README.md

metadata

library_name: transformers
license: mit
base_model: microsoft/deberta-v3-small
tags:
  - generated_from_trainer
metrics:
  - accuracy
  - f1
  - precision
  - recall
model-index:
  - name: disi-unibo-nlp
    results: []
datasets:
  - disi-unibo-nlp/foodex2-clean

DeBERTa FoodEx2 Coder

This model is a fine-tuned version of microsoft/deberta-v3-small on the train_task1 split of the dataset foodex2-clean. It achieves the following results on the evaluation set:

• Loss: 0.0548
• Accuracy: 0.9822
• F1: 0.8507
• Precision: 0.9301
• Recall: 0.7838

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: 8
• eval_batch_size: 8
• seed: 42
• 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
• num_epochs: 3

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	F1	Precision	Recall
0.124	0.2899	1000	0.1032	0.9671	0.7090	0.8324	0.6174
0.1004	0.5799	2000	0.0855	0.9721	0.7551	0.8769	0.6631
0.0858	0.8698	3000	0.0737	0.9757	0.7873	0.9102	0.6937
0.0736	1.1598	4000	0.0696	0.9786	0.8196	0.9031	0.7502
0.0696	1.4497	5000	0.0639	0.9795	0.8294	0.8996	0.7694
0.068	1.7396	6000	0.0606	0.9812	0.8401	0.9385	0.7604
0.0634	2.0296	7000	0.0593	0.9809	0.8414	0.9123	0.7808
0.0565	2.3195	8000	0.0568	0.9820	0.8485	0.9318	0.7790
0.0584	2.6095	9000	0.0553	0.9822	0.8512	0.9296	0.7850
0.0568	2.8994	10000	0.0548	0.9822	0.8507	0.9301	0.7838

Framework versions

• Transformers 4.48.3
• Pytorch 2.5.1+cu124
• Datasets 3.3.2
• Tokenizers 0.21.0