metadata

license: apache-2.0
base_model: facebook/wav2vec2-large-xlsr-53
tags:
  - generated_from_trainer
datasets:
  - common_voice
metrics:
  - wer
model-index:
  - name: Model_G_2
    results:
      - task:
          name: Automatic Speech Recognition
          type: automatic-speech-recognition
        dataset:
          name: common_voice
          type: common_voice
          config: id
          split: test
          args: id
        metrics:
          - name: Wer
            type: wer
            value: 0.9852694387469699

Model_G_2

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the common_voice dataset. It achieves the following results on the evaluation set:

Loss: 0.0359
Wer: 0.9853
Cer: 0.7143

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0003
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: 500
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
3.8996	0.81	400	0.7268	1.0008	0.7672
0.5216	1.61	800	0.2765	1.0171	0.7602
0.3112	2.42	1200	0.1712	0.9965	0.7335
0.2343	3.23	1600	0.1169	0.9984	0.7262
0.1911	4.03	2000	0.0970	0.9970	0.7447
0.1625	4.84	2400	0.0834	0.9941	0.7245
0.1471	5.65	2800	0.0771	0.9936	0.7239
0.1301	6.45	3200	0.0645	0.9940	0.7330
0.1241	7.26	3600	0.0621	0.9912	0.7208
0.1128	8.06	4000	0.0672	0.9892	0.7188
0.1035	8.87	4400	0.0531	0.9895	0.7332
0.0993	9.68	4800	0.0541	0.9912	0.7374
0.0917	10.48	5200	0.0516	0.9883	0.7276
0.0879	11.29	5600	0.0507	0.9841	0.7246
0.0836	12.1	6000	0.0490	0.9858	0.7335
0.0767	12.9	6400	0.0464	0.9844	0.7231
0.0744	13.71	6800	0.0458	0.9855	0.7170
0.0695	14.52	7200	0.0506	0.9893	0.7145
0.0676	15.32	7600	0.0443	0.9892	0.7151
0.0621	16.13	8000	0.0457	0.9831	0.7188
0.0593	16.94	8400	0.0437	0.9905	0.7251
0.0558	17.74	8800	0.0419	0.9881	0.7160
0.0539	18.55	9200	0.0403	0.9897	0.7128
0.0509	19.35	9600	0.0435	0.9853	0.7195
0.0482	20.16	10000	0.0451	0.9863	0.7170
0.0452	20.97	10400	0.0397	0.9874	0.7128
0.0438	21.77	10800	0.0378	0.9874	0.7108
0.0419	22.58	11200	0.0394	0.9881	0.7096
0.0389	23.39	11600	0.0412	0.9874	0.7105
0.0377	24.19	12000	0.0388	0.9847	0.7180
0.0362	25.0	12400	0.0365	0.9848	0.7149
0.0336	25.81	12800	0.0363	0.9840	0.7144
0.0315	26.61	13200	0.0366	0.9855	0.7138
0.031	27.42	13600	0.0381	0.9864	0.7171
0.0303	28.23	14000	0.0363	0.9857	0.7145
0.0276	29.03	14400	0.0365	0.9854	0.7136
0.0282	29.84	14800	0.0359	0.9853	0.7143

Framework versions

Transformers 4.31.0
Pytorch 2.0.1+cu117
Datasets 1.18.3
Tokenizers 0.13.3