Edit model card

mbart-large-cc25-en-ar-evaluated-en-to-ar-1000instancesopus-leaningRate2e-05-batchSize2

This model is a fine-tuned version of akhooli/mbart-large-cc25-en-ar on the opus100 dataset. It achieves the following results on the evaluation set:

  • Loss: 0.4673
  • Bleu: 10.5645
  • Meteor: 0.0783
  • Gen Len: 10.23

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: 2e-05
  • train_batch_size: 2
  • eval_batch_size: 2
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • num_epochs: 11
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Bleu Meteor Gen Len
8.1731 0.25 100 2.8417 0.9599 0.028 230.885
0.6743 0.5 200 0.4726 6.4055 0.0692 14.81
0.3028 0.75 300 0.4572 6.7544 0.0822 23.92
0.2555 1.0 400 0.4172 8.4078 0.0742 13.655
0.1644 1.25 500 0.4236 9.284 0.071 13.03
0.1916 1.5 600 0.4222 4.8976 0.0779 32.225
0.2011 1.75 700 0.4305 7.6909 0.0738 16.675
0.1612 2.0 800 0.4416 10.8622 0.0855 10.91
0.116 2.25 900 0.4673 10.5645 0.0783 10.23

Framework versions

  • Transformers 4.18.0
  • Pytorch 1.11.0
  • Datasets 2.1.0
  • Tokenizers 0.12.1
Downloads last month
2
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.

Dataset used to train abdoutony207/mbart-large-cc25-en-ar-evaluated-en-to-ar-1000instancesopus-leaningRate2e-05-batchSize2

Evaluation results