Edit model card

v2 is out! https://huggingface.co/recoilme/insomnia_v2

Project by https://aiartlab.org/
A GPT2 model to generate prompts for SDXL or similar models.
Trained from the GPT2 small model. Attach a Style to affect the render further.
Trained on Syntetic prompts generated with Mistral7b.
Dataset: https://huggingface.co/datasets/recoilme/SyntheticPrompts

from transformers import pipeline, GPT2Tokenizer,GPT2LMHeadModel

checkpoint_path = "recoilme/insomnia_v2"
model = GPT2LMHeadModel.from_pretrained(checkpoint_path)
tokenizer = GPT2Tokenizer.from_pretrained('gpt2')

text_generator = pipeline('text-generation', model=model, tokenizer=tokenizer)

texts = [
    "Frog in a Harry Potter costume",
    "Cat, art by ",
    "a photo of woman underwater",
    "thunderstorms on the alien planet, very shocking",
    "Standing on the land of a new planet, the Female astronaut dances",
    "The face of the cat woman, a face beautiful, young. The head is adorned with the Egyptian crown of Bastet.",
]

for text in texts:
    print(f"Input: {text}:")
    out = text_generator(text, max_length=150, num_return_sequences=2,temperature=1.0,)
    print(f"Output 1: {out[0]['generated_text']}\n\n")
    print(f"Output 2: {out[1]['generated_text']}")
    print("\n")
Input: Frog in a Harry Potter costume:

Output 1: Frog in a Harry Potter costume, detailed with a touch of magical realism, highlight bulging eyes, slick skin, webbed feet, add atmospheric detail  misty breath, dawns first light at lilycovered pond, end with a nod to Gabriel Garca Mrquezs wizarding world.
Output 2: Frog in a Harry Potter costume, detailed and exact, persona reminiscent of a dragon or wizard duel, setting for a graveyard, atmosphere charged with suspense and anticipation, mystical creatures looming, cinematic style emphasizing amphibious grace.```
Downloads last month
13
Safetensors
Model size
124M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Dataset used to train recoilme/insomnia_v1