mms-1b-allFT-QASR-CS-ara

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: inf
• Wer: 0.6871

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: 1e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_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: 500
• num_epochs: 7
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.145	1.0	398	inf	0.6692
2.8066	2.0	796	inf	0.6609
2.1619	1.9994	1588	inf	0.6670
2.1074	3.0	2383	inf	0.6871
2.1159	4.0	3178	inf	0.6871
2.1113	5.0	3973	inf	0.6874
2.1168	6.0	4768	inf	0.6872
2.1044	6.9943	5558	inf	0.6871

Framework versions

• Transformers 4.50.3
• Pytorch 2.2.2+cu121
• Datasets 3.0.0
• Tokenizers 0.21.1