中文Sign up

DrishtiSharma/wav2vec2-large-xls-r-300m-vot-final-a2

🤗 Hugging Face sourceautomatic-speech-recognitionapache-2.03.8 GBother✓ 4 checksumsupdated today
Submit in one command

Run it next to your model folder. It makes the torrent, checks your files against Hugging Face, and submits it. You just start seeding and paste your key from your account. It only reads your files and never changes them. Read the script first if you like.

curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo DrishtiSharma/wav2vec2-large-xls-r-300m-vot-final-a2 ./model-folder
Needs a seeder →

wav2vec2-large-xls-r-300m-vot-final-a2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the MOZILLA-FOUNDATION/COMMON_VOICE_8_0 - VOT dataset. It achieves the following results on the evaluation set:

  • Loss: 2.8745
  • Wer: 0.8333

Evaluation Commands

  1. To evaluate on mozilla-foundation/common_voice_8_0 with test split

python eval.py --model_id DrishtiSharma/wav2vec2-large-xls-r-300m-vot-final-a2 --dataset mozilla-foundation/common_voice_8_0 --config vot --split test --log_outputs

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

Votic language isn't available in speech-recognition-community-v2/dev_data

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0004
  • train_batch_size: 16
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 32
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 340
  • num_epochs: 200
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
11.1216 33.33 100 4.2848 1.0
2.9982 66.67 200 2.8665 1.0
1.5476 100.0 300 2.3022 0.8889
0.2776 133.33 400 2.7480 0.8889
0.1136 166.67 500 2.5383 0.8889
0.0489 200.0 600 2.8745 0.8333

Framework versions

  • Transformers 4.16.2
  • Pytorch 1.10.0+cu111
  • Datasets 1.18.3
  • Tokenizers 0.11.0