cjwbw / distilgpt2-stable-diffusion-v2

Descriptive stable diffusion prompts generation using GPT2

  • Public
  • 605 runs
  • T4

Input

string
Shift + Return to add a new line

The beginning of the prompt

Default: "a cat sitting"

integer

Number of tokens to sample from at each step with top_k sampling

Default: 50

integer

Maximum number of tokens for the output of the model

Default: 80

number

A higher temperature will produce more diverse results, but with a higher risk of less coherent text

Default: 0.9

number

Penalty value for each repetition of a token

Default: 1.2

integer

Number of results to generate

Default: 5

Output

a cat sitting in a coffee shop on the moon surface, closeup photo-real delicate ceramic porcelain sculpture of an ornate detailed mermaid as emma watson by rafael and lee madgwick with intricate white flowery details :: clockwork orange color scheme:: full body composition: volumetric lightsand foggy atmosphere. epic elegant pose!!! anime lightning!! cinematic shot a cat sitting on a bench in the park, beautiful digital art by greg rutkowski and thomas kinkade with ambient lighting. highly detailed painting of an orange colored fox standing up next to him holding white fluffy dog that is looking at it from bay windowsill while wearing headphones!!! trending artwork 4k hd render very coherent symmetrical perfect anatomy portrait!! smooth sharp focus illustration highlights light a cat sitting on a table drinking beer from an orange mushroom in the desert. digital art, trending at pixiv's - n 9 5 4 0 s horror movie still by dan mumford and thomas kinkade 8 3 2 w 7 6 hq sharp contrast intricate details award winning detailed octane render unreal engine beautiful colors masterpiece many cake foggy like ghibliak tachik a cat sitting in a chair, digital art by edward hopper and james gilleard painting highly detailed trending on arstation station cinematic lighting close up shot photograph detail cyberpunk 2077 colors octane render excellent composition dramatic studio action scene aesthetic 8k ultra hd resolution smooth sharp focus oil painted with colored paint splashes of canvas concept design matte drawing stylized vegetation forest fisheye a cat sitting on a chair reading an old book, digital art in the style of pixar trending by james gurney 1 0 8 7 studio lighting product photography highly detailed realism hd shadows depth 3 2 k. 5 d ue 6 as background bokeh sigma f / 4 - 50 mm nikon z 9 canon eos r | ghost shipwra during golden
Generated in

Run time and cost

This model costs approximately $0.020 to run on Replicate, or 50 runs per $1, but this varies depending on your inputs. It is also open source and you can run it on your own computer with Docker.

This model runs on Nvidia T4 GPU hardware. Predictions typically complete within 90 seconds. The predict time for this model varies significantly based on the inputs.

Readme

weights from: https://huggingface.co/FredZhang7/distilgpt2-stable-diffusion-v2 cog implementation: https://github.com/chenxwh/cog-themed-diffusion/tree/distilgpt2

DistilGPT2 Stable Diffusion V2

This model was trained on 2,470,000 descriptive stable diffusion prompts on the FredZhang7/distilgpt2-stable-diffusion checkpoint for another 4,270,000 steps.

Compared to other prompt generation models using GPT2, this one runs with 50% faster forwardpropagation and 40% less disk space & RAM.

Major improvements from v1 are: - 25% more variations - faster and more fluent prompt generation - cleaned training data * removed prompts that generate images with nsfw scores > 0.5 * removed duplicates, including prompts that differ by capitalization and punctuations * removed punctuations at random places * removed prompts shorter than 15 characters