Whisper-small-FullFinetuning-CV-train-test

This model is a fine-tuned version of mohammadsp99/whisper-small on the common_voice_13_0 dataset. It achieves the following results on the evaluation set:

• Loss: 0.4865
• Wer: 37.3 The evaluation was done after training

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: 50
• training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.886	0.05	100	2.1958	101.5152
0.6142	0.1	200	2.2113	110.6061
0.5544	0.15	300	2.2247	215.1515
0.4809	0.2	400	1.8149	104.5455
0.393	0.25	500	1.8802	96.9697
0.4191	0.3	600	1.9056	107.5758
0.3515	0.35	700	1.9166	89.3939
0.2671	0.4	800	1.9010	86.3636
0.2763	0.45	900	1.8574	96.9697
0.2896	0.5	1000	1.8940	95.4545
0.2201	0.55	1100	1.6264	96.9697
0.1937	0.6	1200	1.8990	98.4848
0.1787	0.65	1300	1.7999	100.0
0.1138	0.7	1400	1.8118	96.9697
0.1759	0.75	1500	1.9026	93.9394
0.1276	0.8	1600	1.8715	195.4545
0.1437	0.85	1700	1.7353	92.4242
0.1593	1.02	1800	1.7307	95.4545
0.1617	1.07	1900	1.7732	96.9697
0.1737	1.12	2000	1.7646	93.9394

Framework versions

• Transformers 4.31.0
• Pytorch 2.0.1+cu118
• Datasets 2.14.3
• Tokenizers 0.13.3