Instructions to use catplusplus/Qwen21_Text_Encoder_Heretic with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use catplusplus/Qwen21_Text_Encoder_Heretic with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("catplusplus/Qwen21_Text_Encoder_Heretic", dtype=torch.bfloat16, device_map="cuda") prompt = "Astronaut in a jungle, cold color palette, muted colors, detailed, 8k" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
snake oil
This isn't true:
The primary purpose of this model is maximum instruction following and prompt adherence: it banishes geometric representation deflection ("internal blush" / hesitation vectors) that otherwise causes safety-tuned VLMs to corrupt diffusion conditioning on dynamic poses, human figures, athletic wear, and complex scenes.
from the author of heretic.
and if it works, provide sample images that demonstrate it π
Why should I provide anything for you? I made a model for personal use and made it available for folks to try it out and see if it seems to work for them too. If you are offering to materially compensate me for doing extra work that does not advance my personal hobbies, please reach out with the terms. Otherwise I am not interested in debating appeals to authority.
Why should I provide anything for you? ... Otherwise I am not interested in debating appeals to authority.
This has nothing to do with debating with one person. The OP posted a direct dismissal of the effectiveness of heretic ablation on VL models, and unless you provide evidence that it works (contrary to the evidence) your work here is utterly worthless and has no value. It's up to you to decide if you should provide anything; if you don't, people have no reason to take seriously anything in this repo.