whisper-small-coser-fono

This model is a fine-tuned version of openai/whisper-small on the johnatanebonilla/coser dataset specifically utilizing the sentence_fono transcriptions. It achieves the following results on the evaluation set: It achieves the following results on the evaluation set:

• Loss: 0.7937
• Wer: 95.5939

Model description

whisper-small-coser-fono is an adaptation of OpenAI's Whisper model, specifically tailored to understand and transcribe rural Spanish dialects as captured in the sentence_fono transcriptions from the johnatanebonilla/coser This fine-tuning aims to enhance the model's ability to accurately transcribe audio that contains various dialectal phonological characteristics typical of rural Spanish areas. The model's small size makes it suitable for applications where computational resources are limited, yet it remains robust enough to handle the complexities of dialectal variations.

Intended uses & limitations

The primary use of this model is to transcribe rural Spanish dialects with high phonological accuracy. It can be particularly useful in linguistic research, dialectal studies, and applications requiring understanding of non-standard Spanish speech patterns. However, the model might exhibit limitations in understanding standard Spanish or other dialects not represented in the johnatanebonilla/coser It is also less suitable for tasks requiring understanding of context beyond the phonological level.

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.8697	0.3	1000	0.8991	79.1363
0.7742	0.59	2000	0.8372	91.9980
0.7888	0.89	3000	0.8035	101.5357
0.6478	1.19	4000	0.7937	95.5939

Framework versions

• Transformers 4.36.2
• Pytorch 2.1.0+cu121
• Datasets 2.16.1
• Tokenizers 0.15.0