qwerrwe / scripts

Commit History

refactor scripts/finetune.py into new cli modules (#550)
861ceca
unverified

winglian Nanobit commited on

check for the existence of the default accelerate config that can create headaches (#561)
fdb777b
unverified

winglian commited on

fix wandb so mypy doesn't complain (#562)
bf08044
unverified

winglian commited on

move is_llama_derived_model into normalize_config (#524)
44454ae
unverified

tmm1 commited on

Debug tokenization output: Add ability to output text only (no tokens), and/or specify num samples to see (#511)
48434be
unverified

Tom Jobbins commited on

split train from other cli options (#503)
b21e4a2
unverified

winglian commited on

customizable ascii art (#506)
548787d
unverified

winglian commited on

tweak: use default config file when only one file is present (#501)
36b2e1c
unverified

Maxime commited on

Refactor train cfg cli (#499)
125cccb
unverified

winglian commited on

fix: inference did not move the model to the correct device (#483)
17605b8
unverified

Maxime commited on

ReLoRA implementation (with quantization) (#322)
bde3c5a
unverified

chargoddard winglian commited on

feat(docs): improve user customized prompts (#443)
04a42b6
unverified

Nanobit commited on

Ax art (#405)
29241cf
unverified

winglian commited on

add utils.data.prepare_dataset
2e22404

tmm1 commited on

use context manager to run things on rank0 before others (#397)
fc2d6be
unverified

winglian commited on

Feat(config): add max steps (#387)
3c2ad00
unverified

ittailup commited on

save tokenizer before training starts (#380)
86a91e2
unverified

winglian commited on

simplify `load_tokenizer`
efb3b2c

tmm1 commited on

improve GPU logging to break out pytorch cache and system mem
7b55fe6

tmm1 commited on

extract module for working with cfg
8cec513

tmm1 commited on

Attention mask and position id fixes for packing (#285)
2bb0b78
unverified

winglian commited on

Fix(save): Save as safetensors (#363)
a276c9c
unverified

Nanobit commited on

feat(merge): save tokenizer on merge (#362)
289d5c4
unverified

Nanobit commited on

Merge pull request #356 from tmm1/load_model-args
11ddccb
unverified

tmm1 commited on

simplify load_model signature
7181022

tmm1 commited on

log GPU memory usage
e303d64

tmm1 commited on

fix FSDP save of final model (#329)
894cba0
unverified

winglian commited on

add runpod envs to .bashrc, fix bnb env (#316)
cf62cfd
unverified

winglian commited on

misc fixes
d75adb9

winglian commited on

Fixed pre-commit problems, fixed small bug in logging_config to handle LOG_LEVEL env var
b1f4f7a

theobjectivedad commited on

Merge pull request #92 from OpenAccess-AI-Collective/flash-optimum
16bb627
unverified

winglian commited on

chore: Refactor inf_kwargs out
dc77c8e

Nanobit commited on

Merge branch 'main' into flash-optimum
fd2c981
unverified

winglian commited on

Merge pull request #177 from NanoCode012/fix/landmark-patch
8002ffb
unverified

winglian commited on

Merge pull request #159 from AngainorDev/patch-1
8e568bb
unverified

Nanobit commited on

Fix strict and Lint
b565ecf

Angainor commited on

Fix set mem_id for inference and refactor
974dc00

Nanobit commited on

Set mem cache args on inference
572d114

Nanobit commited on

fix formatting
958da70

winglian commited on

pass a prompt in from stdin for inference
c4e4f81

winglian commited on

Update scripts/finetune.py
759e867
unverified

winglian Nanobit commited on

address PR feedback
0c6f928

winglian commited on

add streaming dataset support for pretraining datasets
eea2731

winglian commited on

more tweaks to do pre-training with bettertransformers
1210dc8

winglian commited on

experimental expansion of ctx len
488a67d

winglian commited on

add flash attn context for efficient training and attempt setting model to train mode:
8792199

winglian commited on

add support for opimum bettertransformers
1edc30c

winglian commited on

Merge branch 'main' into patch-1
79e2a6f
unverified

Angainor Development commited on

Remove explicit definition of cfg.inference
c250898
unverified

Angainor Development commited on