cjwbw/stable-diffusion-v2 – Run with an API on Replicate

cjwbw / stable-diffusion-v2

sd-v2 with diffusers, test version!

Cold

Public
280.3K runs
A100 (80GB)

Run with an API

Playground API Examples README Versions

Input

prompt

string

Shift + Return to add a new line

a photo of an astronaut riding a horse on marsa photo of an astronaut riding a horse on mars

Input prompt

Default: "a photo of an astronaut riding a horse on mars"

negative_prompt

string

Shift + Return to add a new line

The prompt NOT to guide the image generation. Ignored when not using guidance

width

integer

Width of output image. Maximum size is 1024x768 or 768x1024 because of memory limits

Default: 768

height

integer

Height of output image. Maximum size is 1024x768 or 768x1024 because of memory limits

Default: 768

init_image

file

Initial image to generate variations of. Will be resized to the specified width and height

prompt_strength

number

Prompt strength when using init image. 1.0 corresponds to full destruction of information in init image

Default: 0.8

num_outputs

integer

(minimum: 1, maximum: 3)

Number of images to output. Currenly allowing 1-3, otherwise would OOM.

Default: 1

num_inference_steps

integer

(minimum: 1, maximum: 500)

Number of denoising steps

Default: 50

guidance_scale

number

(minimum: 1, maximum: 20)

Scale for classifier-free guidance

Default: 7.5

scheduler

string

Choose a scheduler. Seems only DDIM and K_EULER and DPMSolverMultistep work for sd-v2 now.

Default: "K_EULER"

seed

integer

Random seed. Leave blank to randomize the seed

Run this model in Node.js with one line of code:

npx create-replicate --model=cjwbw/stable-diffusion-v2

or set up a project from scratch

Install Replicate’s Node.js client library:

npm install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import and set up the client:

import Replicate from "replicate";

const replicate = new Replicate({
  auth: process.env.REPLICATE_API_TOKEN,
});

Run cjwbw/stable-diffusion-v2 using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "cjwbw/stable-diffusion-v2:e5e1fd333a08c8035974a01dd42f799f1cca4625aec374643d716d9ae40cf2e4",
  {
    input: {
      width: 768,
      height: 768,
      prompt: "a photo of an astronaut riding a horse on mars",
      scheduler: "K_EULER",
      num_outputs: 1,
      guidance_scale: 7.5,
      prompt_strength: 0.8,
      num_inference_steps: 50
    }
  }
);
console.log(output);

To learn more, take a look at the guide on getting started with Node.js.

Install Replicate’s Python client library:

pip install replicate

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Import the client:

import replicate

Run cjwbw/stable-diffusion-v2 using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "cjwbw/stable-diffusion-v2:e5e1fd333a08c8035974a01dd42f799f1cca4625aec374643d716d9ae40cf2e4",
    input={
        "width": 768,
        "height": 768,
        "prompt": "a photo of an astronaut riding a horse on mars",
        "scheduler": "K_EULER",
        "num_outputs": 1,
        "guidance_scale": 7.5,
        "prompt_strength": 0.8,
        "num_inference_steps": 50
    }
)
print(output)

To learn more, take a look at the guide on getting started with Python.

Set the REPLICATE_API_TOKEN environment variable:

export REPLICATE_API_TOKEN=<paste-your-token-here>

Find your API token in your account settings.

Run cjwbw/stable-diffusion-v2 using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

curl -s -X POST \
  -H "Authorization: Bearer $REPLICATE_API_TOKEN" \
  -H "Content-Type: application/json" \
  -H "Prefer: wait" \
  -d $'{
    "version": "e5e1fd333a08c8035974a01dd42f799f1cca4625aec374643d716d9ae40cf2e4",
    "input": {
      "width": 768,
      "height": 768,
      "prompt": "a photo of an astronaut riding a horse on mars",
      "scheduler": "K_EULER",
      "num_outputs": 1,
      "guidance_scale": 7.5,
      "prompt_strength": 0.8,
      "num_inference_steps": 50
    }
  }' \
  https://api.replicate.com/v1/predictions

To learn more, take a look at Replicate’s HTTP API reference docs.

