Почетна Преглед Помоћ
Регистрација Пријавите се
AIForce
/
train-dalle-together
огледало од https://github.com/learning-at-home/dalle-hivemind.git
5
0
Креирај огранак 0
Датотеке Дискусије 0 Вики

Нема описа

Дрво: b67665d0eb
Гране Ознаке
train-dalle-tog...
ZIP TAR.GZ
Alexander Borzunov b67665d0eb Add LICENSE пре 2 година
inference acf688eac2 Use --top_p and --top_k options in run_inference.py пре 3 година
lib 64dee420da Upgrade to using hivemind.optim.experimental пре 3 година
.gitignore 72fc0bcdb7 Initial commit (ru-max branch without private code) пре 3 година
LICENSE b67665d0eb Add LICENSE пре 2 година
README.md d4140807f4 Update readme пре 3 година
arguments.py 19e3d2d060 Make aux peers fetch checkpoints every 2 steps by default пре 3 година
callback.py 4918c58cb6 Polish stdout пре 3 година
data.py c61c61b20d Use t5-small tokenizer пре 3 година
huggingface_auth.py 3e604bc1f5 fix auth пре 3 година
manage_scaleset.py c365b2ec9f Tweak settings for the upcoming demo (#2) пре 3 година
requirements.txt 872db56e0b Fix bitsandbytes version пре 3 година
run_aux_peer.py 748f20991c Make runners executable and improve their shebangs пре 3 година
run_trainer.py 748f20991c Make runners executable and improve their shebangs пре 3 година
run_trainer_tpu.py 748f20991c Make runners executable and improve their shebangs пре 3 година
task.py 09240991cc Make model uploading use access token from authorizer (#7) пре 3 година
utils.py f621362466 Make logging less verbose пре 3 година

README.md

Training DALL-E with volunteers from all over the Internet

This repository is a part of the NeurIPS 2021 demonstration "Training Transformers Together".

In this demo, we train a model similar to OpenAI DALL-E — a Transformer "language model" that generates images from text descriptions. Training happens collaboratively — volunteers from all over the Internet contribute to the training using hardware available to them. We use LAION-400M, the world's largest openly available image-text-pair dataset with 400 million samples. Our model is based on the dalle‑pytorch implementation by Phil Wang with a few tweaks to make it communication-efficient.

See details about how to join and how it works on our website.