中文Sign up

DrishtiSharma/wav2vec2-large-xls-r-300m-gn-k1

🤗 Hugging Face sourceautomatic-speech-recognitionapache-2.05.0 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-gn-k1 ./model-folder
Needs a seeder →

wav2vec2-large-xls-r-300m-gn-k1

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

  • Loss: 0.9220
  • Wer: 0.6631

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-gn-k1 --dataset mozilla-foundation/common_voice_8_0 --config gn --split test --log_outputs

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

NA

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.00018
  • 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: 600
  • num_epochs: 200
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer
15.9402 8.32 100 6.9185 1.0
4.6367 16.64 200 3.7416 1.0
3.4337 24.96 300 3.2581 1.0
3.2307 33.32 400 2.8008 1.0
1.3182 41.64 500 0.8359 0.8171
0.409 49.96 600 0.8470 0.8323
0.2573 58.32 700 0.7823 0.7576
0.1969 66.64 800 0.8306 0.7424
0.1469 74.96 900 0.9225 0.7713
0.1172 83.32 1000 0.7903 0.6951
0.1017 91.64 1100 0.8519 0.6921
0.0851 99.96 1200 0.8129 0.6646
0.071 108.32 1300 0.8614 0.7043
0.061 116.64 1400 0.8414 0.6921
0.0552 124.96 1500 0.8649 0.6905
0.0465 133.32 1600 0.8575 0.6646
0.0381 141.64 1700 0.8802 0.6723
0.0338 149.96 1800 0.8731 0.6845
0.0306 158.32 1900 0.9003 0.6585
0.0236 166.64 2000 0.9408 0.6616
0.021 174.96 2100 0.9353 0.6723
0.0212 183.32 2200 0.9269 0.6570
0.0191 191.64 2300 0.9277 0.6662
0.0161 199.96 2400 0.9220 0.6631

Framework versions

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