You can run this model locally using Cog. First, install Cog:

brew install cog

If you don’t have Homebrew, there are other installation options available.

Run this to download the model and run it in your local environment:

cog predict r8.im/chenxwh/stable-diffusion-v2@sha256:e5e1fd333a08c8035974a01dd42f799f1cca4625aec374643d716d9ae40cf2e4 \
  -i 'width=768' \
  -i 'height=768' \
  -i 'prompt="a photo of an astronaut riding a horse on mars"' \
  -i 'scheduler="K_EULER"' \
  -i 'num_outputs=1' \
  -i 'guidance_scale=7.5' \
  -i 'prompt_strength=0.8' \
  -i 'num_inference_steps=50'

To learn more, take a look at the Cog documentation.

Run this to download the model and run it in your local environment:

docker run -d -p 5000:5000 --gpus=all r8.im/chenxwh/stable-diffusion-v2@sha256:e5e1fd333a08c8035974a01dd42f799f1cca4625aec374643d716d9ae40cf2e4
curl -s -X POST \
  -H "Content-Type: application/json" \
  -d $'{
    "input": {
      "width": 768,
      "height": 768,
      "prompt": "a photo of an astronaut riding a horse on mars",
      "scheduler": "K_EULER",
      "num_outputs": 1,
      "guidance_scale": 7.5,
      "prompt_strength": 0.8,
      "num_inference_steps": 50
    }
  }' \
  http://localhost:5000/predictions

To learn more, take a look at the Cog documentation.

