README.md

---
license: mit
library_name: transformers
tags:
  - token-classification
  - ner
  - mountains
  - bert
language:
  - en
---

# Mountain NER (BERT fine-tuned)

Fine-tuned BERT token classifier for **mountain** named entity recognition
(`O`, `B-MOUNTAIN`, `I-MOUNTAIN`).

## Metrics (held-out test, 831 sentences)

| Metric | Value |
|--------|-------|
| Precision | 0.9448 |
| Recall | 0.9377 |
| F1 | **0.9413** |

Source: `Task_1_NLP/artifacts/eval/notebook_eval_results.csv` in the assessment repo.

## Load

```python
from transformers import AutoModelForTokenClassification, AutoTokenizer

repo = "<YOU>/mountain-ner-bert"  # after upload
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForTokenClassification.from_pretrained(repo)
```

Local path before upload: `publish/hf-task1-bert-ner/`.

## Training notes

- Started from a CoNLL-2003 NER BERT checkpoint and adapted to the mountain label set.
- Full pipeline: see `Task_1_NLP/` in the GitHub assessment repository.
- Intermediate Trainer checkpoints were **excluded** from this upload (final `model.safetensors` only).

## Citation / context

DS internship assessment — Task 1 (Mountain NER).