README + Docs

Author: saddam213

Docs/Memory.md

@@ -0,0 +1,16 @@
+## Memory Modes
+
+Memory modes control **how models are placed across GPUs and CPU memory** during inference. They are designed to simplify setup while offering fine-grained control when needed.
+
+| Mode | Description |
+|------|-------------|
+| **Auto** | Automatically selects the best memory strategy for the selected device(s). |
+| **Balanced** | Distributes model weights across all available GPUs and the CPU (multi-GPU setups). |
+| **Lowest** | Sequential CPU offload for minimum GPU memory usage (slowest, lowest VRAM). |
+| **Low** | Model CPU offload with VAE slicing and tiling enabled. |
+| **Medium** | Model CPU offload without VAE slicing or tiling. |
+| **High** | All models loaded on the selected device, with VAE slicing and tiling enabled. |
+| **Highest** | All models fully loaded on the selected device (fastest, highest VRAM usage). |
+
+> **Tip:**  
+> If you’re unsure which mode to use, start with **Auto** — it handles most cases well.

Docs/Quantization.md

@@ -0,0 +1,24 @@
+## Quantization
+
+Diffuse supports **automatic INT8 quantization** during model load to reduce VRAM usage.
+
+### Supported Backends
+
+Diffuse supports two quantization backends:
+
+1. **quanto**  
+   - Used in the default environments  
+   - Supports both **CUDA** and **ROCm**
+
+2. **torchao**  
+   - Optional CUDA-only environment  
+   - Requires a custom environment build
+
+### Key Notes
+
+- Only **INT8 quantization** is currently supported
+- Quantization is **automatic** and happens during model loading
+- INT8 can reduce VRAM usage by **~30–40%**
+- Inference may be **slightly slower** when quantization is enabled
+
+> Quantization is best suited for memory-constrained systems where VRAM is more important than raw speed.

README.md

@@ -69,17 +69,21 @@ Proof of concept, Focus on core functionality.
 - ~~Portable Python installation and management~~
 - ~~Device-specific virtual environments~~
 - ~~Minimal but functional Windows UI~~
-- B~~asic Diffusers pipeline support~~
+- ~~Basic Diffusers pipeline support~~
 
 ### Beta
 Focus on usability, stability, and feature expansion.
 - ~~Fully isolated Python execution~~
 - ~~Multiple virtual environments~~
 - ~~Installer and deployment tooling~~
+- ~~Upscaling and interpolation support~~
+- ~~Extractor pipeline support~~
 - Advanced UI and workflow options
 - ControlNet support
-- Upscaling and interpolation support
-- Extractor pipeline support
+- GGUF model support
+- Weighted prompt support
+- Inpaint/Outpaint processes
+- Model Manager, download queuing, online templates
 - Stability, performance, and reliability improvements
 
 ---