Udoy's picture
Model save
b1c50f5
metadata
license: apache-2.0
base_model: google/mt5-small
tags:
  - generated_from_trainer
metrics:
  - wer
model-index:
  - name: Finetuned-mT5-Bangla2IPA
    results: []

Finetuned-mT5-Bangla2IPA

This model is a fine-tuned version of google/mt5-small on the None dataset. It achieves the following results on the evaluation set:

  • Loss: 0.0529
  • Wer: 0.0530

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: 0.0003
  • train_batch_size: 4
  • eval_batch_size: 16
  • seed: 42
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 1000
  • num_epochs: 10

Training results

Training Loss Epoch Step Validation Loss Wer
4.407 0.75 2000 1.4085 0.9726
0.9875 1.5 4000 0.2242 0.2198
0.3856 2.25 6000 0.1268 0.1266
0.2443 2.99 8000 0.0929 0.0958
0.1753 3.74 10000 0.0798 0.0827
0.1431 4.49 12000 0.0701 0.0731
0.121 5.24 14000 0.0649 0.0665
0.1046 5.99 16000 0.0614 0.0625
0.0914 6.74 18000 0.0585 0.0588
0.0821 7.49 20000 0.0556 0.0564
0.0747 8.23 22000 0.0554 0.0548
0.0707 8.98 24000 0.0535 0.0540
0.0646 9.73 26000 0.0529 0.0530

Framework versions

  • Transformers 4.33.0
  • Pytorch 2.0.0
  • Datasets 2.1.0
  • Tokenizers 0.13.3