pulaar-V2

This model is a fine-tuned version of abdouaziiz/fpulaar_v3 on the abdouaziz/pul_V2 dataset. It achieves the following results on the evaluation set:

• Loss: 0.8696
• Wer: 0.3605

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: 2e-05
• train_batch_size: 32
• eval_batch_size: 32
• seed: 42
• 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_steps: 50
• training_steps: 12000
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss	Wer
0.9023	2.1368	500	0.6921	0.3786
0.2246	4.2735	1000	0.7561	0.3686
0.0453	6.4103	1500	0.8414	0.3664
0.0139	8.5470	2000	0.8696	0.3605
0.0065	10.6838	2500	0.9175	0.3627
0.0049	12.8205	3000	0.9408	0.3605
0.0045	14.9573	3500	0.9553	0.3607

Framework versions

• Transformers 4.46.3
• Pytorch 2.4.1+cu121
• Datasets 3.1.0
• Tokenizers 0.20.3