Add comprehensive fine-tuning guide to README

Add Google Colab / Kaggle getting started section with complete install steps

Improve model card with comprehensive documentation

Add tokenizer

Fine-tuned FinBERT on 20K macro sentiment dataset (2 epochs)

Training in progress, epoch 2

Add smart router (TF-IDF+LR, 0.78 F1)

Add meta-classifier (dict+TF features, 90.5% acc)

v0.3.0: Add OOD eval, custom fine-tuning, multilingual sections to README

v0.3.0: export MultilingualHead and DEFAULT_MULTILINGUAL

v0.3.0: Add multilingual head (XLM-RoBERTa) with language-aware routing

Add out-of-domain evaluation results

Add custom fine-tuning CLI script

Add custom fine-tuning CLI (python -m macro_sentiment.finetune)"

Training in progress, epoch 1

Add YAML metadata to fix repo card warning

Rewrite README: plain-language intro, simple vs advanced usage, clear 3-line quickstart

Update eval_results.json with v0.2.0 fine-tuned model results

Update README: v0.2.0 with fine-tuned models, explicit usage instructions, updated eval results

Bump version to 0.2.0 (fine-tuned models)"

Add custom model_name support to all heads + update defaults to fine-tuned models

Remove untrained model weights from interrupted training run. Pipeline uses ProsusAI/finbert directly.

Update README with complete evaluation results and usage guide

Update transformers_ensemble with custom model_name support for FinBERTHead

Add evaluation results

Add meta-classifier training module

Update requirements

Add data preparation module

Fine-tuned FinBERT on macro sentiment dataset

Training in progress, epoch 4

Training in progress, epoch 3

Training in progress, epoch 2

Training in progress, epoch 1

Upload macro_sentiment/pipeline.py

Upload macro_sentiment/transformers_ensemble.py

Upload requirements.txt

Upload macro_sentiment/dictionaries.py

Add pipeline source code: dictionaries, transformers ensemble, unified pipeline

Add README with pipeline documentation

initial commit