mms-zeroshot-300m-natbed-combined-model

This model is a fine-tuned version of mms-meta/mms-zeroshot-300m on the NATBED - BEM dataset. It achieves the following results on the evaluation set:

• Loss: 0.5928
• Wer: 0.5278

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: 8
• 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: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	0.2503	200	2.6612	1.0
No log	0.5006	400	0.7787	0.6601
3.3034	0.7509	600	0.7196	0.6194
3.3034	1.0013	800	0.6961	0.5966
0.8261	1.2516	1000	0.6695	0.5762
0.8261	1.5019	1200	0.6314	0.5728
0.8261	1.7522	1400	0.6478	0.5575
0.7513	2.0025	1600	0.6374	0.5554
0.7513	2.2528	1800	0.6033	0.5484
0.7173	2.5031	2000	0.6270	0.5419
0.7173	2.7534	2200	0.6057	0.5433
0.7173	3.0038	2400	0.5928	0.5278
0.7092	3.2541	2600	0.5980	0.5321
0.7092	3.5044	2800	0.5976	0.5261
0.696	3.7547	3000	0.6369	0.5248

Framework versions

• Transformers 4.46.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0