024-05-01 00:33:53 +09:00
  • e05a7c0fd0 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-27 16:29:10 +00:00
  • 5294653a2d PoSE context length ext (#1567) Wing Lian 2024-04-27 12:28:20 -04:00
  • 17f1e93c5f Built site for gh-pages Quarto GHA Workflow Runner 2024-04-27 16:08:00 +00:00
  • 98c25e15cb Add ORPO example and e2e test (#1572) Motoki Wu 2024-04-27 09:07:06 -07:00
  • 7a7c56f018 fixes to support fsdp-qdora fsdp-qdora Wing Lian 2024-04-23 08:37:04 -04:00
  • 8574fa264b Built site for gh-pages Quarto GHA Workflow Runner 2024-04-22 20:00:55 +00:00
  • 68601ec6ad make sure everything stays in the same dtype when using dpo + FSDP (#1559) Wing Lian 2024-04-22 16:00:05 -04:00
  • 51c1457619 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-21 23:56:37 +00:00
  • 40c7223aa2 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-21 23:56:10 +00:00
  • 60f5ce0569 Add support for Gemma chat template (#1530) Haoxiang Wang 2024-04-21 18:55:40 -05:00
  • 7477a53287 wrap prepared_ds_path in str() to avoid TypeError in fsspec package (#1548) Frank Ruis 2024-04-22 01:55:20 +02:00
  • 2df0b3be1f Built site for gh-pages Quarto GHA Workflow Runner 2024-04-19 21:26:28 +00:00
  • 7d1d22f72f ORPO Trainer replacement (#1551) Wing Lian 2024-04-19 17:25:36 -04:00
  • 0e8f340945 fix(yml): update llama-3 config (#1543) [skip ci] NanoCode012 2024-04-19 20:44:46 +09:00
  • c653ed0634 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-19 11:43:04 +00:00
  • 59ef25470c fix(packages): lock datasets version (#1545) NanoCode012 2024-04-19 20:42:10 +09:00
  • 3ce9b0760b fix the lora yaml for l3 fix-l3-lora Wing Lian 2024-04-19 07:28:07 -04:00
  • 4dbb561597 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-19 05:04:01 +00:00
  • c10563c444 fix broken linting (#1541) Wing Lian 2024-04-19 01:03:04 -04:00
  • 4d0a4738ea Built site for gh-pages Quarto GHA Workflow Runner 2024-04-18 19:28:25 +00:00
  • 37c037c69d Adding Llama-3 qlora (#1536) Monk (looking for PhD Fall’24) 2024-04-19 00:57:32 +05:30
  • 36d093789b Built site for gh-pages Quarto GHA Workflow Runner 2024-04-18 18:28:58 +00:00
  • 15f7910d33 llama-3 examples (#1537) Wing Lian 2024-04-18 14:28:03 -04:00
  • fed83311f8 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-18 17:21:10 +00:00
  • d28ba2e405 feat(doc): Add example for pad_token (#1535) NanoCode012 2024-04-19 02:20:20 +09:00
  • 0eadfc8c86 Create mixtral_22.yml (#1514) [skip ci] Atlas 2024-04-16 23:16:00 -06:00
  • bcaa92325d Update Readme to include support for Mixtral8X22B (#1518) [skip ci] Atlas 2024-04-16 23:15:30 -06:00
  • 7d9bafcb88 Update README.md (#1521) [skip ci] YTING 2024-04-17 13:15:05 +08:00
  • 9397b5376f Built site for gh-pages Quarto GHA Workflow Runner 2024-04-16 23:46:43 +00:00
  • e07dcb288c add docs around pre-processing (#1529) Wing Lian 2024-04-16 19:45:46 -04:00
  • dcab184145 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-16 18:54:55 +00:00
  • 6319da1f9b Unsloth gradient checkpointing offload (#1528) Wing Lian 2024-04-16 14:53:57 -04:00
  • a98fe5469e Built site for gh-pages Quarto GHA Workflow Runner 2024-04-12 13:03:34 +00:00
  • 132eb740f0 DBRX Model Support (#1462) Wing Lian 2024-04-12 09:02:36 -04:00
  • 4c92b51cd5 fix the torch dtype check merge-lora-tests Wing Lian 2024-04-11 08:56:46 -04:00
  • 3202f19f52 add save_only_model arg save_only_model Wing Lian 2024-04-10 16:09:08 -04:00
  • 5767eea874 add tests for merging lora and validating the dtype Wing Lian 2024-04-10 13:00:37 -04:00
  • b4f51e2d13 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-09 22:59:36 +00:00
  • 5ed29393e3 Update SaveAxolotlConfigtoWandBCallback to use artifact instead of save (#1483) Thomas Capelle 2024-04-10 00:58:38 +02:00
  • 6eddf7f4d8 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-09 21:29:34 +00:00
  • da9b1a3196 use locale agnostic seperator to make large nums easier to read (#1503) Wing Lian 2024-04-09 17:28:43 -04:00
  • 057fa44191 WIP: Support table logging for mlflow, too (#1506) DavidFarago 2024-04-09 23:28:27 +02:00
  • 986a6123fb Built site for gh-pages Quarto GHA Workflow Runner 2024-04-09 20:41:23 +00:00
  • 8fa0785f74 Correctly handle splits for datasets.arrow_dataset.Dataset objects (#1504) Scott Fleming 2024-04-09 16:40:26 -04:00
  • 5fb6186f85 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-09 15:06:09 +00:00
  • 4313b1a6a0 Print versions (#1496) Wing Lian 2024-04-09 11:05:15 -04:00
  • 7f17eff81a Fix the wrong adapter in qwen2-moe-qlora example (#1501) [skip ci] Maziyar Panahi 2024-04-09 16:57:24 +02:00
  • 98f6ff748b Built site for gh-pages Quarto GHA Workflow Runner 2024-04-09 01:38:51 +00:00
  • ff01c45127 add field to sft dataset pydantic for completion support (#1497) Wing Lian 2024-04-08 21:37:54 -04:00
  • f8bb4185bc skip s2 attention test due to timeout pytest-skip-s2 Wing Lian 2024-04-08 18:33:33 -04:00
  • 2c064bea61 Built site for gh-pages Quarto GHA Workflow Runner 2024-04-08 15:05:27 +00:00
  • 2fa65b9599 ignore issues with calculating # params when printing (#1493) Wing Lian 2024-04-08 11:04:22 -04:00