Thecoder3281f/whisper-small-hi-nscpart123-2000

Whisper Small NSC part 1,2,3 (2000 steps) - Jarrett Er

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: 0.0702
• Wer: 2.7673

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: 0.0001
• train_batch_size: 16
• 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: 200
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.45	0.2008	50	1.2406	152.8748
0.674	0.4016	100	0.5503	37.9635
0.5474	0.6024	150	0.4958	20.0161
0.5325	0.8032	200	0.5054	18.4041
0.6039	1.0040	250	0.5225	18.9952
0.4602	1.2048	300	0.4821	18.0011
0.492	1.4056	350	0.4468	16.7920
0.45	1.6064	400	0.4248	16.0129
0.435	1.8072	450	0.3914	14.9382
0.4395	2.0080	500	0.3767	13.7829
0.2993	2.2088	550	0.3492	13.2724
0.3069	2.4096	600	0.3280	12.4127
0.2861	2.6104	650	0.2930	12.0903
0.2507	2.8112	700	0.2764	11.4723
0.2301	3.0120	750	0.2562	11.0962
0.1827	3.2129	800	0.2455	10.1827
0.1877	3.4137	850	0.2174	8.9468
0.1805	3.6145	900	0.2072	8.7587
0.1535	3.8153	950	0.2028	8.4095
0.1502	4.0161	1000	0.1726	7.1736
0.1175	4.2169	1050	0.1639	6.7706
0.1088	4.4177	1100	0.1600	6.7437
0.129	4.6185	1150	0.1489	6.1795
0.1103	4.8193	1200	0.1331	6.2063
0.0905	5.0201	1250	0.1309	5.0510
0.0709	5.2209	1300	0.1191	5.1854
0.0733	5.4217	1350	0.1109	4.1644
0.0704	5.6225	1400	0.1047	3.9226
0.0636	5.8233	1450	0.1044	3.7883
0.0572	6.0241	1500	0.0977	3.9495
0.0447	6.2249	1550	0.0931	3.9226
0.0408	6.4257	1600	0.0871	3.3047
0.0438	6.6265	1650	0.0861	3.5196
0.0396	6.8273	1700	0.0818	3.3584
0.038	7.0281	1750	0.0794	3.0360
0.0304	7.2289	1800	0.0761	2.7942
0.0332	7.4297	1850	0.0731	2.6061
0.0459	7.6305	1900	0.0717	2.5793
0.0288	7.8313	1950	0.0706	2.6599
0.0224	8.0321	2000	0.0702	2.7673

Framework versions

• PEFT 0.14.0
• Transformers 4.45.2
• Pytorch 2.5.1+cu124
• Datasets 3.2.1.dev0
• Tokenizers 0.20.3