PeppePasti/IT-FINANCIAL-BERT

model

This model is a fine-tuned version of dlicari/Italian-Legal-BERT-SC on a custom dataset about financial norms.

Intended uses & limitations

This mode has just been further pre-trained, if you intend to use it for downstream task you should fine-tune it with your data.

Usage

To use this model you can use the following script:

from transformers import AutoTokenizer, AutoModel,CamembertModel,pipeline
import torch
model = CamembertModel.from_pretrained("PeppePasti/IT-FINANCIAL-BERT")
tokenizer = AutoTokenizer.from_pretrained("dlicari/Italian-Legal-BERT-SC")
tokenizer.model_max_length=512
classifier = pipeline("fill-mask", model=model, tokenizer=tokenizer)
classifier(" La Repubblica riconosce a tutti i <mask> il diritto al lavoro e promuove le condizioni che rendano effettivo questo diritto.")

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-06
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 256
• total_train_batch_size: 2048
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.5
• num_epochs: 1
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Input Tokens Seen
0.4096	0.9982	202	211812352

Framework versions

• Transformers 4.40.0
• Pytorch 2.3.0+cu118
• Datasets 2.20.0
• Tokenizers 0.19.1