mms-1b-bigcgen-male-5hrs-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: 0.4408
• Wer: 0.4520

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
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
12.4451	0.3106	100	1.2035	0.8329
1.6342	0.6211	200	0.6175	0.5759
1.5403	0.9317	300	0.5695	0.5535
1.3778	1.2422	400	0.5524	0.5359
1.4572	1.5528	500	0.5302	0.5172
1.4042	1.8634	600	0.5179	0.5266
1.4053	2.1739	700	0.5029	0.5143
1.2782	2.4845	800	0.4701	0.4864
1.2541	2.7950	900	0.4585	0.4867
1.1672	3.1056	1000	0.4728	0.4862
1.1205	3.4161	1100	0.4558	0.4794
1.1699	3.7267	1200	0.4520	0.4811
1.2418	4.0373	1300	0.4495	0.4751
1.071	4.3478	1400	0.4487	0.4737
1.078	4.6584	1500	0.4446	0.4761
1.2474	4.9689	1600	0.4437	0.4626
1.1127	5.2795	1700	0.4380	0.4657
1.1761	5.5901	1800	0.4480	0.4674
1.0997	5.9006	1900	0.4470	0.4653
1.1203	6.2112	2000	0.4421	0.4614
1.0749	6.5217	2100	0.4344	0.4506
1.1156	6.8323	2200	0.4354	0.4511
1.0404	7.1429	2300	0.4364	0.4535
1.1081	7.4534	2400	0.4377	0.4516
1.0535	7.7640	2500	0.4407	0.4520

Framework versions

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