w2v-bert-bem-genbed-f-model

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the GENBED - BEM dataset. It achieves the following results on the evaluation set:

• Loss: 0.2525
• Wer: 0.4159

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• num_epochs: 30.0
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.4652	1.0959	200	0.4014	0.6859
0.3232	2.1918	400	0.4205	0.6648
0.2583	3.2877	600	0.2882	0.5244
0.2005	4.3836	800	0.2846	0.4935
0.1707	5.4795	1000	0.3055	0.5254
0.1448	6.5753	1200	0.2750	0.4459
0.1147	7.6712	1400	0.2650	0.4418
0.1086	8.7671	1600	0.2656	0.4789
0.0872	9.8630	1800	0.2525	0.4159
0.0631	10.9589	2000	0.3105	0.4286
0.0609	12.0548	2200	0.2801	0.4273
0.043	13.1507	2400	0.3265	0.4177

Framework versions

• Transformers 4.46.0.dev0
• Pytorch 2.4.1+cu121
• Datasets 3.0.1
• Tokenizers 0.20.0