Machine Learning
Everything on this site that shares this tag — one index instead of scanning the whole blog.
1 item
End-to-end sentiment pipeline: noisy CSV → cleaning & dedup → XLM-RoBERTa fine-tuning with class-weighted loss (max_len 128–256). Decision threshold tuned on VALID. Experiments tracked in W&B/MLflow. Public deliverable: a reusable W&B dashboard you can duplicate and feed with your own runs.