ykleeee/wav2vec2-10epochs-3e4

wav2vec2-15epochs-3e4

This model is a fine-tuned version of facebook/wav2vec2-large-xlsr-53 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2620
• Wer: 0.2097

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: 0.0003
• train_batch_size: 4
• eval_batch_size: 4
• 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
• num_epochs: 15
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.0349	0.36	100	0.2561	0.2127
0.0248	0.72	200	0.2822	0.2133
0.0298	1.08	300	0.2751	0.2317
0.0459	1.44	400	0.3767	0.2538
0.0693	1.8	500	0.4385	0.3031
0.06	2.16	600	0.4180	0.2919
0.0663	2.52	700	0.3613	0.2707
0.0638	2.88	800	0.3703	0.2887
0.0823	3.24	900	0.3058	0.2815
0.0673	3.6	1000	0.3425	0.2793
0.049	3.96	1100	0.3508	0.2597
0.0588	4.32	1200	0.3257	0.2542
0.0527	4.68	1300	0.3560	0.2660
0.0799	5.04	1400	0.3480	0.2544
0.0526	5.4	1500	0.3653	0.2634
0.0688	5.76	1600	0.3146	0.2499
0.0651	6.12	1700	0.3179	0.2479
0.0752	6.47	1800	0.3011	0.2359
0.0833	6.83	1900	0.3444	0.2469
0.055	7.19	2000	0.3242	0.2438
0.0626	7.55	2100	0.3166	0.2343
0.0558	7.91	2200	0.3307	0.2385
0.061	8.27	2300	0.3255	0.2304
0.0529	8.63	2400	0.2864	0.2365
0.0431	8.99	2500	0.3058	0.2276
0.0737	9.35	2600	0.2943	0.2284
0.0514	9.71	2700	0.3015	0.2277
0.0893	10.07	2800	0.2709	0.2287
0.0752	10.43	2900	0.2897	0.2264
0.0604	10.79	3000	0.2772	0.2238
0.0433	11.15	3100	0.2577	0.2188
0.0325	11.51	3200	0.2671	0.2162
0.0389	11.87	3300	0.2694	0.2177
0.0487	12.23	3400	0.2835	0.2210
0.0318	12.59	3500	0.2754	0.2153
0.0268	12.95	3600	0.2668	0.2153
0.0494	13.31	3700	0.2734	0.2149
0.0645	13.67	3800	0.2450	0.2111
0.0318	14.03	3900	0.2580	0.2099
0.0454	14.39	4000	0.2656	0.2092
0.0187	14.75	4100	0.2620	0.2097

Framework versions

• Transformers 4.11.3
• Pytorch 1.10.1
• Datasets 2.9.0
• Tokenizers 0.10.3