wav2vec2-large-mms-1b-ngn-on-bam-17122024

This model is a fine-tuned version of facebook/mms-1b-all on the None dataset. It achieves the following results on the evaluation set:

• Loss: 13.3362
• Wer: 0.9984

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: 0.0001
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
49.1683	0.2632	10	41.9386	1.4812
42.0061	0.5263	20	40.4597	1.6168
40.2869	0.7895	30	37.6978	1.7132
35.1294	1.0526	40	33.5613	1.6152
28.9708	1.3158	50	28.0153	1.3113
25.4906	1.5789	60	20.8692	1.0637
18.3476	1.8421	70	13.3362	0.9984

Framework versions

• Transformers 4.48.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0