Output

{
  "completed_at": "2022-11-24T19:33:41.949967Z",
  "created_at": "2022-11-24T19:33:34.583744Z",
  "data_removed": false,
  "error": null,
  "id": "w56jkpbhuvcm5brzppvzatscme",
  "input": {
    "width": 768,
    "height": 768,
    "prompt": "a photo of an astronaut riding a horse on mars",
    "scheduler": "K_EULER",
    "num_outputs": 1,
    "guidance_scale": 7.5,
    "prompt_strength": 0.8,
    "num_inference_steps": 50
  },
  "logs": "Using seed: 32989\n  0%|          | 0/50 [00:00<?, ?it/s]\n  2%|▏         | 1/50 [00:00<00:07,  6.62it/s]\n  4%|▍         | 2/50 [00:00<00:06,  7.22it/s]\n  6%|▌         | 3/50 [00:00<00:06,  7.68it/s]\n  8%|▊         | 4/50 [00:00<00:05,  7.90it/s]\n 10%|█         | 5/50 [00:00<00:05,  8.04it/s]\n 12%|█▏        | 6/50 [00:00<00:05,  8.13it/s]\n 14%|█▍        | 7/50 [00:00<00:05,  8.18it/s]\n 16%|█▌        | 8/50 [00:01<00:05,  8.21it/s]\n 18%|█▊        | 9/50 [00:01<00:04,  8.23it/s]\n 20%|██        | 10/50 [00:01<00:04,  8.24it/s]\n 22%|██▏       | 11/50 [00:01<00:04,  8.26it/s]\n 24%|██▍       | 12/50 [00:01<00:04,  8.26it/s]\n 26%|██▌       | 13/50 [00:01<00:04,  8.28it/s]\n 28%|██▊       | 14/50 [00:01<00:04,  8.28it/s]\n 30%|███       | 15/50 [00:01<00:04,  8.27it/s]\n 32%|███▏      | 16/50 [00:01<00:04,  8.27it/s]\n 34%|███▍      | 17/50 [00:02<00:03,  8.26it/s]\n 36%|███▌      | 18/50 [00:02<00:03,  8.26it/s]\n 38%|███▊      | 19/50 [00:02<00:03,  8.27it/s]\n 40%|████      | 20/50 [00:02<00:03,  8.27it/s]\n 42%|████▏     | 21/50 [00:02<00:03,  8.27it/s]\n 44%|████▍     | 22/50 [00:02<00:03,  8.28it/s]\n 46%|████▌     | 23/50 [00:02<00:03,  8.28it/s]\n 48%|████▊     | 24/50 [00:02<00:03,  8.28it/s]\n 50%|█████     | 25/50 [00:03<00:03,  8.28it/s]\n 52%|█████▏    | 26/50 [00:03<00:02,  8.28it/s]\n 54%|█████▍    | 27/50 [00:03<00:02,  8.28it/s]\n 56%|█████▌    | 28/50 [00:03<00:02,  8.28it/s]\n 58%|█████▊    | 29/50 [00:03<00:02,  8.28it/s]\n 60%|██████    | 30/50 [00:03<00:02,  8.29it/s]\n 62%|██████▏   | 31/50 [00:03<00:02,  8.28it/s]\n 64%|██████▍   | 32/50 [00:03<00:02,  8.28it/s]\n 66%|██████▌   | 33/50 [00:04<00:02,  8.28it/s]\n 68%|██████▊   | 34/50 [00:04<00:01,  8.29it/s]\n 70%|███████   | 35/50 [00:04<00:01,  8.29it/s]\n 72%|███████▏  | 36/50 [00:04<00:01,  8.30it/s]\n 74%|███████▍  | 37/50 [00:04<00:01,  8.30it/s]\n 76%|███████▌  | 38/50 [00:04<00:01,  8.30it/s]\n 78%|███████▊  | 39/50 [00:04<00:01,  8.29it/s]\n 80%|████████  | 40/50 [00:04<00:01,  8.29it/s]\n 82%|████████▏ | 41/50 [00:04<00:01,  8.29it/s]\n 84%|████████▍ | 42/50 [00:05<00:00,  8.29it/s]\n 86%|████████▌ | 43/50 [00:05<00:00,  8.29it/s]\n 88%|████████▊ | 44/50 [00:05<00:00,  8.29it/s]\n 90%|█████████ | 45/50 [00:05<00:00,  8.29it/s]\n 92%|█████████▏| 46/50 [00:05<00:00,  8.30it/s]\n 94%|█████████▍| 47/50 [00:05<00:00,  8.30it/s]\n 96%|█████████▌| 48/50 [00:05<00:00,  8.30it/s]\n 98%|█████████▊| 49/50 [00:05<00:00,  8.30it/s]\n100%|██████████| 50/50 [00:06<00:00,  8.30it/s]\n100%|██████████| 50/50 [00:06<00:00,  8.23it/s]",
  "metrics": {
    "predict_time": 7.331184,
    "total_time": 7.366223
  },
  "output": [
    "https://replicate.delivery/pbxt/Z1qXRKCfPlXuXKl2nhfb3eL3Y1TfqGylycuig9eClEOsi8aAC/out-0.png"
  ],
  "started_at": "2022-11-24T19:33:34.618783Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/w56jkpbhuvcm5brzppvzatscme",
    "cancel": "https://api.replicate.com/v1/predictions/w56jkpbhuvcm5brzppvzatscme/cancel"
  },
  "version": "867f87d0d9a9bbef873e37bdf36c4835a01411c3bb18e43653d7de63fb536096"
}

