Commit History
verbose failure message (#1694)
cbbf039
unverified
fix for when sample_packing and eval_sample_packing are different (#1695)
18cabc0
unverified
add back packing efficiency estimate so epochs and multi-gpu works properly (#1697)
ed8ef65
unverified
ensure explicit eval_sample_packing to avoid mismatch issues (#1692)
9c1af1a
unverified
Phi-3 conversation format, example training script and perplexity metric (#1582)
cf64284
unverified
add support for rpo_alpha (#1681)
c996881
unverified
re-enable DPO for tests in modal ci (#1374)
1f151c0
unverified
need to add back drop_last for sampler (#1676)
05b0bd0
unverified
cleanup the deepspeed proxy model at the end of training (#1675)
d4f6c65
unverified
load explicit splits on datasets (#1652)
a944f7b
unverified
set chat_template in datasets config automatically (#1664)
9d4225a
unverified
use mixins for orpo and kto configs so they work with axolotl customizations (#1674)
f7332ac
unverified
revert multipack batch sampler changes (#1672)
a6b37bd
unverified
handle the system role too for chat templates (#1671)
b752080
unverified
make sure the CI fails when pytest script fails (#1669)
fe650dd
unverified
Correct name of MixtralBlockSparseTop2MLP (L -> l) (#1667)
65db903
unverified
Fix: ensure correct handling of `val_set_size` as `float` or `int` (#1655)
6a5a725
unverified
Generalizing the chat_template prompt strategy (#1660) [skip ci]
cc11c6b
unverified
support for custom messages field in sharegpt (#1651)
bbfed31
unverified
enable loraplus setting for dpo trainer (#1646)
a27d5e1
unverified
allow report_to for multiple providers (#1647)
6299eb5
unverified
Fix llama3 chat_template (extra <|eot_id|> on last turn) (#1635)
7c2bf30
unverified
Add KTO support (#1640)
22ae21a
unverified
fixes to save on fractional save_steps (#1643)
ba45531
unverified
Unsloth optims for Llama (#1609)
8a1572a
unverified
add save_only_model option (#1634)
702a669
unverified
Fix `total_num_steps` (#1566)
81da7d2
unverified
FIX: max_length and max_prompt_length was not being sent to ORPOTrainer (#1584)
1e1921b
unverified
make sure to save on the last step (#1615)
1634ac8
unverified
fix attention mask collation (#1603)
0298273
unverified
feat: Add LLaMA-3 instruct prompt strategies for fine-tuning (#1553)
50421c8
unverified
adding llama3 fastchat conversation monkeypatch (#1539)
b32c08f
unverified
ignore the fsdp_config section too (#1606) [skip ci]
fff06af
unverified
make sure to save the lora adapter at the end of RL/dpo training (#1573)
796a085
unverified
improve tool handling roles (#1587)
cb78a36
unverified
feat: exclude mamba blocks for jamba (#1578)
8b9c15b
unverified
Pass deepspeed and fsdp as None explicitly when merging adapters to allow custom device_map (#1575)
9e1480e
unverified
improve save callbacks (#1592)
29cf15a
unverified
FIX: TRL trainer preprocessing step was running in one process (#1583)
b9bb169
unverified
Ali Mosavian
Ali Mosavian
commited on
