wav2vec2-large-mms-1b-twi-colab

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: 0.5085
• Wer: 0.4664

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 100

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.6457	7.69	100	1.0843	0.8419
0.7832	15.38	200	0.4796	0.5296
0.5053	23.08	300	0.4547	0.4783
0.4083	30.77	400	0.4349	0.4545
0.3281	38.46	500	0.4467	0.4506
0.276	46.15	600	0.4482	0.4427
0.2269	53.85	700	0.4847	0.4664
0.2081	61.54	800	0.4704	0.4150
0.1974	69.23	900	0.5047	0.4506
0.1695	76.92	1000	0.5142	0.4625
0.1642	84.62	1100	0.5028	0.4625
0.159	92.31	1200	0.5085	0.4664
0.1581	100.0	1300	0.5085	0.4664

Framework versions

• Transformers 4.33.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.14.4
• Tokenizers 0.13.3