Configuration Parsing Warning: In adapter_config.json: "peft.base_model_name_or_path" must be a string
Configuration Parsing Warning: In adapter_config.json: "peft.task_type" must be a string

✏️ Doodle to Magic

📢 This Model was carried out as part of the AIKU (🇰🇷 Korea Univ Deep-Learning CLUB) Summer 2025 activities - 🥉 Bronze Prize Winner!!

📌 Github Link
Doodle-to-Magic

🏀 Deployment Link
vercel/doodle-to-magic


Model Introduction

This model aims to automatically generate Pokemon characters from personal doodles.


Datasets

Dataset Description
Pokémon dataset Pokémon illustrations with image-caption pairs

Training Info

  • Hardware: 4 × Titan XP GPUs
  • Total Steps: 10,425 steps (~15 hours)
  • Model: LoRA finetuning on control_v11p_sd15_scribble

Methodology

Data Pre-processing

Data Preprocessing Example

Training (LoRA)

Training Example

Example Results

Prompt
pokemon prompt : pokemon style, cute {input} pokemon character, no background
image image image image

license: mit

Downloads last month
5
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for hiyseo/pokemon_scribble

Adapter
(6236)
this model

Dataset used to train hiyseo/pokemon_scribble