mms-1b-bigcgen-female-20hrs-model

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

• Loss: inf
• Wer: 0.5260

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• 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
• training_steps: 2500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
13.9528	0.0752	100	inf	1.0431
6.1846	0.1505	200	inf	1.0030
5.4651	0.2257	300	inf	1.0386
4.4356	0.3010	400	inf	0.8831
2.2016	0.3762	500	inf	0.6218
1.8013	0.4515	600	inf	0.5746
1.7499	0.5267	700	inf	0.5793
1.6979	0.6020	800	inf	0.5501
1.5567	0.6772	900	inf	0.5439
1.6301	0.7524	1000	inf	0.5355
1.6362	0.8277	1100	inf	0.5367
1.5247	0.9029	1200	inf	0.5326
1.4012	0.9782	1300	inf	0.5346
1.6397	1.0534	1400	inf	0.5301
1.5258	1.1287	1500	inf	0.5285
1.4144	1.2039	1600	inf	0.5244
1.4363	1.2792	1700	inf	0.5144
1.3733	1.3544	1800	inf	0.5358
1.4592	1.4296	1900	inf	0.5598
1.3499	1.5049	2000	inf	0.5192
1.4039	1.5801	2100	inf	0.5228
1.4057	1.6554	2200	inf	0.5289
1.4961	1.7306	2300	inf	0.5323
1.3975	1.8059	2400	inf	0.5119
1.4725	1.8811	2500	inf	0.5260

Framework versions

• Transformers 4.47.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0