mdeberta-baseline-fiqhqa-ML-384

This model is a fine-tuned version of microsoft/mdeberta-v3-base on the FiqhQA dataset. It achieves the following results on the evaluation set:

• Loss: 1.1627
• F1: 50.76
• Exact Match: 30.1

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: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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
• num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	F1	Exact Match
1.1308	1.0	258	1.0216	46.39	21.36
0.9302	2.0	516	0.9366	51.21	28.16
0.688	3.0	774	0.9667	52.56	29.13
0.5068	4.0	1032	1.0910	51.64	28.16
0.3239	5.0	1290	1.1627	50.76	30.1

Framework versions

• Transformers 4.57.1
• Pytorch 2.9.1+cu128
• Datasets 4.4.1
• Tokenizers 0.22.1