mms-1b-swagen-combined-15hrs-model

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

• Loss: 0.2307
• Wer: 0.1929

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
14.8801	0.0797	100	0.7377	0.4426
0.6766	0.1594	200	0.2688	0.2006
0.5153	0.2391	300	0.2484	0.1975
0.526	0.3189	400	0.2398	0.1949
0.4874	0.3986	500	0.2398	0.1958
0.4666	0.4783	600	0.2358	0.1909
0.4406	0.5580	700	0.2391	0.1944
0.4689	0.6377	800	0.2334	0.1926
0.462	0.7174	900	0.2293	0.1927
0.4407	0.7971	1000	0.2293	0.1931
0.4567	0.8768	1100	0.2298	0.1928
0.4711	0.9566	1200	0.2305	0.1972
0.4444	1.0359	1300	0.2307	0.1929

Framework versions

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