automatic-speech-recognition sil-ai/bloom-speech generated_from_trainer

wav2vec2-bloom-speech-tpi

logo for Bloom Library sil-ai logo

Model description

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the SIL-AI/bloom-speech - TPI (Tok Pisin) dataset. It achieves the following results on the evaluation set:

Users should refer to the original model for tutorials on using a trained model for inference.

Intended uses & limitations

Users of this model must abide by the SIL RAIL-M License.

This model is created as a proof of concept and no guarantees are made regarding the performance of the model is specific situations.

Training and evaluation data

Training, Validation, and Test datasets were generated from the same corpus, ensuring that no duplicate files were used.

Training procedure

Standard finetuning of XLS-R was used based on the examples in the Hugging Face Transformers Github

Training hyperparameters

The following hyperparameters were used during training:

Training results

Training Loss Epoch Step Validation Loss Wer Cer
No log 7.58 250 0.4165 0.2821 0.0651
2.2947 15.15 500 0.1794 0.1590 0.0325
2.2947 22.73 750 0.1842 0.0923 0.0218
0.1337 30.3 1000 0.1713 0.1060 0.0225
0.1337 37.88 1250 0.1047 0.0923 0.0200
0.0815 45.45 1500 0.1176 0.0991 0.0236
0.0815 53.03 1750 0.2213 0.1162 0.0286
0.0616 60.61 2000 0.1363 0.0906 0.0222
0.0616 68.18 2250 0.2185 0.1077 0.0229
0.0516 75.76 2500 0.1177 0.0872 0.0190
0.0516 83.33 2750 0.1790 0.0991 0.0232
0.0424 90.91 3000 0.1505 0.0923 0.0236
0.0424 98.48 3250 0.1620 0.0923 0.0215

Framework versions