whisper-large-v3-atco2-asr

This model is a fine-tuned version of openai/whisper-large-v3 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7695
• Wer: 17.0374

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
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 2800
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.1388	3.57	100	0.5488	20.1957
0.0313	7.14	200	0.5830	17.5712
0.0173	10.71	300	0.5898	20.4181
0.004	14.29	400	0.6201	16.3256
0.001	17.86	500	0.6543	18.4164
0.002	21.43	600	0.6499	17.8381
0.0003	25.0	700	0.6724	17.1263
0.0002	28.57	800	0.6890	16.9929
0.0002	32.14	900	0.7012	16.8594
0.0001	35.71	1000	0.7104	16.9484
0.0001	39.29	1100	0.7178	16.9039
0.0001	42.86	1200	0.7241	17.4377
0.0001	46.43	1300	0.7305	17.3488
0.0001	50.0	1400	0.7358	17.3043
0.0001	53.57	1500	0.7407	17.3043
0.0001	57.14	1600	0.7451	17.1263
0.0001	60.71	1700	0.7495	17.2598
0.0001	64.29	1800	0.7529	17.2153
0.0001	67.86	1900	0.7563	17.2598
0.0001	71.43	2000	0.7593	17.4377
0.0001	75.0	2100	0.7612	17.3932
0.0001	78.57	2200	0.7632	17.2598
0.0	82.14	2300	0.7651	17.1263
0.0	85.71	2400	0.7666	17.0819
0.0	89.29	2500	0.7681	17.0374
0.0	92.86	2600	0.7686	17.0374
0.0	96.43	2700	0.7695	17.1263
0.0	100.0	2800	0.7695	17.0374

Framework versions

• Transformers 4.35.0
• Pytorch 2.0.1+cu117
• Datasets 2.12.0
• Tokenizers 0.14.1