Clean up: remove intermediate checkpoints and training artifacts

Auto-improve cycle 8 — F1=95.7%

Auto-improve cycle 7 — F1=?%

Auto-improve cycle 6 — F1=?%

Auto-improve cycle 5 — F1=?%

Auto-improve cycle 4 — F1=?%

Auto-improve cycle 3 — F1=?%

Upload Egide toxicity model (fine-tuned xlm-roberta-base)

Update model card with v2 benchmark results (96% F1, 1000-msg benchmark)

Fine-tuned v2: 96% F1, fixed false positives on Twitch chat

Clean repo: remove checkpoints and training artifacts

Upload Egide toxicity model (fine-tuned xlm-roberta-base)

Upload Egide toxicity model (fine-tuned xlm-roberta-base)

Upload Egide toxicity model (fine-tuned xlm-roberta-base)

Fix config.json: proper label names for widget display

Update model card: remove base model references, add widget examples

Remove training_args.bin

Remove training checkpoints (not needed for inference)

Add model card with full documentation

Upload Egide toxicity model (fine-tuned xlm-roberta-base)

initial commit