Commit History
chore: Clean up repetitive model kwargs (#670)
e62d590
unverified
Feat: Allow usage of native Mistral FA when no sample_packing (#669)
697c50d
unverified
refactor to set eval_batch_size earlier if unset, so we can warn if mismatched (#662)
2642cae
unverified
remove patch fix for phi (#664)
f34648c
unverified
prepared dataset caching, other misc fixes (#665)
e50a64e
unverified
Make dataset_processes configurable (#651)
9ec2077
unverified
Fix bug when using pretokenized datasets (#652)
590d603
unverified
ich
commited on
add support for defined train split (#654)
409ca0f
unverified
Feat: Add example for Mistral (#644)
eb41f76
unverified
Fix(cfg): Add validation for save_strategy and eval_strategy (#633)
383f88d
unverified
Mistral flash attn packing (#646)
b6ab8aa
unverified
Fix bug in dataset loading (#284)
8fe0e63
unverified
skip some flash attn patches unless explicitly enabled (#643)
895f0a0
unverified
use fastchat conversations template (#578)
e7d3e2d
unverified
attention_mask not needed for training (#642)
e8cbf50
unverified
Feat: Add support for upstream FA2 (#626)
19a600a
unverified
Fix: Fail bf16 check when running on cpu during merge (#631)
cfbce02
unverified
chore(callback): Remove old peft saving code (#510)
d5f8589
unverified
misc fixes to add gptq tests (#621)
03e5907
unverified
run eval on the first step to get a baseline (#617)
2844eb2
unverified
skip the gpu memory checks if the device is set to 'auto' (#609)
196ff11
unverified
fix distributed devices (#612)
2fe95cd
unverified
Maxime
commited on
support to disable exllama for gptq (#604)
faecff9
unverified
Delete duplicate lines (#606)
aa656e0
unverified
minor tweaks to simplify (#597)
31b9e0c
unverified
btlm and falcon monkey patches for flash attn (#566)
6b9b229
unverified
add bf16 check (#587)
131afdb
unverified
Feat(data): Allow loading local csv and text (#594)
00dce35
unverified
gather/broadcast the max value of the packing efficiency automatically (#463)
b15b19e
unverified
don't add position_ids for evals (#591)
ab534d7
unverified
optionally configure sample packing for evals (#589)
21ec195
unverified
make phi training work with Loras (#588)
62eaee7
unverified
don't resize embeddings if it's already large enough (#577)
3607882
unverified
Support Sample packing for phi arch (#586)
12a2dbb
unverified
support custom field for completion from yml (#580)
f7a2263
unverified
E2e device cuda (#575)
2414673
unverified
mypy wandb ignore (#572)
c6d870b
unverified
remove columns after tokenizing for pretraining (#571)
1157950
unverified
fix save_steps so it doesn't get duplicated (#567)
3fbde76
unverified
Model parallel (#538)
f6060a6
unverified
improve how we setup eval/save strategies and steps (#547)
36e53c7
unverified
gracefully handle length feature used for group by (#565)
e7aa7b1
unverified
add optimization for group-by-len (#563)
e5bb22a
unverified
fix wandb so mypy doesn't complain (#562)
bf08044
unverified
Add training callback to send predictions to WandB table (#521)
5b67ea9
unverified
Fix pretraining with iterable/streaming Dataset (#556)
2f586d1
unverified
Jan Philipp Harries
Jan Philipp Harries
commited on