test_v4

This model is a fine-tuned version of ./models/entities/test_v4 on the adalbertojunior/entities dataset. It achieves the following results on the evaluation set:

  • eval_loss: 0.1648
  • eval_model_preparation_time: 0.0009
  • eval_precision: 0.8411
  • eval_recall: 0.8900
  • eval_f1: 0.8649
  • eval_accuracy: 0.9744
  • eval_runtime: 177.3832
  • eval_samples_per_second: 27.962
  • eval_steps_per_second: 27.962
  • step: 0

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: 0.0002
  • train_batch_size: 8
  • eval_batch_size: 1
  • seed: 42
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 64
  • optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 500
  • num_epochs: 20.0

Testset Results

Label Precision Recall F1-Score Support
ATRIBUICAO 0.82 0.82 0.82 221
DECISAO 0.81 0.82 0.82 544
FUNCAO 0.94 0.89 0.91 486
FUNDAMENTO 0.89 0.83 0.86 1501
LOCAL 0.85 0.84 0.85 245
ORGANIZACAO 0.90 0.86 0.88 626
PEDIDO 0.86 0.81 0.83 4341
PESSOA 0.95 0.94 0.95 654
REFLEXO 0.85 0.84 0.85 358
TIPO_ACAO 0.93 0.89 0.91 341
TRIBUNAL 0.96 0.92 0.94 190
VALOR_ACORDO 0.91 0.71 0.79 41
VALOR_CAUSA 0.89 0.92 0.90 62
VALOR_CONDENACAO 0.89 0.76 0.82 72
VALOR_CUSTAS 0.95 0.93 0.94 134
VALOR_PEDIDO 0.94 0.81 0.87 308
VARA 0.95 0.96 0.96 81
micro avg 0.88 0.84 0.86 10205
macro avg 0.90 0.86 0.88 10205
weighted avg 0.88 0.84 0.86 10205

Framework versions

  • Transformers 4.47.1
  • Pytorch 2.5.1+cu121
  • Datasets 3.0.1
  • Tokenizers 0.21.0
Downloads last month
16
Safetensors
Model size
87.5M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.