natW0lf/whisper-et

Medium Amharic - Biniyam Daniel

This model is a fine-tuned version of openai/whisper-medium on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0532
• Wer: 19.1091

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
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• 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: cosine
• lr_scheduler_warmup_ratio: 0.05
• num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1306	0.0741	1000	0.1286	35.3980
0.0923	0.1482	2000	0.0923	26.9373
0.0742	0.2224	3000	0.0797	23.8271
0.0727	0.2965	4000	0.0748	23.6953
0.0706	0.3706	5000	0.0708	22.7201
0.0614	0.4447	6000	0.0689	22.0611
0.065	0.5189	7000	0.0671	21.5340
0.0594	0.5930	8000	0.0640	21.1650
0.0563	0.6671	9000	0.0619	21.1386
0.06	0.7412	10000	0.0618	21.1650
0.055	0.8153	11000	0.0597	20.6378
0.0558	0.8895	12000	0.0595	20.6115
0.0513	0.9636	13000	0.0584	19.8208
0.0521	1.0377	14000	0.0567	20.2161
0.0482	1.1118	15000	0.0567	20.0316
0.0458	1.1859	16000	0.0563	19.9789
0.0487	1.2600	17000	0.0559	19.7153
0.0448	1.3341	18000	0.0554	19.7417
0.0411	1.4083	19000	0.0553	19.4254
0.0407	1.4824	20000	0.0542	19.2145
0.0455	1.5565	21000	0.0539	19.1355
0.0439	1.6306	22000	0.0537	18.8719
0.0427	1.7048	23000	0.0538	19.2936
0.0389	1.7789	24000	0.0534	18.9773
0.0385	1.8530	25000	0.0533	19.0828
0.0366	1.9271	26000	0.0532	19.1091

Framework versions

• Transformers 4.57.3
• Pytorch 2.7.1+cu128
• Datasets 3.6.0
• Tokenizers 0.22.1