flux_dev / README.md
RyanJDick's picture
Add base and bnb_nf4 model variants.
2b3ce48
|
raw
history blame
691 Bytes
---
license: other
license_name: flux-1-dev-non-commercial-license
license_link: LICENSE.md
---
# FLUX dev Quantized Models
This repo contains quantized versions of the FLUX dev transformer for use in [InvokeAI](https://github.com/invoke-ai/InvokeAI).
Contents:
- `transformer/base/` - Transformer in bfloat16 copied from [here](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/0ef5fff789c832c5c7f4e127f94c8b54bbcced44/flux1-dev.safetensors)
- `transformer/bnb_nf4/` - Transformer quantized to bitsandbytes NF4 format using [this script](https://github.com/invoke-ai/InvokeAI/blob/b8ccd53dd33aaaa6d19b780d5f11bef6142155dc/invokeai/backend/quantization/load_flux_model_bnb_nf4.py)