First, we generate the training data by running the following command:
./generate_training_data.sh
Then, we train the model by running the following command:
python train_pytorch.py crnn_mobilenet_v3_large --train_path train --val_path val -j 2 --pretrained --vocab tamazight --epochs 10 --name crnn_mobilenet_v3_large --wb
Run summary:
-
exact_match 99.45%
-
partial_match 99.57%
-
val_loss 0.00942
Check out This notebook To use the model for inference.
Check out the domo on Hugging Face 