Flux.1 image generation hasn't yet implemented iGPU and Compile-mode but PRs are ready

[JamieVC]

Describe the bug
We saw the PR make iGPU run Flux.1.
openvinotoolkit/openvino#27265

We saw PRs that make compile-only mode works.
huggingface/optimum-intel#873
huggingface/optimum-intel#1101

Then, we did some tests on Flux.1 https://github.com/openvinotoolkit/openvino_notebooks/blob/latest/notebooks/flux.1-image-generation/flux.1-image-generation.ipynb

Can you help enable Flux.1 sample code to support GPU as well?
The config below is for CPU, GPU, GPU(with compile-only mode).

Expected behavior
Flux.1 sample code can support GPU

[andrei-kochin]

@JamieVC compile-only mode is not the default path and will not be present in the notebook which covers the general model usecase. GPU can run flux.1 that is why it is listed among supported

[JamieVC]

hi @andrei-kochin

Sure! understood we don't present the compile-only mode on the notebook.

However, I believe we need this config to inference on GPU. Could we modify the notebook's ipynb code to support GPU as well? Thanks

from optimum.intel.openvino import OVDiffusionPipeline
model_dir = model_base_dir / "INT4" if use_quantized_models.value else model_base_dir / "FP16"
ov_config = {"INFERENCE_PRECISION_HINT" : "f16", "ACTIVATIONS_SCALE_FACTOR": "8.0", "CACHE_DIR" : ""}
ov_pipe = OVDiffusionPipeline.from_pretrained(model_dir, device=device.value, ov_config=ov_config)

[andrei-kochin]

@JamieVC we don't "need" it for running the notebook on GPU. This option "can" be provided and that is a key point. As I've mentioned previously we demonstrate the general scenario of the notebook execution and it is aligned now with the existing content.