ComfyUI_bitsandbytes_NF4  by comfyanonymous

ComfyUI custom node for NF4 checkpoint loading

Created 1 year ago
416 stars

Top 70.6% on SourcePulse

GitHubView on GitHub
Project Summary

This custom ComfyUI node enables loading of NF4 quantized checkpoints, specifically the flux dev and flux schnell variants. It is intended for users of ComfyUI who want to leverage these memory-efficient model formats.

How It Works

The node CheckpointLoaderNF4 directly integrates bitsandbytes functionality into ComfyUI, allowing it to load NF4 quantized models. This approach aims to reduce VRAM usage compared to standard FP16 or FP32 checkpoints, facilitating the use of larger models on less powerful hardware.

Quick Start & Requirements

  • Install via ComfyUI Manager (select dev channel) or git clone the repository.
  • Requires bitsandbytes installation.
  • Ensure ComfyUI is updated.

Highlighted Details

  • Supports NF4 quantized checkpoints (flux dev, flux schnell).
  • Aims to reduce VRAM requirements.

Maintenance & Community

The project notes that it is "very likely Deprecated in favor of GGUF" and points to city96/ComfyUI-GGUF as a preferred alternative.

Licensing & Compatibility

License information is not specified in the README.

Limitations & Caveats

The project is explicitly marked as likely deprecated, recommending a newer GGUF-based solution. Compatibility is limited to specific NF4 checkpoint variants.

Health Check
Last Commit

1 year ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
5 stars in the last 30 days

Explore Similar Projects

Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems") and Jeremy Howard Jeremy Howard(Cofounder of fast.ai).

GPTFast by MDK8888

0%
687
HF Transformers accelerator for faster inference
Created 1 year ago
Updated 1 year ago
Starred by Tobi Lutke Tobi Lutke(Cofounder of Shopify), Andrej Karpathy Andrej Karpathy(Founder of Eureka Labs; Formerly at Tesla, OpenAI; Author of CS 231n), and
36 more.

unsloth by unslothai

0.6%
46k
Finetuning tool for LLMs, targeting speed and memory efficiency
Created 1 year ago
Updated 16 hours ago
Feedback? Help us improve.