wav2vec2-large-mms-1b-urmi-christian-nostress

This model is a fine-tuned version of facebook/mms-1b-all on the nena_speech_1_0_test dataset. It achieves the following results on the evaluation set:

• Loss: 0.6617
• Wer: 0.9450
• Cer: 0.1846

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: 0.001
• train_batch_size: 32
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
14.5154	0.78	25	10.7711	1.0	0.9743
7.3688	1.56	50	6.3951	1.0	0.9408
3.895	2.34	75	2.4755	1.0	0.7579
1.6699	3.12	100	0.8689	0.9725	0.2286
1.1293	3.91	125	0.7194	0.9633	0.1981
0.9637	4.69	150	0.6617	0.9450	0.1846

Framework versions

• Transformers 4.34.1
• Pytorch 2.1.0+cu121
• Datasets 2.14.6
• Tokenizers 0.14.1