Go to file
2023-03-17 14:38:19 +01:00
notepads naming change 2023-01-10 18:24:44 +01:00
old new training script with config, files split 2023-01-04 11:10:24 +01:00
utils random choice for sequences in streaming dataset 2023-03-17 14:38:19 +01:00
.gitignore gitignore, reqs, train_stream test 2023-03-14 21:43:21 +00:00
config-eval.yaml config eval and donut eval better params 2023-01-04 10:00:50 +01:00
config-train.yaml stream trainning 2023-03-14 15:48:49 +01:00
eval.py naming change 2023-01-10 18:24:44 +01:00
README.md readme 2023-01-10 18:29:00 +01:00
requirements.txt gitignore, reqs, train_stream test 2023-03-14 21:43:21 +00:00
train_stream.py train_len 2023-03-14 21:44:39 +00:00
train.py restore donut 2023-01-25 17:44:25 +01:00

Donut script for train and eval

Instalation/setup:

-> pip install -r requirements.txt -> export HUG_TOKEN=YourToken - token needed for pushing model to huggingface after training

Usage:

Train

-> python train --config config-train.yaml

Evaluation

-> python eval --config config-eval.yaml

Configuration

Need to suply config file like config-eval.yaml for evaluation or config-train.yaml for traning