---
tags:
- generated_from_trainer
datasets:
- cassandra-themis/ner-phrases
model-index:
- name: lsg-ner-phrases-16384
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

# lsg-ner-phrases-16384

This model is a fine-tuned version of [lsg-base-16384-juri](https://huggingface.co/cassandra-themis/lsg-base-4096-juri) on the cassandra-themis/ner-phrases dataset.
It achieves the following results on the evaluation set:
- Loss: 0.0058
- New Sentence Precision: 0.9955
- New Sentence Recall: 0.9932
- New Sentence F1: 0.9943
- New Sentence Number: 442
- Overall Precision: 0.9955
- Overall Recall: 0.9932
- Overall F1: 0.9943
- Overall Accuracy: 0.9996

## Usage

```python
from transformers import AutoTokenizer, AutoModelForTokenClassification, pipeline
import re

model_path = "cassandra-themis/lsg-ner-phrases-16384"

model = AutoModelForTokenClassification.from_pretrained(model_path, trust_remote_code=True, use_auth_token=True)
tokenizer = AutoTokenizer.from_pretrained(model_path, use_auth_token=True)
ner_pipe = pipeline("token-classification", model=model, tokenizer=tokenizer)


document = "My document"
document_flattened = re.sub(r'(\s|\t|\n)+', r' ', document).strip()

prediction = ner_pipe(document_flattened, aggregation_strategy="simple")

sentences = []
for i in range(len(prediction) - 1):
    sentences.append(document_flattened[prediction[i]["start"]:prediction[i+1]["start"]].strip())
print("\n".join(sentences))
```

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 8e-05
- train_batch_size: 2
- eval_batch_size: 16
- seed: 42
- gradient_accumulation_steps: 16
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 150.0

### Training results



### Framework versions

- Transformers 4.25.1
- Pytorch 1.13.1+cu117
- Datasets 2.9.0
- Tokenizers 0.11.6