wav2vec2-1

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: 0.8870
• Wer: 0.3805

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: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 1000
• num_epochs: 30
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
4.6457	0.5	500	2.8866	0.9999
2.863	1.0	1000	2.8676	1.0
1.8085	1.5	1500	0.9396	0.6602
0.8828	2.0	2000	0.7278	0.5699
0.6659	2.5	2500	0.7000	0.5401
0.6085	3.0	3000	0.7143	0.4939
0.4878	3.5	3500	0.5845	0.4717
0.4888	4.0	4000	0.6201	0.4677
0.4022	4.5	4500	0.5984	0.4532
0.3947	5.0	5000	0.5874	0.4378
0.3415	5.5	5500	0.6486	0.4405
0.3413	6.0	6000	0.5988	0.4355
0.2903	6.5	6500	0.6584	0.4304
0.3046	7.0	7000	0.6602	0.4189
0.2625	7.5	7500	0.5924	0.4235
0.2625	8.0	8000	0.6541	0.4212
0.2341	8.5	8500	0.6365	0.4171
0.2384	9.0	9000	0.6095	0.4182
0.2052	9.5	9500	0.6675	0.4091
0.2124	10.0	10000	0.6524	0.4110
0.1915	10.5	10500	0.6877	0.4122
0.1922	11.0	11000	0.6857	0.4122
0.1719	11.5	11500	0.6881	0.4056
0.1811	12.0	12000	0.6832	0.4083
0.1554	12.5	12500	0.7378	0.4103
0.163	13.0	13000	0.6940	0.4019
0.1452	13.5	13500	0.6811	0.3993
0.1457	14.0	14000	0.7216	0.4007
0.1319	14.5	14500	0.7243	0.3996
0.1367	15.0	15000	0.7332	0.4006
0.118	15.5	15500	0.7609	0.4050
0.121	16.0	16000	0.7585	0.4021
0.1096	16.5	16500	0.7583	0.4003
0.112	17.0	17000	0.7928	0.4011
0.1063	17.5	17500	0.7794	0.4038
0.1009	18.0	18000	0.7474	0.3982
0.0931	18.5	18500	0.8143	0.3980
0.0943	19.0	19000	0.7873	0.4000
0.0847	19.5	19500	0.8064	0.3991
0.0831	20.0	20000	0.8564	0.3967
0.0821	20.5	20500	0.8632	0.3956
0.0807	21.0	21000	0.8250	0.3928
0.0748	21.5	21500	0.8389	0.3949
0.0751	22.0	22000	0.8355	0.3943
0.072	22.5	22500	0.8568	0.3930
0.0696	23.0	23000	0.8396	0.3912
0.0678	23.5	23500	0.8634	0.3901
0.0671	24.0	24000	0.8576	0.3880
0.063	24.5	24500	0.8303	0.3876
0.0575	25.0	25000	0.9125	0.3847
0.0572	25.5	25500	0.8745	0.3839
0.0572	26.0	26000	0.8714	0.3844
0.0533	26.5	26500	0.8824	0.3840
0.0496	27.0	27000	0.8993	0.3830
0.0525	27.5	27500	0.8818	0.3830
0.0514	28.0	28000	0.8874	0.3819
0.0464	28.5	28500	0.8947	0.3802
0.0473	29.0	29000	0.9028	0.3805
0.048	29.5	29500	0.8899	0.3801
0.0458	30.0	30000	0.8870	0.3805

Framework versions

• Transformers 4.35.2
• Pytorch 2.1.0+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0