w2v-bert-2.0-luo_19_19h

This model is a fine-tuned version of facebook/w2v-bert-2.0 on the CLEAR-GLOBAL/LUO_19_19H - NA dataset. It achieves the following results on the evaluation set:

• Loss: 0.2606
• Wer: 0.2928
• Cer: 0.0957

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: 3e-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: 0.1
• training_steps: 100000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer	Cer
0.4378	2.4277	1000	0.7760	0.5952	0.1907
0.5611	4.8554	2000	0.4460	0.4255	0.1422
0.099	7.2819	3000	0.3206	0.3329	0.1064
0.0731	9.7096	4000	0.2828	0.3197	0.1003
0.1424	12.1361	5000	0.2606	0.2941	0.0958
0.0376	14.5638	6000	0.2693	0.2787	0.0897
0.0545	16.9915	7000	0.2793	0.2879	0.0897
0.0309	19.4180	8000	0.3190	0.2840	0.0921
0.0262	21.8457	9000	0.2881	0.2897	0.0924
0.0288	24.2722	10000	0.3093	0.3003	0.0955

Framework versions

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