w2v-bert-bem-genbed-combined-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: 0.2969
• Wer: 0.4669

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
0.6407	0.5495	200	0.6847	0.8381
0.458	1.0989	400	0.4856	0.6787
0.4014	1.6484	600	0.4310	0.6258
0.3523	2.1978	800	0.3654	0.5422
0.3298	2.7473	1000	0.3534	0.5374
0.2749	3.2967	1200	0.3402	0.5196
0.2705	3.8462	1400	0.3284	0.5250
0.249	4.3956	1600	0.3499	0.5299
0.2508	4.9451	1800	0.3512	0.5582
0.2081	5.4945	2000	0.3217	0.4808
0.2176	6.0440	2200	0.3141	0.472
0.1784	6.5934	2400	0.2969	0.4669
0.166	7.1429	2600	0.3367	0.4914
0.157	7.6923	2800	0.3206	0.4903
0.1398	8.2418	3000	0.3260	0.4617

Framework versions

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