whisper-small-ablation-3

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

• Loss: 0.2778
• Wer: 11.1006

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: 32
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 10000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.3813	0.6693	500	0.2758	17.4282
0.1944	1.3387	1000	0.2244	14.1938
0.1419	2.0080	1500	0.2037	12.3724
0.0757	2.6774	2000	0.2087	11.6659
0.0315	3.3467	2500	0.2192	11.9328
0.0236	4.0161	3000	0.2198	11.6659
0.0142	4.6854	3500	0.2310	11.3833
0.0058	5.3548	4000	0.2405	11.9014
0.0039	6.0241	4500	0.2508	11.1791
0.0046	6.6934	5000	0.2509	11.2734
0.0037	7.3628	5500	0.2554	11.3676
0.0015	8.0321	6000	0.2524	11.2420
0.0013	8.7015	6500	0.2638	11.1791
0.0009	9.3708	7000	0.2653	11.2106
0.0008	10.0402	7500	0.2667	11.0221
0.0009	10.7095	8000	0.2702	11.0378
0.0006	11.3788	8500	0.2734	10.9750
0.0005	12.0482	9000	0.2757	11.0378
0.0006	12.7175	9500	0.2772	11.1163
0.0005	13.3869	10000	0.2778	11.1006

Framework versions

• Transformers 4.45.0
• Pytorch 2.9.1+cu130
• Datasets 3.0.1
• Tokenizers 0.20.0