Jun 1 15:24 Add xlmr-udd1 safetensors weights (UDD-1 test UAS 89.25 / LAS 82.87) + safetensors loader Feb 7 04:46 Update src/__init__.py 04:38 Add model checkpoints and src/models to repo rain1024 +1: Claude Opus 4.6 - Update .gitignore to allow models/*.pt
- Add bamboo-1.0.0-20260202-phobert-vndt.pt (521M)
- Add bamboo-1.0.0-20260202-xlmr-udd1.pt (1.1G)
- Add src/models/ (MST, transformer_parser)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> 04:29 Consolidate project: merge scripts/, bamboo1/ into src/, optimize training rain1024 +1: Claude Opus 4.6 - Move all scripts/*.py and *.sh to src/
- Move bamboo1/ models to src/models/
- Move parse_sentences.py and references/paper_db.py to src/
- Remove requirements.txt (pyproject.toml is superset)
- Remove bamboo1/, scripts/, references/ directories
- Optimize trankit training: pre-tokenize dataset, torch.gather,
larger batch size (64), higher LR, reduced patience/warmup
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> 03:20 Move models to models/, add evaluation tools, update training infra rain1024 +1: Claude Opus 4.6 - Move .pt model files to models/ (gitignored, stored on HF LFS)
- Add cost_estimate.py and evaluate_detailed.py
- Fix torch.save for network filesystem (tempfile + shutil.move)
- Update TECHNICAL_REPORT with VnDT PhoBERT results (84.29% UAS, 77.22% LAS)
- Add more RunPod datacenters to runpod_setup.py
- Update CLAUDE.md with RunPod volume and GPU info
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Feb 2 08:41 Upload src/evaluate.py with huggingface_hub 08:41 Upload src/train.py with huggingface_hub 08:41 Upload src/vndt_corpus.py with huggingface_hub 08:41 Upload src/ud_corpus.py with huggingface_hub 08:41 Upload src/corpus.py with huggingface_hub 08:41 Upload src/inference.py with huggingface_hub 08:41 Upload src/__init__.py with huggingface_hub 
