mms-1b-dhivehi-finetuned

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

• Loss: 0.4735
• Wer: 0.5684

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: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 5
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8561	0.4813	200	0.6015	0.5990
0.7882	0.9627	400	0.5385	0.5823
0.7576	1.4428	600	0.5117	0.5758
0.75	1.9242	800	0.4980	0.5754
0.7907	2.4043	1000	0.4895	0.5727
0.7268	2.8857	1200	0.4817	0.5732
0.6884	3.3658	1400	0.4788	0.5692
0.7694	3.8472	1600	0.4750	0.5709
0.7055	4.3273	1800	0.4735	0.5684
0.8217	4.8087	2000	0.4721	0.5692

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.2
• Tokenizers 0.22.1