lelapa/mms-1b-fl102-xho-15

This model is a fine-tuned version of facebook/mms-1b-fl102 on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.2132
• Wer: 0.3206

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: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 32
• optimizer: Use OptimizerNames.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: 20
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
10.6268	1.0	10	6.2896	1.6214
4.8047	2.0	20	3.2731	0.9996
1.8458	3.0	30	0.4804	0.4990
0.3363	4.0	40	0.2824	0.3707
0.2325	5.0	50	0.2379	0.3477
0.2061	6.0	60	0.2314	0.3389
0.1923	7.0	70	0.2218	0.3325
0.1849	8.0	80	0.2184	0.3308
0.1787	9.0	90	0.2176	0.3329
0.1743	10.0	100	0.2137	0.3214
0.172	11.0	110	0.2138	0.3229
0.1679	12.0	120	0.2144	0.3222
0.1674	13.0	130	0.2132	0.3206

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.5.1+cu121
• Datasets 3.2.0
• Tokenizers 0.21.0