joelniklaus's picture
update model card README.md
5b10a21
metadata
tags:
  - generated_from_trainer
model-index:
  - name: legal-french-roberta-base
    results: []

legal-french-roberta-base

This model was trained from scratch on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4293

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: 0.0001
  • train_batch_size: 16
  • eval_batch_size: 16
  • seed: 42
  • distributed_type: tpu
  • num_devices: 8
  • gradient_accumulation_steps: 4
  • total_train_batch_size: 512
  • total_eval_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: cosine
  • lr_scheduler_warmup_ratio: 0.05
  • training_steps: 1000000

Training results

Training Loss Epoch Step Validation Loss
0.8649 0.05 50000 0.7819
0.7852 0.1 100000 0.6027
0.5898 1.02 150000 0.5842
0.6136 1.07 200000 0.5343
0.6135 1.12 250000 0.5461
0.5804 2.03 300000 0.5295
0.5602 2.08 350000 0.5120
0.5446 2.13 400000 0.4904
0.5414 3.05 450000 0.4853
0.5765 3.1 500000 0.4788
0.6903 4.01 550000 0.4597
0.6149 4.06 600000 0.4556
0.5649 4.11 650000 0.4543
0.6449 5.03 700000 0.4489
0.6425 5.08 750000 0.4386
0.6263 5.13 800000 0.4344
0.6035 6.05 850000 0.4317
0.607 6.1 900000 0.4332
0.5899 7.01 950000 0.4321
0.5751 7.06 1000000 0.4293

Framework versions

  • Transformers 4.20.1
  • Pytorch 1.12.0+cu102
  • Datasets 2.8.0
  • Tokenizers 0.12.1