Generated in

7.3 seconds

Tweak itReport

Using seed: 32989
  0%|          | 0/50 [00:00<?, ?it/s]
  2%|▏         | 1/50 [00:00<00:07,  6.62it/s]
  4%|▍         | 2/50 [00:00<00:06,  7.22it/s]
  6%|▌         | 3/50 [00:00<00:06,  7.68it/s]
  8%|▊         | 4/50 [00:00<00:05,  7.90it/s]
 10%|█         | 5/50 [00:00<00:05,  8.04it/s]
 12%|█▏        | 6/50 [00:00<00:05,  8.13it/s]
 14%|█▍        | 7/50 [00:00<00:05,  8.18it/s]
 16%|█▌        | 8/50 [00:01<00:05,  8.21it/s]
 18%|█▊        | 9/50 [00:01<00:04,  8.23it/s]
 20%|██        | 10/50 [00:01<00:04,  8.24it/s]
 22%|██▏       | 11/50 [00:01<00:04,  8.26it/s]
 24%|██▍       | 12/50 [00:01<00:04,  8.26it/s]
 26%|██▌       | 13/50 [00:01<00:04,  8.28it/s]
 28%|██▊       | 14/50 [00:01<00:04,  8.28it/s]
 30%|███       | 15/50 [00:01<00:04,  8.27it/s]
 32%|███▏      | 16/50 [00:01<00:04,  8.27it/s]
 34%|███▍      | 17/50 [00:02<00:03,  8.26it/s]
 36%|███▌      | 18/50 [00:02<00:03,  8.26it/s]
 38%|███▊      | 19/50 [00:02<00:03,  8.27it/s]
 40%|████      | 20/50 [00:02<00:03,  8.27it/s]
 42%|████▏     | 21/50 [00:02<00:03,  8.27it/s]
 44%|████▍     | 22/50 [00:02<00:03,  8.28it/s]
 46%|████▌     | 23/50 [00:02<00:03,  8.28it/s]
 48%|████▊     | 24/50 [00:02<00:03,  8.28it/s]
 50%|█████     | 25/50 [00:03<00:03,  8.28it/s]
 52%|█████▏    | 26/50 [00:03<00:02,  8.28it/s]
 54%|█████▍    | 27/50 [00:03<00:02,  8.28it/s]
 56%|█████▌    | 28/50 [00:03<00:02,  8.28it/s]
 58%|█████▊    | 29/50 [00:03<00:02,  8.28it/s]
 60%|██████    | 30/50 [00:03<00:02,  8.29it/s]
 62%|██████▏   | 31/50 [00:03<00:02,  8.28it/s]
 64%|██████▍   | 32/50 [00:03<00:02,  8.28it/s]
 66%|██████▌   | 33/50 [00:04<00:02,  8.28it/s]
 68%|██████▊   | 34/50 [00:04<00:01,  8.29it/s]
 70%|███████   | 35/50 [00:04<00:01,  8.29it/s]
 72%|███████▏  | 36/50 [00:04<00:01,  8.30it/s]
 74%|███████▍  | 37/50 [00:04<00:01,  8.30it/s]
 76%|███████▌  | 38/50 [00:04<00:01,  8.30it/s]
 78%|███████▊  | 39/50 [00:04<00:01,  8.29it/s]
 80%|████████  | 40/50 [00:04<00:01,  8.29it/s]
 82%|████████▏ | 41/50 [00:04<00:01,  8.29it/s]
 84%|████████▍ | 42/50 [00:05<00:00,  8.29it/s]
 86%|████████▌ | 43/50 [00:05<00:00,  8.29it/s]
 88%|████████▊ | 44/50 [00:05<00:00,  8.29it/s]
 90%|█████████ | 45/50 [00:05<00:00,  8.29it/s]
 92%|█████████▏| 46/50 [00:05<00:00,  8.30it/s]
 94%|█████████▍| 47/50 [00:05<00:00,  8.30it/s]
 96%|█████████▌| 48/50 [00:05<00:00,  8.30it/s]
 98%|█████████▊| 49/50 [00:05<00:00,  8.30it/s]
100%|██████████| 50/50 [00:06<00:00,  8.30it/s]
100%|██████████| 50/50 [00:06<00:00,  8.23it/s]

This example was created by a different version, cjwbw/stable-diffusion-v2:867f87d0.

Examples

View more examples

Run time and cost

This model costs approximately $0.032 to run on Replicate, or 31 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 A100 (80GB) GPU hardware. Predictions typically complete within 23 seconds. The predict time for this model varies significantly based on the inputs.

Readme

weights from: https://huggingface.co/stabilityai/stable-diffusion-2, fp32 version code for the demo: https://github.com/chenxwh/cog-stable-diffusion/tree/sd-v2

TEST version!

This stable-diffusion-2 model is resumed from stable-diffusion-2-base (512-base-ema.ckpt) and trained for 150k steps using a v-objective on the same dataset. Resumed for another 140k steps on 768x768 images.

Compared to the mainline stable-diffusion https://replicate.com/stability-ai/stable-diffusion, the differences are: - seems only DDIM, K_EULER, and DPMSolverMultistep schedulers work properly - removed image inpainting pipeline - Default size set 768X768 - less num_output is allowed to fit to on V100