whisper-small-ar-eos-v6-mulaw

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

• Loss: 0.7786
• Accuracy: 0.7073

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: 5e-07
• train_batch_size: 32
• eval_batch_size: 32
• 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: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 30
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.8494	1.0	557	0.8372	0.5987
0.7418	2.0	1114	0.7249	0.7012
0.7201	3.0	1671	0.7034	0.7144
0.7032	4.0	2228	0.6996	0.7123
0.7016	5.0	2785	0.6956	0.7169
0.64	6.0	3342	0.7002	0.7184
0.6507	7.0	3899	0.6937	0.7245
0.6972	8.0	4456	0.6966	0.7235
0.6435	9.0	5013	0.6954	0.7260
0.6437	10.0	5570	0.6969	0.7194
0.675	11.0	6127	0.7022	0.7225
0.6486	12.0	6684	0.7024	0.7220
0.5699	13.0	7241	0.7103	0.7199
0.6024	14.0	7798	0.7070	0.7220
0.6019	15.0	8355	0.7145	0.7220
0.6209	16.0	8912	0.7201	0.7230
0.5811	17.0	9469	0.7241	0.7210
0.5941	18.0	10026	0.7281	0.7199
0.5847	19.0	10583	0.7371	0.7210
0.5309	20.0	11140	0.7435	0.7098
0.5697	21.0	11697	0.7490	0.7128
0.5536	22.0	12254	0.7534	0.7103
0.5517	23.0	12811	0.7636	0.7118
0.5907	24.0	13368	0.7624	0.7123
0.5081	25.0	13925	0.7674	0.7103
0.5305	26.0	14482	0.7706	0.7103
0.4848	27.0	15039	0.7740	0.7083
0.4801	28.0	15596	0.7761	0.7093
0.5059	29.0	16153	0.7776	0.7042
0.51	30.0	16710	0.7786	0.7073

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0