English注册

kingabzpro/wav2vec2-large-xls-r-300m-Indonesian

🤗 Hugging Face 来源automatic-speech-recognitionapache-2.0315M 参数1.3 GBsafetensors✓ 3 个校验和今天更新
一条命令提交

在你的模型文件夹旁边运行它。它会制作种子、将文件与 Hugging Face 比对，然后提交。你只需开始做种，并粘贴你账户中的密钥。它只读取你的文件，绝不修改。如果愿意，可以先阅读脚本。

curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo kingabzpro/wav2vec2-large-xls-r-300m-Indonesian ./model-folder
需要做种者 →

wav2vec2-large-xls-r-300m-Indonesian

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

  • Loss: 0.4087
  • Wer: 0.2461
  • Cer: 0.0666

Training hyperparameters

The following hyperparameters were used during training:

  • learning_rate: 0.0003
  • train_batch_size: 64
  • eval_batch_size: 8
  • seed: 42
  • gradient_accumulation_steps: 2
  • total_train_batch_size: 128
  • optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
  • lr_scheduler_type: linear
  • lr_scheduler_warmup_steps: 400
  • num_epochs: 50
  • mixed_precision_training: Native AMP

Training results

Training Loss Epoch Step Validation Loss Wer Cer
5.0788 4.26 200 2.9389 1.0 1.0
2.8288 8.51 400 2.2535 1.0 0.8004
0.907 12.77 600 0.4558 0.4243 0.1095
0.4071 17.02 800 0.4013 0.3468 0.0913
0.3 21.28 1000 0.4167 0.3075 0.0816
0.2544 25.53 1200 0.4132 0.2835 0.0762
0.2145 29.79 1400 0.3878 0.2693 0.0729
0.1923 34.04 1600 0.4023 0.2623 0.0702
0.1681 38.3 1800 0.3984 0.2581 0.0686
0.1598 42.55 2000 0.3982 0.2493 0.0663
0.1464 46.81 2200 0.4087 0.2461 0.0666

Framework versions

  • Transformers 4.17.0.dev0
  • Pytorch 1.10.2+cu102
  • Datasets 1.18.2.dev0
  • Tokenizers 0.11.0