food → Text refers to food, meals, beverages, ingredients, or edible items.
not_food → Text refers to non-edible objects, places, activities, concepts, or entities.

The model is intended as an educational and experimental NLP project demonstrating the complete workflow of:

Synthetic dataset generation
Dataset preparation using Hugging Face Datasets
Tokenization with DistilBERT Tokenizer
Fine-tuning using Hugging Face Transformers
Model evaluation and deployment on Hugging Face Hub

Base Model

This model is fine-tuned from:

distilbert-base-uncased

DistilBERT is a lightweight Transformer model that retains much of BERT's language understanding capability while requiring fewer parameters and faster inference.

Dataset

The model was trained on:

Food vs Not Food Synthetic Dataset

Dataset Link:

https://huggingface.co/datasets/ashutosh-kedar/food_not_food_synthetic_dataset

The dataset consists of text samples labeled as:

Label	Description
food	Sentence refers to an edible item
not_food	Sentence refers to a non-edible item

Example Inputs

Food

"A slice of pepperoni pizza."
"Fresh orange juice was served."
"The chef prepared vegetable biryani."

Not Food

"The laptop screen is cracked."
"A red bicycle is parked outside."
"The meeting starts at 2 PM."

Intended Uses

This model can be used for:

Binary text classification
NLP learning projects
Hugging Face fine-tuning demonstrations
Dataset validation experiments
Educational purposes

Training Procedure

Preprocessing

Tokenizer: DistilBERT Tokenizer
Truncation enabled
Padding applied during batching

Fine-Tuning

The model was fine-tuned using the Hugging Face Transformers Trainer API.

Training notebook:

https://github.com/ashutosh-kedar/food_not_food_distil_bert_uncase_fine_tuned/blob/main/Finetuned_DistilBert_Food_Not_Food_Classification_Model_Training.ipynb

Project repository:

https://github.com/ashutosh-kedar/food_not_food_distil_bert_uncase_fine_tuned

Usage

Framework versions

Transformers 5.0.0
Pytorch 2.11.0+cu128
Datasets 4.0.0
Tokenizers 0.22.2

Load Model

python

from transformers import pipeline

classifier = pipeline(
    "text-classification",
    model="ashutosh-kedar/finetuned_distil_bert_food_not_food_classification_model"
)

result = classifier("A bowl of hot tomato soup.")
print(result)

Example Output

python

[
    {
        'label': 'food',
        'score': 0.99
    }
]

Limitations

The model was trained on a synthetic dataset.
Real-world performance may differ from benchmark results.
Predictions are limited to the binary categories food and not_food.
Performance may degrade on ambiguous or context-dependent sentences.

Author

Ashutosh Kedar

This project demonstrates the complete NLP pipeline from synthetic dataset creation to Transformer model fine-tuning and deployment using the Hugging Face ecosystem.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 0.0001
train_batch_size: 32
eval_batch_size: 32
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
num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.4674	1.0	7	0.2076	1.0
0.1237	2.0	14	0.0151	1.0
0.0105	3.0	21	0.0202	0.98
0.0026	4.0	28	0.0024	1.0
0.0013	5.0	35	0.0008	1.0
0.0009	6.0	42	0.0006	1.0
0.0007	7.0	49	0.0005	1.0
0.0006	8.0	56	0.0005	1.0
0.0006	9.0	63	0.0005	1.0
0.0006	10.0	70	0.0004	1.0