miosipof/asr_IT_AUG_Synth

Whisper Medium

This model is a fine-tuned version of openai/whisper-medium on the ASR_Synthetic_Speecht5_TTS dataset. It achieves the following results on the evaluation set:

• Loss: 3.1269
• Wer: 75.9657

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: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 300
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
7.8335	0.0244	25	7.0413	196.1373
6.5445	0.0489	50	5.3231	160.5150
4.2166	0.0733	75	3.6955	144.6352
2.2081	0.0978	100	3.5441	84.9785
1.815	0.1222	125	3.4335	83.4049
1.6465	0.1467	150	3.2851	136.9099
1.5241	0.1711	175	3.3021	357.3677
1.3811	0.1956	200	3.2476	81.5451
1.2553	0.2200	225	3.1495	132.1888
1.3158	0.2445	250	3.1816	76.9671
1.2942	0.2689	275	3.1349	74.3920
1.1935	0.2934	300	3.1269	75.9657

Framework versions

• PEFT 0.13.2
• Transformers 4.44.2
• Pytorch 2.2.0
• Datasets 3.1.0
• Tokenizers 0.19.1