gopdataset_phonome_base

This model is a fine-tuned version of facebook/wav2vec2-xls-r-300m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2722
• Cer: 0.1144

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: 32
• 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	Cer
10.4672	0.84	100	12.3501	0.9750
5.6053	1.68	200	4.4724	0.9750
3.6779	2.52	300	3.5514	0.9750
3.3108	3.36	400	3.4045	0.9750
3.2684	4.2	500	3.4435	0.7828
3.1223	5.04	600	3.0123	0.7864
2.663	5.88	700	2.1177	0.6216
1.8146	6.72	800	0.9518	0.2387
1.0305	7.56	900	0.5432	0.1662
0.7835	8.4	1000	0.4268	0.1500
0.6468	9.24	1100	0.3911	0.1422
0.564	10.08	1200	0.3544	0.1378
0.5089	10.92	1300	0.3322	0.1356
0.4667	11.76	1400	0.3058	0.1277
0.4304	12.61	1500	0.2984	0.1248
0.4248	13.45	1600	0.3040	0.1270
0.4041	14.29	1700	0.2886	0.1223
0.3641	15.13	1800	0.2860	0.1215
0.3611	15.97	1900	0.2868	0.1220
0.3336	16.81	2000	0.2906	0.1217
0.3329	17.65	2100	0.2908	0.1213
0.3264	18.49	2200	0.2933	0.1204
0.3059	19.33	2300	0.2818	0.1193
0.2966	20.17	2400	0.2924	0.1196
0.2948	21.01	2500	0.2851	0.1186
0.2833	21.85	2600	0.2818	0.1181
0.2724	22.69	2700	0.2884	0.1183
0.2693	23.53	2800	0.2905	0.1179
0.2593	24.37	2900	0.2894	0.1184
0.2515	25.21	3000	0.2931	0.1169
0.2487	26.05	3100	0.2915	0.1176
0.2518	26.89	3200	0.2900	0.1176
0.2467	27.73	3300	0.2934	0.1175
0.246	28.57	3400	0.2965	0.1182
0.2537	29.41	3500	0.2948	0.1183

Framework versions

• Transformers 4.17.0
• Pytorch 2.4.0
• Datasets 1.18.3
• Tokenizers 0.20.3