flx8lora / README.md
fantos's picture
Update README.md
79afe16 verified
---
title: FLUX Fast & Furious
emoji: πŸ–ΌπŸ†
colorFrom: purple
colorTo: red
sdk: gradio
sdk_version: 5.35.0
app_file: app.py
pinned: false
license: openrail++
short_description: 'FLUX 8 Step Fast & High Quality Mode'
---
I'll create comprehensive documentation for this FLUX Fast & Furious image generation code in both English and Korean.
## English Documentation
### FLUX: Fast & Furious - Hyper-Speed Image Generation
This application implements an accelerated version of the FLUX.1-dev image generation model, optimized by ByteDance's AutoML team using their Hyper-SD technology to achieve high-quality image generation in just 8 steps instead of the typical 20-50 steps.
#### Key Features
1. **Hyper-Speed Generation**
- Utilizes Hyper-SD LoRA (Low-Rank Adaptation) technology from ByteDance
- Reduces inference steps from 20-50 to just 6-25 steps (default: 8)
- Maintains high image quality while dramatically reducing generation time
- Optimized for CUDA with TF32 precision enabled for maximum performance
2. **Neon-Themed User Interface**
- Custom cyberpunk-inspired design with glowing neon effects
- Animated hover effects and dynamic visual feedback
- Dark theme with blue, cyan, and magenta color accents
- Responsive layout optimized for both desktop and mobile devices
3. **User-Friendly Features**
- **Example Prompts**: Five pre-written creative prompts covering various genres:
- Cyberpunk cityscapes
- Fantasy fairy scenes
- Epic dragon imagery
- Sci-fi space stations
- Underwater ancient cities
- **Click-to-Use Examples**: Simply click any example to instantly populate the prompt field
- **Advanced Settings**: Collapsible panel for fine-tuning generation parameters
4. **Customizable Generation Parameters**
- **Image Dimensions**: Adjustable width and height (256-1152 pixels)
- **Inference Steps**: Control speed vs. quality trade-off (6-25 steps)
- **Guidance Scale**: Adjust prompt adherence (0.0-5.0)
- **Seed Control**: Reproducible results with manual seed input
#### Technical Implementation
The application leverages cutting-edge technologies:
- **FLUX.1-dev**: State-of-the-art diffusion model from Black Forest Labs
- **Hyper-SD LoRA**: ByteDance's acceleration technology achieving 5-10x speedup
- **BFloat16 Precision**: Reduced memory usage while maintaining quality
- **Gradio Spaces**: GPU-accelerated deployment with automatic resource management
- **Custom CSS**: Neon-themed styling with glow effects and animations
The generation pipeline:
1. Loads the base FLUX.1-dev model in bfloat16 precision
2. Applies Hyper-SD LoRA weights with 0.125 scaling factor
3. Fuses LoRA weights for optimal performance
4. Generates images using accelerated inference with custom parameters
5. Outputs high-quality 1024x1024 images (default) in seconds
#### Performance Optimization
- **GPU Acceleration**: Automatic CUDA optimization with @spaces.GPU decorator
- **Memory Efficiency**: BFloat16 precision reduces VRAM usage by 50%
- **Inference Mode**: Torch inference mode and autocast for maximum speed
- **TF32 Support**: Enabled for compatible GPUs for additional speedup
- **Cached Models**: Local model caching to reduce loading times
#### Use Cases
Perfect for:
- Rapid prototyping of visual concepts
- Creative exploration with instant feedback
- Production of high-quality images for various projects
- Testing different artistic styles and compositions
- Educational purposes to understand AI image generation
---
## ν•œκΈ€ μ„€λͺ…μ„œ
### FLUX: Fast & Furious - μ΄ˆκ³ μ† 이미지 생성기
이 μ• ν”Œλ¦¬μΌ€μ΄μ…˜μ€ ByteDance의 AutoML νŒ€μ΄ κ°œλ°œν•œ Hyper-SD κΈ°μˆ μ„ ν™œμš©ν•˜μ—¬ FLUX.1-dev 이미지 생성 λͺ¨λΈμ„ κ°€μ†ν™”ν•œ λ²„μ „μœΌλ‘œ, κΈ°μ‘΄ 20-50단계가 ν•„μš”ν–ˆλ˜ 과정을 단 8λ‹¨κ³„λ‘œ 쀄여 κ³ ν’ˆμ§ˆ 이미지λ₯Ό μƒμ„±ν•©λ‹ˆλ‹€.
