|
--- |
|
license: apache-2.0 |
|
language: |
|
- fr |
|
metrics: |
|
- accuracy |
|
tags: |
|
- French |
|
- proverb |
|
- nlp |
|
- bert |
|
- fine-tune |
|
--- |
|
|
|
|
|
# bert-base-french-europeana-cased |
|
|
|
This model is a fine-tuned version of [bert-base-french-europeana-cased](https://huggingface.co/dbmdz/bert-base-french-europeana-cased) on a manually created dataset. |
|
It achieves the following results on the evaluation set: |
|
- Loss: 1.21 |
|
- Accuracy: 0.85 |
|
|
|
### Training hyperparameters |
|
|
|
The following hyperparameters were used during training: |
|
- learning_rate: 2e-05 |
|
- train_batch_size: 64 |
|
- eval_batch_size: 64 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- num_epochs: 10 |
|
|
|
### Training results |
|
|
|
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|
|:-------------:|:-----:|:----:|:---------------:|:--------:| |
|
| 4.2 | 1.0 | 47 | 4.15156 | 0.174 | |
|
... |
|
| 1.216 | 10 | 490 | 1.2586 | 0.856 | |
|
|
|
|
|
## How to use |
|
|
|
```python |
|
from transformers import pipeline, AutoTokenizer |
|
|
|
model_checkpoint = "dbmdz/bert-base-french-europeana-cased" |
|
tokenizer = AutoTokenizer.from_pretrained(model_checkpoint, use_fast=True) |
|
model= "rasta/proverbes-french-IFT-7022" |
|
|
|
generator = pipeline(task="fill-mask", model=model, tokenizer=tokenizer) |
|
|
|
sentence = 'quand la poire est mûre, elle [MASK]' |
|
results = generator(sentence) |
|
|
|
``` |
|
|
|
|
|
### Framework versions |
|
|
|
- Transformers 4.18.0 |
|
- Pytorch 1.11.0+cu113 |
|
- Datasets 2.1.0 |
|
- Tokenizers 0.12.1 |
|
|
