wav2vec2-Pratyush-poly

This model is a fine-tuned version of Maverick1713/wav2vec2-Pratyush-poly on the minds14 dataset. It achieves the following results on the evaluation set:

• Loss: 13.1938
• Wer: 0.9648

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: 0.0002
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 6
• total_train_batch_size: 24
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 15
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
No log	1.3274	25	12.2640	0.9648
No log	2.6549	50	12.1580	0.9648
No log	3.9823	75	12.1267	0.9648
No log	5.3097	100	12.2379	0.9648
No log	6.6372	125	12.8694	0.9648
No log	7.9646	150	12.5758	0.9648
No log	9.2920	175	12.4888	0.9648
9.4864	10.6195	200	13.2420	0.9648
9.4864	11.9469	225	13.3185	0.9648
9.4864	13.2743	250	13.1938	0.9648

Framework versions

• Transformers 4.44.2
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.19.1