Understanding RunDiffusion Pathing for Open Source Apps
When working in opensource ai applications like Kohya or ComfyUI through RunDiffusion, knowing the correct file pathing is essential. This article covers the importance of using proper pathing when uploading or accessing files within your sessions, helping you avoid common errors and streamline your workflow.
Why Pathing Matters on RunDiffusion
RunDiffusion allows users to manage datasets, checkpoints, and outputs directly from its cloud environment. When accessing these files from within an application, a consistent file pathing structure must be used. Failing to follow this format will cause nodes or scripts to fail in locating the file, halting your project.
Correct File Pathing Format: /mnt/private/Folder/Subfolder
To access uploaded files or generated outputs, always reference the full path:
/mnt/private/FolderName/SubfolderNameKey details:
- Always begin with
/mnt/private/ - Match folder and file names exactly (case-sensitive)
This structure applies universally across tools like:
- Kohya (for LoRA training datasets)
- ComfyUI (nodes that require paths)
- Auto1111 and Deforum (batch image processing etc)
Common File Locations
Checkpoints:
- Flux:
/models/checkpoints/custom/flux/ - Flux in ComfyUI Load Diffusion Model Node:
/mnt/private/models/unet - SDXL:
/models/checkpoints/custom/sdxl/ - 1.5 Checkpoints:
/models/checkpoints/custom/v1/
LoRa Files:
- Flux LoRa:
/models/lora/custom/flux/ - SDXL LoRa:
/models/lora/custom/sdxl/ - 1.5 LoRa:
/models/lora/custom/v1/
Tips to Avoid Common Pathing Issues
- Double-check spelling and capitalization.
- Use the Files browser to confirm folder structure.
- Always preview file paths in nodes before running workflows.
