my_awesome_asr_mind_model

This model is a fine-tuned version of facebook/wav2vec2-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8925
• Wer: 0.4558

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: 4
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 8
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.5119	1.77	100	4.1083	1.0
3.287	3.54	200	3.2437	1.0
3.1513	5.31	300	3.1230	1.0
3.0487	7.08	400	3.0786	1.0
3.0241	8.85	500	3.0934	1.0
2.9968	10.62	600	2.9948	1.0
2.9601	12.39	700	2.9549	1.0
2.9061	14.16	800	2.8990	1.0
2.3543	15.93	900	2.2582	0.9272
1.3794	17.7	1000	1.7532	0.8179
0.8947	19.47	1100	1.2148	0.6710
0.5989	21.24	1200	1.3229	0.5579
0.5861	23.01	1300	1.4233	0.5267
0.4311	24.78	1400	1.5458	0.5104
0.3286	26.55	1500	1.6509	0.5039
0.2765	28.32	1600	1.6818	0.4948
0.2541	30.09	1700	1.7650	0.4629
0.2151	31.86	1800	1.7185	0.4460
0.1959	33.63	1900	1.9164	0.4577
0.1909	35.4	2000	1.8925	0.4558

Framework versions

• Transformers 4.37.1
• Pytorch 2.1.2+cu121
• Datasets 2.16.1
• Tokenizers 0.15.1