airKlizz's picture
update model card README.md
910dd53
metadata
license: apache-2.0
tags:
  - summarization
  - generated_from_trainer
metrics:
  - rouge
model-index:
  - name: mt5-base-wikinewssum-polish
    results: []

mt5-base-wikinewssum-polish

This model is a fine-tuned version of google/mt5-base on an unknown dataset. It achieves the following results on the evaluation set:

  • Loss: 2.3179
  • Rouge1: 7.911
  • Rouge2: 3.2189
  • Rougel: 6.7856
  • Rougelsum: 7.4485

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: 5.6e-05
  • train_batch_size: 4
  • eval_batch_size: 4
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 8
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 8

Training results

Training Loss Epoch Step Validation Loss Rouge1 Rouge2 Rougel Rougelsum
No log 1.0 315 2.5391 5.9874 2.3594 5.1303 5.6116
No log 2.0 630 2.4446 7.7294 3.0152 6.6024 7.2757
No log 3.0 945 2.3912 7.6451 2.9785 6.5714 7.2011
3.5311 4.0 1260 2.3720 7.8007 3.0913 6.7067 7.3451
3.5311 5.0 1575 2.3411 7.8374 3.1208 6.7288 7.3459
3.5311 6.0 1890 2.3354 7.8664 3.1655 6.762 7.4364
3.5311 7.0 2205 2.3175 7.9529 3.2225 6.8438 7.4904
2.692 8.0 2520 2.3179 7.911 3.2189 6.7856 7.4485

Framework versions

  • Transformers 4.13.0
  • Pytorch 1.10.1
  • Datasets 1.16.1
  • Tokenizers 0.10.3