moonshine_tiny_pt_v02

This model is a fine-tuned version of aomocelin/moonshine_tiny_pt on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 11.8080
• Wer: 30.5170

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: 8
• eval_batch_size: 64
• 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: 0.03
• training_steps: 15000
• mixed_precision_training: Native AMP
• label_smoothing_factor: 0.1

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.8983	0.0319	500	11.6646	31.4356
1.9170	0.0638	1000	11.6557	31.5321
1.9151	0.0957	1500	11.6914	31.3633
1.9107	0.1276	2000	11.7577	31.7097
1.8966	0.1595	2500	11.7166	31.9800
1.8583	0.1914	3000	11.8045	31.3570
1.8798	0.2233	3500	11.7613	31.5245
1.8912	0.2552	4000	11.8076	31.3988
1.8865	0.2870	4500	11.7252	31.0639
1.8911	0.3189	5000	11.8129	31.3354
1.9007	0.3508	5500	11.7334	31.3608
1.8651	0.3827	6000	11.7288	31.0969
1.9072	0.4146	6500	11.7975	31.2326
1.8440	0.4465	7000	11.7890	30.9941
1.8773	0.4784	7500	11.7768	31.0652
1.8737	0.5103	8000	11.7275	31.1146
1.8520	0.5422	8500	11.7756	31.0842
1.8446	0.5741	9000	11.7749	30.9256
1.8907	0.6060	9500	11.7880	30.6591
1.8785	0.6379	10000	11.7906	30.7733
1.8916	0.6698	10500	11.8329	30.7429
1.8674	0.7017	11000	11.8269	30.6325
1.8811	0.7336	11500	11.8236	30.4726
1.8546	0.7655	12000	11.8306	30.5741
1.9147	0.7973	12500	11.8017	30.3432
1.8926	0.8292	13000	11.8024	30.4777
1.9099	0.8611	13500	11.8052	30.4244
1.8751	0.8930	14000	11.8008	30.5373
1.9035	0.9249	14500	11.8152	30.4714
1.9062	0.9568	15000	11.8080	30.5170

Framework versions

• Transformers 5.12.1
• Pytorch 2.11.0+cu128
• Datasets 5.0.0
• Tokenizers 0.22.2