#### μ£Όμš” κΈ°λŠ₯
1. **μ΄ˆκ³ μ† 생성**
- ByteDance의 Hyper-SD LoRA(Low-Rank Adaptation) 기술 ν™œμš©
- μΆ”λ‘  단계λ₯Ό 20-50λ‹¨κ³„μ—μ„œ 6-25λ‹¨κ³„λ‘œ λŒ€ν­ μΆ•μ†Œ (κΈ°λ³Έκ°’: 8단계)
- 생성 μ‹œκ°„μ„ 획기적으둜 λ‹¨μΆ•ν•˜λ©΄μ„œλ„ 높은 이미지 ν’ˆμ§ˆ μœ μ§€
- μ΅œλŒ€ μ„±λŠ₯을 μœ„ν•œ TF32 정밀도가 ν™œμ„±ν™”λœ CUDA μ΅œμ ν™”
2. **λ„€μ˜¨ ν…Œλ§ˆ μ‚¬μš©μž μΈν„°νŽ˜μ΄μŠ€**
- λ°œκ΄‘ λ„€μ˜¨ νš¨κ³Όκ°€ 적용된 μ‚¬μ΄λ²„νŽ‘ν¬ μŠ€νƒ€μΌμ˜ λ§žμΆ€ν˜• λ””μžμΈ
- μ• λ‹ˆλ©”μ΄μ…˜ ν˜Έλ²„ νš¨κ³Όμ™€ 동적 μ‹œκ° ν”Όλ“œλ°±
- νŒŒλž€μƒ‰, 청둝색, λ§ˆμ  νƒ€ 색상 μ•…μ„ΌνŠΈκ°€ μžˆλŠ” 닀크 ν…Œλ§ˆ
- λ°μŠ€ν¬ν†±κ³Ό λͺ¨λ°”일 κΈ°κΈ° λͺ¨λ‘μ— μ΅œμ ν™”λœ λ°˜μ‘ν˜• λ ˆμ΄μ•„μ›ƒ
3. **μ‚¬μš©μž μΉœν™”μ  κΈ°λŠ₯**
- **μ˜ˆμ‹œ ν”„λ‘¬ν”„νŠΈ**: λ‹€μ–‘ν•œ μž₯λ₯΄λ₯Ό λ‹€λ£¨λŠ” 5개의 창의적인 ν”„λ‘¬ν”„νŠΈ 제곡:
- μ‚¬μ΄λ²„νŽ‘ν¬ λ„μ‹œ 풍경
- νŒνƒ€μ§€ μš”μ • μž₯λ©΄
- μ›…μž₯ν•œ λ“œλž˜κ³€ 이미지
- SF 우주 μ •κ±°μž₯
- μˆ˜μ€‘ κ³ λŒ€ λ„μ‹œ
- **ν΄λ¦­ν•˜μ—¬ μ‚¬μš©**: μ˜ˆμ‹œλ₯Ό ν΄λ¦­ν•˜λ©΄ μ¦‰μ‹œ ν”„λ‘¬ν”„νŠΈ ν•„λ“œμ— μž…λ ₯
- **κ³ κΈ‰ μ„€μ •**: 생성 λ§€κ°œλ³€μˆ˜ λ―Έμ„Έ 쑰정을 μœ„ν•œ 접을 수 μžˆλŠ” νŒ¨λ„
4. **λ§žμΆ€ν˜• 생성 λ§€κ°œλ³€μˆ˜**
- **이미지 크기**: μ‘°μ • κ°€λŠ₯ν•œ λ„ˆλΉ„μ™€ 높이 (256-1152 ν”½μ…€)
- **μΆ”λ‘  단계**: 속도 λŒ€ ν’ˆμ§ˆ κ· ν˜• 쑰절 (6-25단계)
- **κ°€μ΄λ˜μŠ€ μŠ€μΌ€μΌ**: ν”„λ‘¬ν”„νŠΈ μ€€μˆ˜λ„ μ‘°μ • (0.0-5.0)
- **μ‹œλ“œ μ œμ–΄**: μˆ˜λ™ μ‹œλ“œ μž…λ ₯으둜 μž¬ν˜„ κ°€λŠ₯ν•œ κ²°κ³Ό
