STANDALONE.md

Standalone RMBG-2-Studio Technical Documentation

This document explains the technical details of running RMBG-2-Studio without the Pinokio orchestration platform.

Architecture

The project consists of:

1. Frontend/Backend: A single-file Gradio application (app/app.py).
2. Model: BRIA-RMBG-2.0 (BiRefNet architecture).
3. Environment Management: Scripts (setup.bat, run.bat, etc.) that handle dependencies.

Dependency Management

To solve the significant version conflict issues between Gradio 4.x and modern AI libraries, we use strict version pinning in app/requirements.txt:

Package	Version	Reason
gradio	4.44.0	Pin for gradio_imageslider compatibility.
gradio_client	1.3.0	Matches Gradio 4.44.0 exactly.
huggingface-hub	0.26.2	Prevents ImportError: HfFolder found in newer versions.
numpy	1.26.4	Version 2.0+ breaks many legacy AI libraries.
fsspec	2024.10.0	Required for compatibility with pinned HF and Gradio.
pydantic	2.10.6	Prevents serialization errors in newer Pydantic versions.

Monkeypatching

Due to a known bug in gradio_client (and specifically its interaction with newer pydantic types), app.py includes a monkeypatch at the very top:

import gradio_client.utils
_old_get_type = gradio_client.utils.get_type
def _patched_get_type(schema):
    if isinstance(schema, bool):
        return "boolean"
    return _old_get_type(schema)
gradio_client.utils.get_type = _patched_get_type

This fix prevents the TypeError: argument of type 'bool' is not iterable crash during app startup without requiring users to download additional patches.

GPU Support

The setup scripts automatically detect your hardware:

• Windows NVIDIA: Installs Torch with CUDA 12.4 support (cu124).
• Mac (M1/M2/M3): Installs Torch with Metal Performance Shaders (MPS) support.
• Linux NVIDIA: Installs Torch with CUDA 12.4.
• Linux AMD: Installs Torch with ROCm 6.2.4.

Manual Maintenance

If you need to update the model or dependencies:

1. Modify app/requirements.txt.
2. Delete the app/env folder.
3. Run setup.bat again.