ModelsLab/ NSFW2FP16LORA

This version of NSFW2 offers two ways to generate images:

1. Full NSFW2 Model: Use this for a quick and easy experience.

2. Base Model + NSFW2 LORA: For more customization, combine the base model (Flux1-Dev-FP8) with the NSFW2 LORA. Think of the LORA as a special ingredient that gives the images a unique style.

Why use the LORA?

• Save space: The LORA is much smaller than the full model.

• More flexibility: Experiment with different styles by combining the LORA with other base models.

• Something you can´t do with checkpoints and unets, you can play with the strength of the lora. For some features you can enhance the image, e.g, make a woman more busty (1.2 is enough) 2 is the limit

Using the base model and the NSFW2 LORA together saves you space and makes it easier to experiment with different styles. It's like having a modular system where you can customize your cake with different toppings

This is an attempt to distribute a modification of a basic model in the LORA format instead of a full trained or merged model. Every time we download a trained model, for each model we download we download again: the basic model, the VAE, Clip-L and T5, in total about 17Gb. If you believe that GGUF is a solution, the penalty is only reduced by half. That is to say that if we download "n" models based on FluxDev, we have a redundancy of n x 17Gb. SSD vendors are very happy and grateful. Using a distribution based on LORAs, you just download the base model Flux1.Dev, with the included VAE, Clip-L and T5 once and then the specific LORA.

So, You have two ways to do the same image.

1) using the full NSFW2 FP8 full model

2) use flux1-Dev-FP8 (17Gb) as a base model + NSFW2LORA16FP

In other words, if I merged the lora with the base Flux1.Dev model, I would get the full NSFW2 model.

The base model to make the sample images was:

https://huggingface.co/lllyasviel/flux1_dev/resolve/main/flux1-dev-fp8.safetensors

If you want an image in 6-8 steps download Bytedance also include in the prompt,

https://huggingface.co/ByteDance/Hyper-SD/resolve/main/Hyper-FLUX.1-dev-8steps-lora.safetensors

For those who like GGUF, there are several cuantizations versions with 6-8 steps accelerator included

https://huggingface.co/mhnakif/flux-hyp8/tree/main

Example:

• Prompt: Concept design for a futuristic motorbike in a exposition. Hyper-realistic, 4K, HDR,

• Model: flux1-dev-fp8.safetensors

• Loras: <lora:Flux/NSFW2FP16LORA><lora:Flux/Hyper-FLUX.1-dev-8steps-lora:.125>

• Steps: 6

• CFG: 1

• Flux guidance: 3.5

• Sampler: Euler

Another way to see how the image is generated, is to click "i" on an image and see all the settings that generate the sample image

FAQ - NSFW2 Model

What is NSFW2?

NSFW2 is an image generation model that allows you to create realistic and detailed images. It's designed to be versatile and easy to use, both for beginners and advanced users.

What is a LORA?

A LORA (Lightweight Object Recognition Architecture) is a small neural network used to fine-tune a base model. In the case of NSFW2, the LORA adds a specific style or characteristics to the base model.

Why do I need the LORA?

The LORA allows you to take advantage of the capabilities of the base model (Flux1-Dev-FP8) and customize the results with the NSFW2 style. It's like adding a special ingredient to a recipe to give it a unique flavor.

What's the difference between using the full NSFW2 model and using the base model with the LORA?

• Full NSFW2 Model: It's like using a pre-made recipe. Everything is ready to use.

• Base model + LORA: It's like starting with a basic recipe and adding a special ingredient. It gives you more control over the final result.

Can I use the LORA with other base models?

It is not recommended to use the NSFW2 LORA with other base models as it has been specifically trained to work with Flux1-Dev-FP8}

What is a base model?

A base model is a pre-trained neural network on a massive dataset that serves as a starting point for creating more specific models. In the case of NSFW2, the base model (Flux1-Dev-FP8) provides the ability to generate realistic images

Think of a base model as a blank canvas. You can paint anything on it, but the canvas provides the basic surface and materials. In the case of NSFW2, the base model is the canvas and the LORA is the paint that gives it a specific style.

A base model is trained on a massive dataset to learn general patterns and features. This allows it to generate new content, such as images or text, in a creative way. However, a base model may be too general for specific tasks. That's why LORAs are used to fine-tune the base model and specialize it in a particular style or task