mms-1b-all-bigc-combined-20hrs-42

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

• Loss: 1.5889
• Wer: 0.7335

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: 4
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• 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: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
8.6506	0.1623	100	5.8876	1.6051
4.6398	0.3247	200	4.4276	1.0
3.2168	0.4870	300	1.9280	0.8152
0.8072	0.6494	400	1.5889	0.7337
0.7235	0.8117	500	1.4422	0.7262
0.7213	0.9740	600	1.4460	0.7128
0.6956	1.1364	700	1.4075	0.7043
0.6309	1.2987	800	1.4961	0.7094
0.6998	1.4610	900	1.2500	0.6960
0.6451	1.6234	1000	1.2464	0.6934
0.6637	1.7857	1100	1.3811	0.6833
0.6959	1.9481	1200	1.3056	0.6780
0.6718	2.1104	1300	1.3619	0.6748

Framework versions

• Transformers 4.53.0.dev0
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.0