xlsr-syntesized-turkish-4-hour-llr-2

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on an unknown dataset. It achieves the following results on the evaluation set:

• Loss: 0.3600
• Wer: 0.3908

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: 0.0025
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• 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: 200
• num_epochs: 20
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
3.8281	0.52	100	3.0184	1.0
2.9829	1.04	200	2.7322	1.0
1.9671	1.56	300	0.9112	0.9752
0.846	2.08	400	0.5559	0.7995
0.6572	2.6	500	0.4549	0.7173
0.6061	3.12	600	0.4026	0.6306
0.5277	3.65	700	0.3818	0.5686
0.482	4.17	800	0.3699	0.5591
0.423	4.69	900	0.3543	0.5409
0.4262	5.21	1000	0.3444	0.5281
0.3991	5.73	1100	0.3637	0.5318
0.3954	6.25	1200	0.3564	0.5162
0.3503	6.77	1300	0.3861	0.6283
0.341	7.29	1400	0.3381	0.4984
0.3199	7.81	1500	0.3075	0.4844
0.3068	8.33	1600	0.2926	0.4610
0.3069	8.85	1700	0.3056	0.4693
0.2738	9.38	1800	0.3055	0.4649
0.2783	9.9	1900	0.2932	0.4420
0.2467	10.42	2000	0.2955	0.4283
0.237	10.94	2100	0.3103	0.4512
0.2284	11.46	2200	0.3228	0.4406
0.2296	11.98	2300	0.3204	0.4180
0.2078	12.5	2400	0.3491	0.4189
0.2142	13.02	2500	0.3304	0.4480
0.1983	13.54	2600	0.3364	0.4770
0.1913	14.06	2700	0.3099	0.4328
0.1763	14.58	2800	0.3127	0.3946
0.1749	15.1	2900	0.3274	0.4017
0.1616	15.62	3000	0.3419	0.3930
0.16	16.15	3100	0.3543	0.3904
0.1509	16.67	3200	0.3532	0.3951
0.1459	17.19	3300	0.3593	0.3959
0.1402	17.71	3400	0.3685	0.3964
0.1364	18.23	3500	0.3679	0.3942
0.141	18.75	3600	0.3662	0.3890
0.1348	19.27	3700	0.3656	0.3916
0.1331	19.79	3800	0.3600	0.3908

Framework versions

• Transformers 4.26.0
• Pytorch 2.1.0+cu118
• Datasets 2.9.0
• Tokenizers 0.13.3