metadata

license: apache-2.0
base_model: distilbert/distilbert-base-uncased
library_name: transformers
tags:
  - generated_from_trainer
metrics:
  - accuracy
model-index:
  - name: text-pic-request-identifier
    results: []
datasets:
  - andriadze/pic-text-requests-synth
widget:
  - text: I'd love to see that
    output:
      - label: pic
        score: 0.99
      - label: text
        score: 0.01

text-pic-request-identifier

This model is a fine-tuned version of distilbert/distilbert-base-uncased on an synthetic dataset.

It achieves the following results on the evaluation set:

Loss: 0.0015
Accuracy: 0.9996

Model description

Model identifies if user is asking for a picture or a text.

Intended uses & limitations

Intended use for chat applications to either route the message to a text model or an image model.

Model will return 'pic' or 'text'

Training and evaluation data

Model was trained on synthetic dataset consisting of around ~25k messages. Messages were generated by different LLM's including gpt4,gpt4o,gpt4o-mini,gpt3.5-turbo

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 2e-05
train_batch_size: 16
eval_batch_size: 16
seed: 42
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: linear
num_epochs: 2

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.0391	1.0	844	0.0021	0.9996
0.0021	2.0	1688	0.0015	0.9996

Framework versions

Transformers 4.44.0
Pytorch 2.3.1
Datasets 2.21.0
Tokenizers 0.19.1

How to use

from transformers import (
    pipeline
)

picClassifier = pipeline("text-classification", model="andriadze/text-pic-request-identifier")
res = picClassifier('Can you send me a selfie?')