make sure everything stays in the same dtype when using dpo + FSDP (#1559) 68601ec unverified winglian commited on Apr 22, 2024
Add support for Gemma chat template (#1530) 60f5ce0 unverified Haoxiang-Wang winglian commited on Apr 21, 2024
wrap prepared_ds_path in str() to avoid TypeError in fsspec package (#1548) 7477a53 unverified Frank Ruis winglian commited on Apr 21, 2024
fix(yml): update llama-3 config (#1543) [skip ci] 0e8f340 unverified Nanobit commited on Apr 19, 2024
Update Readme to include support for Mixtral8X22B (#1518) [skip ci] bcaa923 unverified Atlas commited on Apr 17, 2024
Update SaveAxolotlConfigtoWandBCallback to use artifact instead of save (#1483) 5ed2939 unverified tcapelle winglian commited on Apr 9, 2024
use locale agnostic seperator to make large nums easier to read (#1503) da9b1a3 unverified winglian commited on Apr 9, 2024
WIP: Support table logging for mlflow, too (#1506) 057fa44 unverified DavidFarago Dave Farago winglian commited on Apr 9, 2024
Correctly handle splits for datasets.arrow_dataset.Dataset objects (#1504) 8fa0785 unverified scottifer8 winglian commited on Apr 9, 2024
Fix the wrong adapter in qwen2-moe-qlora example (#1501) [skip ci] 7f17eff unverified MaziyarPanahi commited on Apr 9, 2024
add field to sft dataset pydantic for completion support (#1497) ff01c45 unverified winglian commited on Apr 9, 2024
ignore issues with calculating # params when printing (#1493) 2fa65b9 unverified winglian commited on Apr 8, 2024
drop empty token from beginning if tokenizer has no bos_token (in the case of qwen) (#1490) 934fc85 unverified winglian commited on Apr 7, 2024
feat: validate sample packing requires flash_attention (#1465) bf4cd67 unverified Nanobit commited on Apr 5, 2024
don't use deepspeed or fsdp when merging loras (#1479) 87ca3f9 unverified winglian commited on Apr 5, 2024
refactor utils.data module for line count linter (#1476) e0fcef4 unverified winglian commited on Apr 4, 2024
Added pip install ninja to accelerate installation of flash-attn (#1461) cae608f unverified melvinebenezer commited on Apr 2, 2024
feat: add deepspeed 3 with cpuoffload (#1466) 946b497 unverified Nanobit winglian commited on Apr 1, 2024
make sure to install causal_conv1d in docker (#1459) 89134f2 unverified winglian commited on Mar 29, 2024
fix how nightly tag is generated (#1456) [skip ci] da3415b unverified winglian commited on Mar 29, 2024
configure nightly docker builds (#1454) [skip ci] 8cb127a unverified winglian commited on Mar 29, 2024
Support loading datasets saved via save_to_disk (#1432) e634118 unverified fozziethebeat commited on Mar 29, 2024