eslamxm's picture
update model card README.md
aec9b30
metadata
license: apache-2.0
tags:
  - summarization
  - arabic
  - ar
  - mt5
  - Abstractive Summarization
  - generated_from_trainer
datasets:
  - xlsum
model-index:
  - name: mt5-base-finetuned-urdu-finetuned-urdu-arabic
    results: []

mt5-base-finetuned-urdu-finetuned-urdu-arabic

This model is a fine-tuned version of eslamxm/mt5-base-finetuned-urdu on the xlsum dataset. It achieves the following results on the evaluation set:

  • Loss: 3.3744
  • Rouge-1: 22.77
  • Rouge-2: 10.15
  • Rouge-l: 20.71
  • Gen Len: 19.0
  • Bertscore: 71.46

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: 0.0005
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 8
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 5
  • label_smoothing_factor: 0.1

Training results

Training Loss Epoch Step Validation Loss Rouge-1 Rouge-2 Rouge-l Gen Len Bertscore
4.5155 1.0 1172 3.6895 18.81 6.77 17.01 19.0 70.27
3.8315 2.0 2344 3.5047 19.75 7.79 17.95 19.0 70.58
3.6122 3.0 3516 3.4231 20.46 8.44 18.7 19.0 70.8
3.4735 4.0 4688 3.3835 21.12 8.86 19.21 19.0 70.98
3.3855 5.0 5860 3.3744 21.48 9.01 19.57 19.0 71.17

Framework versions

  • Transformers 4.18.0
  • Pytorch 1.11.0+cu113
  • Datasets 2.2.1
  • Tokenizers 0.12.1