w2v-bert-2.0-lwazi-gpu

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

• Loss: 66.8561
• Wer: 0.4472

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: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• optimizer: Use OptimizerNames.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_steps: 200
• training_steps: 2000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
274.7911	0.0848	200	190.5785	0.9105
125.3483	0.1697	400	122.7743	0.7101
107.2834	0.2545	600	100.6818	0.6027
91.1629	0.3393	800	88.3333	0.5560
90.386	0.4242	1000	81.7339	0.5223
79.5318	0.5090	1200	77.8670	0.5000
76.4624	0.5938	1400	73.4570	0.4830
77.446	0.6787	1600	69.9153	0.4550
73.0826	0.7635	1800	68.6940	0.4572
67.0382	0.8484	2000	66.8561	0.4472

Framework versions

• Transformers 4.52.0
• Pytorch 2.8.0+cu128
• Datasets 3.6.0
• Tokenizers 0.21.4