wav2vec2-xls-r-300m-malayalam-colab-CV17.0-v2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the common_voice_17_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.9415
• Wer: 0.7946
• Cer: 0.1990

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_ratio: 0.15
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
8.3824	3.1496	200	3.5244	1.0	1.0
2.8615	6.2992	400	1.4480	0.9716	0.3680
0.8112	9.4488	600	0.9231	0.9188	0.2573
0.4211	12.5984	800	0.9136	0.8843	0.2477
0.2862	15.7480	1000	0.9257	0.8533	0.2370
0.21	18.8976	1200	0.9450	0.8185	0.2188
0.1772	22.0472	1400	0.9285	0.8343	0.2151
0.1432	25.1969	1600	0.9596	0.8262	0.2110
0.117	28.3465	1800	0.9419	0.7985	0.2026
0.1047	31.4961	2000	0.9415	0.7946	0.1990

Framework versions

• Transformers 4.42.4
• Pytorch 2.4.0+cu121
• Datasets 2.21.0
• Tokenizers 0.19.1