README.md

metadata

library_name: transformers
language:
  - en
license: apache-2.0
base_model: openai/whisper-base
tags:
  - generated_from_trainer
metrics:
  - wer
  - precision
  - recall
  - f1
model-index:
  - name: ./whisper-base-ea_base_v2
    results: []

./whisper-base-ea_base_v2

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

• Loss: 1.0084
• Wer Ortho: 0.3014
• Wer: 0.2499
• Cer: 0.1058
• Precision: 0.8391
• Recall: 0.8445
• F1: 0.8412

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: 500
• training_steps: 4000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer Ortho	Wer	Cer	Precision	Recall	F1
0.1669	2.1186	500	0.7136	0.2789	0.2257	0.1014	0.8596	0.8608	0.8593
0.0787	4.2373	1000	0.7690	0.2915	0.2394	0.1081	0.8514	0.8536	0.8515
0.0243	6.3559	1500	0.8373	0.2916	0.2405	0.1010	0.8461	0.8531	0.8491
0.012	8.4746	2000	0.8798	0.2880	0.2372	0.1005	0.8484	0.8551	0.8512
0.0081	10.5932	2500	0.9189	0.2966	0.2459	0.1047	0.8422	0.8477	0.8443
0.0038	12.7119	3000	0.9609	0.2955	0.2460	0.1025	0.8428	0.8493	0.8455
0.0048	14.8305	3500	0.9747	0.2971	0.2483	0.1051	0.8423	0.8478	0.8445
0.0039	16.9492	4000	1.0084	0.3014	0.2499	0.1058	0.8391	0.8445	0.8412

Framework versions

• Transformers 4.51.3
• Pytorch 2.6.0+cu124
• Datasets 3.6.0
• Tokenizers 0.21.1