Hugging Face's logo

Spaces:
rshakked
/
safe-talk
Running

App Files Community
safe-talk / train_abuse_model.py

Commit History

fix: make label_columns explicit parameter in label_row_soft
5071ec6

rshakked commited on

refactor: create predict_pipeline.py and utils.py to modularize app logic
bc56514

rshakked commited on

fix: add missing import for time module in training loop
def4deb

rshakked commited on

add support for hub integration in HF space
c2281d2

rshakked commited on

feat: persist model and logs in Hugging Face Space + add model push to hub
54299e5

rshakked commited on

feat: add support for evaluating saved model without retraining
1294c96

rshakked commited on

fix: pass tokenizer explicitly to AbuseDataset and safeguard evaluation step
5c16708

rshakked commited on

feat(app): run training on button click and display logs after completion
2032430

rshakked commited on

fix: reduce batch size and enable gradient checkpointing to prevent GPU OOM crashes
fedc8f2

rshakked commited on

chore(debug): move device check to top of script for clearer startup logs
285a433

rshakked commited on

chore(debug): print tensor devices and CUDA availability for troubleshooting
d4ac0ac

rshakked commited on

fix: remove .to(device) from dataset to avoid pin_memory error in Trainer
92846e1

rshakked commited on

Add version print for transformers to debug dependency issues
fea5367

rshakked commited on

updated train_abuse_model to save the traind model after training process
5524706

rshakked commited on

Fresh start: safe version with token moved to env secret
80b417a

rshakked commited on

Update training code
7a679b4

rshakked commited on

added app file and enable GPU training
58be06f

rshakked commited on

Add model training script
0618de2

rshakked commited on