This model classifies short food descriptions into semantic classes (e.g., Dish, Ingredient, Beverage). It was fine-tuned on the augmented split and evaluated on both the test subset of augmented and the original split as an external validation set.

Dataset

HF dataset: aedupuga/food-description-text (splits: augmented, original)
We discover the label set dynamically from both splits to stay consistent with the dataset card.

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_FUSED with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: linear
num_epochs: 5

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	F1
0.1081	1.0	80	0.0947	0.975	0.9788	0.975	0.9698
0.0182	2.0	160	0.0139	1.0	1.0	1.0	1.0
0.0118	3.0	240	0.0077	1.0	1.0	1.0	1.0
0.0085	4.0	320	0.0059	1.0	1.0	1.0	1.0
0.0068	5.0	400	0.0054	1.0	1.0	1.0	1.0

Framework versions

Transformers 4.56.1
Pytorch 2.8.0+cu126
Datasets 4.0.0
Tokenizers 0.22.0

Quickstart

python

from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch, numpy as np

model_id = "{cfg.HUB_REPO_ID}"
tok = AutoTokenizer.from_pretrained(model_id)
mdl = AutoModelForSequenceClassification.from_pretrained(model_id)
mdl.eval()

text = "Orange juice is made by squeezing oranges."
inputs = tok(text, return_tensors="pt", truncation=True)
with torch.no_grad():
    logits = mdl(**inputs).logits
probs = torch.softmax(logits, dim=-1)[0].detach().numpy()
pred = int(np.argmax(probs))
print(pred, mdl.config.id2label[pred], probs[pred])

yusenthebot/distilbert_food_text_artifacts

distilbert_food_text_artifacts

Model description