#### 기술적 κ΅¬ν˜„
μ• ν”Œλ¦¬μΌ€μ΄μ…˜μ€ μ΅œμ²¨λ‹¨ κΈ°μˆ μ„ ν™œμš©ν•©λ‹ˆλ‹€:
- **FLUX.1-dev**: Black Forest Labs의 μ΅œμ‹  ν™•μ‚° λͺ¨λΈ
- **Hyper-SD LoRA**: 5-10λ°° 속도 ν–₯상을 λ‹¬μ„±ν•˜λŠ” ByteDance의 가속 기술
- **BFloat16 정밀도**: ν’ˆμ§ˆμ„ μœ μ§€ν•˜λ©΄μ„œ λ©”λͺ¨λ¦¬ μ‚¬μš©λŸ‰ κ°μ†Œ
- **Gradio Spaces**: μžλ™ λ¦¬μ†ŒμŠ€ 관리가 ν¬ν•¨λœ GPU 가속 배포
- **μ»€μŠ€ν…€ CSS**: λ°œκ΄‘ νš¨κ³Όμ™€ μ• λ‹ˆλ©”μ΄μ…˜μ΄ μžˆλŠ” λ„€μ˜¨ ν…Œλ§ˆ μŠ€νƒ€μΌλ§
생성 νŒŒμ΄ν”„λΌμΈ:
1. bfloat16 μ •λ°€λ„λ‘œ κΈ°λ³Έ FLUX.1-dev λͺ¨λΈ λ‘œλ“œ
2. 0.125 μŠ€μΌ€μΌλ§ νŒ©ν„°λ‘œ Hyper-SD LoRA κ°€μ€‘μΉ˜ 적용
3. 졜적 μ„±λŠ₯을 μœ„ν•œ LoRA κ°€μ€‘μΉ˜ μœ΅ν•©
4. μ‚¬μš©μž μ •μ˜ λ§€κ°œλ³€μˆ˜λ‘œ κ°€μ†ν™”λœ 좔둠을 μ‚¬μš©ν•˜μ—¬ 이미지 생성
5. λͺ‡ 초 λ§Œμ— κ³ ν’ˆμ§ˆ 1024x1024 이미지(κΈ°λ³Έκ°’) 좜λ ₯
#### μ„±λŠ₯ μ΅œμ ν™”
- **GPU 가속**: @spaces.GPU λ°μ½”λ ˆμ΄ν„°λ‘œ μžλ™ CUDA μ΅œμ ν™”
- **λ©”λͺ¨λ¦¬ νš¨μœ¨μ„±**: BFloat16 μ •λ°€λ„λ‘œ VRAM μ‚¬μš©λŸ‰ 50% κ°μ†Œ
- **μΆ”λ‘  λͺ¨λ“œ**: μ΅œλŒ€ 속도λ₯Ό μœ„ν•œ Torch μΆ”λ‘  λͺ¨λ“œμ™€ μžλ™ 캐슀트
- **TF32 지원**: ν˜Έν™˜ GPUμ—μ„œ μΆ”κ°€ 속도 ν–₯상을 μœ„ν•΄ ν™œμ„±ν™”
- **μΊμ‹œλœ λͺ¨λΈ**: λ‘œλ”© μ‹œκ°„ 단좕을 μœ„ν•œ 둜컬 λͺ¨λΈ 캐싱
#### μ‚¬μš© 사둀
λ‹€μŒκ³Ό 같은 μš©λ„μ— μ ν•©ν•©λ‹ˆλ‹€:
- μ‹œκ°μ  μ»¨μ…‰μ˜ μ‹ μ†ν•œ ν”„λ‘œν† νƒ€μ΄ν•‘
- 즉각적인 ν”Όλ“œλ°±μœΌλ‘œ 창의적 탐색
- λ‹€μ–‘ν•œ ν”„λ‘œμ νŠΈλ₯Ό μœ„ν•œ κ³ ν’ˆμ§ˆ 이미지 μ œμž‘
- λ‹€μ–‘ν•œ 예술적 μŠ€νƒ€μΌκ³Ό ꡬ성 ν…ŒμŠ€νŠΈ
- AI 이미지 생성 이해λ₯Ό μœ„ν•œ ꡐ윑 λͺ©μ 