README.md

metadata

license: apache-2.0
base_model: google/mobilebert-uncased
tags:
  - dataset tools
  - books
  - book
  - genre
metrics:
  - f1
widget:
  - text: The Quantum Chip
    example_title: Science Fiction & Fantasy
  - text: One Dollar's Journey
    example_title: Business & Finance
  - text: Timmy The Talking Tree
    example_title: idk fiction
  - text: The Cursed Canvas
    example_title: Arts & Design
  - text: Hoops and Hegel
    example_title: Philosophy & Religion
  - text: Overview of Streams in North Dakota
    example_title: Nature
  - text: Advanced Topology
    example_title: Non-fiction/Math
  - text: Cooking Up Love
    example_title: Food & Cooking
  - text: Dr. Doolittle's Extraplanatary Commute
    example_title: Science & Technology
language:
  - en

mobilebert-uncased-title2genre

This model is a fine-tuned version of google/mobilebert-uncased for multi-label classification (18 labels).

Model description

This classifies one or more genre labels in a multi-label setting for a given book title.

The 'standard' way of interpreting the predictions is that the predicted labels for a given example are only the ones with a greater than 50% probability.

Details

Labels

There are 18 labels, these are already integrated into the config.json and should be output by the model:

"id2label": {
    "0": "History & Politics",
    "1": "Health & Medicine",
    "2": "Mystery & Thriller",
    "3": "Arts & Design",
    "4": "Self-Help & Wellness",
    "5": "Sports & Recreation",
    "6": "Non-Fiction",
    "7": "Science Fiction & Fantasy",
    "8": "Countries & Geography",
    "9": "Other",
    "10": "Nature & Environment",
    "11": "Business & Finance",
    "12": "Romance",
    "13": "Philosophy & Religion",
    "14": "Literature & Fiction",
    "15": "Science & Technology",
    "16": "Children & Young Adult",
    "17": "Food & Cooking"
  },

Eval results (validation)

It achieves the following results on the evaluation set:

• Loss: 0.2658
• F1: 0.5395

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 3e-05
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 64
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-10
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.03
• num_epochs: 10.0

Framework versions

• Transformers 4.35.0.dev0
• Pytorch 2.0.1+cpu
• Datasets 2.14.5
• Tokenizers 0.14.0