cuad_qa_model

This model is a fine-tuned version of answerdotai/ModernBERT-base on the cuad-qa dataset. It achieves the following results on the evaluation set:

• Loss: 1.2765
• Jaccard: 0.3298

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: 8e-05
• train_batch_size: 3
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 12
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.98) and epsilon=1e-06 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Jaccard
4.0374	0.3224	300	2.0316	0.3097
2.2477	0.6448	600	1.4555	0.3400
2.0917	0.9672	900	1.3932	0.3600
1.7293	1.2891	1200	1.2932	0.3288
1.709	1.6115	1500	1.2642	0.3280
1.7021	1.9339	1800	1.2142	0.3456
1.3928	2.2558	2100	1.2952	0.3295
1.3582	2.5782	2400	1.2791	0.3298
1.3509	2.9006	2700	1.2765	0.3298

Framework versions

• Transformers 4.49.0.dev0
• Pytorch 2.4.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0