multilingual-w2v-bert-2.0

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: 0.1569
• Wer: 0.1822
• Cer: 0.0391
• Bertscore Precision: 0.9494
• Bertscore Recall: 0.9490
• Bertscore F1: 0.9492

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: 8
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 2
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Use adamw_torch_fused with betas=(0.9,0.98) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 500
• training_steps: 2000

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer	Bertscore Precision	Bertscore Recall	Bertscore F1
0.5819	0.9833	500	0.4666	0.4358	0.1111	0.8899	0.8918	0.8908
0.3311	1.9656	1000	0.3473	0.3259	0.0870	0.9206	0.9182	0.9194
0.2909	2.9479	1500	0.3059	0.2708	0.0721	0.9344	0.9338	0.9340
0.2001	3.9302	2000	0.2830	0.2520	0.0683	0.9388	0.9384	0.9386

Framework versions

• Transformers 4.57.3
• Pytorch 2.9.1+cu128
• Datasets 4.4.2
• Tokenizers 0.22.1