README.md

metadata

library_name: transformers
language:
  - en
license: apache-2.0
base_model: openai/whisper-medium
tags:
  - generated_from_trainer
metrics:
  - wer
  - precision
  - recall
  - f1
model-index:
  - name: ./whisper-medium-ea_5hr_v2
    results: []

./whisper-medium-ea_5hr_v2

This model is a fine-tuned version of openai/whisper-medium on the Afrispeech-200 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4999
• Wer Ortho: 0.2185
• Wer: 0.1579
• Cer: 0.0705
• Precision: 0.9097
• Recall: 0.9048
• F1: 0.9068

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: 16
• eval_batch_size: 8
• seed: 42
• 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: constant_with_warmup
• lr_scheduler_warmup_steps: 50
• training_steps: 500
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer	Cer	Precision	Recall	F1
0.5036	0.1179	50	0.6665	0.2366	0.1752	0.0784	0.9011	0.8964	0.8982
0.5931	0.2358	100	0.5603	0.2211	0.1622	0.0689	0.9073	0.9022	0.9043
0.5468	0.3538	150	0.5329	0.2344	0.1807	0.0911	0.9033	0.8938	0.8974
0.5159	0.4717	200	0.5213	0.2247	0.1675	0.0814	0.9084	0.9044	0.9058
0.4744	0.5896	250	0.5160	0.2332	0.1703	0.0805	0.9064	0.8987	0.9016
0.4753	0.7075	300	0.5132	0.2116	0.1536	0.0661	0.9097	0.9061	0.9074
0.5142	0.8255	350	0.4989	0.2272	0.1646	0.0755	0.9063	0.9023	0.9036
0.4951	0.9434	400	0.4928	0.2152	0.1618	0.0713	0.9098	0.9062	0.9073
0.2467	1.0613	450	0.4990	0.2084	0.1510	0.0648	0.9155	0.9106	0.9126
0.211	1.1792	500	0.4999	0.2185	0.1579	0.0705	0.9097	0.9048	0.9068

Framework versions

• Transformers 4.52.1
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1