DrishtiSharma/wav2vec2-large-xls-r-300m-hsb-v3

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

wav2vec2-large-xls-r-300m-hsb-v3

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMONVOICE8_0 - HSB dataset. It achieves the following results on the evaluation set:

• —Loss: 0.6549
• —Wer: 0.4827

Evaluation Commands

1. 1.To evaluate on mozilla-foundation/commonvoice8_0 with test split

python eval.py --modelid DrishtiSharma/wav2vec2-large-xls-r-300m-hsb-v3 --dataset mozilla-foundation/commonvoice80 --config hsb --split test --log_outputs

1. 1.To evaluate on speech-recognition-community-v2/dev_data

Upper Sorbian (hsb) language not found in speech-recognition-community-v2/dev_data!

Training hyperparameters

The following hyperparameters were used during training:

• —learning_rate: 0.00045
• —trainbatchsize: 16
• —evalbatchsize: 8
• —seed: 42
• —gradientaccumulationsteps: 2
• —totaltrainbatch_size: 32
• —optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• —lrschedulertype: linear
• —lrschedulerwarmup_steps: 500
• —num_epochs: 50
• —mixedprecisiontraining: Native AMP

Training results

Framework versions

• —Transformers 4.16.1
• —Pytorch 1.10.0+cu111
• —Datasets 1.18.2
• —Tokenizers 0.11.0