Sagicc/whisper-medium-sr-v2

Whisper Medium sr v2

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

• Loss: 0.2216
• Wer Ortho: 0.1663
• Wer: 0.0738

Model description

This is a fine tunned on merged datasets Common Voice 16 + Fleurs + Juzne vesti (South news) + LBM

Rupnik, Peter and Ljubešić, Nikola, 2022,
ASR training dataset for Serbian JuzneVesti-SR v1.0, Slovenian language resource repository CLARIN.SI, ISSN 2820-4042,
http://hdl.handle.net/11356/1679.

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 50
• training_steps: 1500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer
0.3634	0.40	500	0.1619	0.1953	0.0921
0.3185	0.81	1000	0.1423	0.175	0.0800
0.2216	1.21	1500	0.137	0.1663	0.0738

Framework versions

• Transformers 4.35.2
• Pytorch 2.0.1+cu117
• Datasets 2.14.5
• Tokenizers 0.14.1