huggingface / huggingface/optimum-intel

Add OpenVINO export support for FLUX.2-dev / Flux2Pipeline

Open
#1,767 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Jupyter Notebook
Stars
620
Forks
270
Avg merge
3d 10h
Merged PRs (30d)
22

Description

I’m trying to run black-forest-labs/FLUX.2-dev through Optimum-Intel’s OpenVINO export path, but export fails before conversion completes.

Environment:
optimum-intel 1.27.0
optimum 2.1.0
openvino 2026.2.0
diffusers 0.38.0
transformers 4.57.6

Model:
black-forest-labs/FLUX.2-dev

Config uses:
Flux2Pipeline
Mistral3ForConditionalGeneration
PixtralProcessor
Flux2Transformer2DModel
AutoencoderKLFlux2

Error:
AttributeError: Could not find the attribute named "vocab_size" in the normalized config.

I'm using an Intel Arc Pro B50. The price/performance and images/energy for that GPU is well above all others hence the want.

Contributor guide

No contributing guide indexed for this repository

Research direction

Start by reproducing the failure on black-forest-labs/FLUX.2-dev with Flux2Pipeline and the listed Optimum-Intel, OpenVINO, Diffusers, and Transformers versions. Trace the OpenVINO export path for the named model components and investigate the normalized-config vocab_size error. Done means the model exports successfully through that path.

Written by the indexing model from the issue text.

Assessment

Domain
machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Quiet
Clarity
Mostly clear
Newbie friendliness
45/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.