Bajiyo/w2v-bert-2_7_datasets

w2v-bert-2_6_datasets

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3804
• Wer: 0.2629

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: 5e-05
• train_batch_size: 16
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• num_epochs: 10
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
1.1149	0.3795	600	0.5531	0.4947
0.2052	0.7590	1200	0.4347	0.4689
0.1576	1.1385	1800	0.3204	0.3717
0.1263	1.5180	2400	0.3928	0.4128
0.1205	1.8975	3000	0.3214	0.3607
0.0993	2.2770	3600	0.3063	0.3514
0.091	2.6565	4200	0.3078	0.3390
0.0877	3.0361	4800	0.2673	0.3165
0.0716	3.4156	5400	0.2798	0.3039
0.0681	3.7951	6000	0.2710	0.2948
0.0592	4.1746	6600	0.2728	0.3072
0.0525	4.5541	7200	0.2828	0.3133
0.0497	4.9336	7800	0.3039	0.3132
0.0402	5.3131	8400	0.2741	0.2832
0.0389	5.6926	9000	0.2837	0.3018
0.0371	6.0721	9600	0.2732	0.2830
0.0286	6.4516	10200	0.2998	0.2794
0.028	6.8311	10800	0.2904	0.2769
0.0232	7.2106	11400	0.3183	0.2752
0.0201	7.5901	12000	0.3045	0.2665
0.0197	7.9696	12600	0.3137	0.2733
0.0139	8.3491	13200	0.3438	0.2670
0.0128	8.7287	13800	0.3385	0.2651
0.0115	9.1082	14400	0.3669	0.2671
0.0079	9.4877	15000	0.3695	0.2613
0.008	9.8672	15600	0.3804	0.2629

Framework versions

• Transformers 4.44.2
• Pytorch 2.1.1+cu121
• Datasets 2.20.0
• Tokenizers 0.19.1