whisper-large-v2-atco2-asr-atcosim

This model is a fine-tuned version of openai/whisper-large-v2 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1063
• Wer: 5.5528

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: 16
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 250
• training_steps: 12644

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0503	1.97	250	0.0602	8.5346
0.0172	3.94	500	0.0602	4.1352
0.0084	5.91	750	0.0608	3.3803
0.0046	7.87	1000	0.0624	3.5523
0.0024	9.84	1250	0.0635	3.5774
0.0019	11.81	1500	0.0704	4.0933
0.0019	13.78	1750	0.0712	6.3832
0.0026	15.75	2000	0.0677	3.3635
0.0016	17.72	2250	0.0706	3.2000
0.0009	19.69	2500	0.0709	4.0597
0.0003	21.65	2750	0.0735	3.2922
0.0001	23.62	3000	0.0771	3.8836
0.0001	25.59	3250	0.0791	4.0178
0.0001	27.56	3500	0.0804	3.7913
0.0002	29.53	3750	0.0792	4.0597
0.0	31.5	4000	0.0831	4.1059
0.0	33.46	4250	0.0847	3.9507
0.0	35.43	4500	0.0859	4.1059
0.0	37.4	4750	0.0871	4.1688
0.0	39.37	5000	0.0883	4.2820
0.0	41.34	5250	0.0891	4.3449
0.0	43.31	5500	0.0898	4.5378
0.0	45.28	5750	0.0908	4.5546
0.0	47.24	6000	0.0915	4.7433
0.0	49.21	6250	0.0923	4.7643
0.0	51.18	6500	0.0933	4.8146
0.0	53.15	6750	0.0939	4.7140
0.0	55.12	7000	0.0947	4.7475
0.0	57.09	7250	0.0955	4.7266
0.0	59.06	7500	0.0962	4.8188
0.0	61.02	7750	0.0969	4.8775
0.0	62.99	8000	0.0976	5.0159
0.0	64.96	8250	0.0982	5.0872
0.0	66.93	8500	0.0989	5.1669
0.0	68.9	8750	0.0996	5.1208
0.0	70.87	9000	0.1002	5.1795
0.0	72.83	9250	0.1009	5.2969
0.0	74.8	9500	0.1014	5.2969
0.0	76.77	9750	0.1020	5.3892
0.0	78.74	10000	0.1027	5.4269
0.0	80.71	10250	0.1031	5.3431
0.0	82.68	10500	0.1038	5.4479
0.0	84.65	10750	0.1043	5.4940
0.0	86.61	11000	0.1047	5.4563
0.0	88.58	11250	0.1052	5.4857
0.0	90.55	11500	0.1055	5.4857
0.0	92.52	11750	0.1058	5.5024
0.0	94.49	12000	0.1060	5.5108
0.0	96.46	12250	0.1062	5.5150
0.0	98.43	12500	0.1063	5.5528

Framework versions

• Transformers 4.30.0.dev0
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.13.3