w2v-bert-2.0-luo_cv_fleurs_19h-v4

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the CLEAR-GLOBAL/LUO_19H - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.2866
• Wer: 0.3289
• Cer: 0.0998

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: 1e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 64
• optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.025
• training_steps: 100000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.5991	6.4935	1000	0.6712	0.5595	0.1797
0.231	12.9870	2000	0.3213	0.3638	0.1045
0.1231	19.4805	3000	0.2866	0.3285	0.0990
0.0514	25.9740	4000	0.2907	0.3122	0.0961
0.0294	32.4675	5000	0.3262	0.3073	0.0932
0.0264	38.9610	6000	0.3543	0.3047	0.0945
0.0116	45.4545	7000	0.3592	0.3104	0.0963
0.009	51.9481	8000	0.3849	0.3355	0.0949

Framework versions

• Transformers 4.48.1
• Pytorch 2.6.0+cu124
• Datasets 3.5.0
• Tokenizers 0.21.1