nightmareai/k-diffusion – Run with an API on Replicate

nightmareai / k-diffusion

CLIP Guided latent k-diffusion

Cold

Public
7.4K runs
T4
GitHub
Paper
License

Run with an API

Playground API Examples README Versions

Input

text_prompt

string

Shift + Return to add a new line

Prompt

Default: "A mysterious orb by Ernst Fuchs"

sampling_mode

string

Sampling mode

Default: "DPM-2"

churn

number

The amount of noise to add during sampling

Default: 50

init_image

file

Initial image for the generation

sigma_start

integer

The starting noise level when using an init image

Default: 10

init_scale

integer

This enhances the effect of the init image, a good value is 1000.

Default: 1000

image_prompt

file

Image prompt (CLIP only)

n_steps

integer

(minimum: 50, maximum: 1000)

The number of timesteps to use

Default: 500

latent_scale

integer

Latent guidance scale, higher for stronger latent guidance.

Default: 5

clip_guidance_scale

integer

CLIP guidance scale, higher for stronger clip guidance. 0 to disable CLIP guidance.

Default: 1000

cutn

integer

The number of random crops per step.

Default: 16

cut_pow

number

Cut power

Default: 0.5

seed

integer

(minimum: 0, maximum: 4294967295)

Seed (leave empty to use a random seed)

output_steps

integer

(minimum: 0, maximum: 50)

Display an image after this many steps.

Default: 10

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

npx create-replicate --model=nightmareai/k-diffusion

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 nightmareai/k-diffusion using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

const output = await replicate.run(
  "nightmareai/k-diffusion:1535edc6074b8d68da1c1ff92786ba1465bf062c46adc2dfa5161e5d6f4a18ac",
  {
    input: {
      cutn: 16,
      churn: 50,
      cut_pow: 0.5,
      n_steps: 200,
      init_scale: 1000,
      sigma_start: 10,
      text_prompt: "A mysterious orb by Ernst Fuchs",
      latent_scale: 5,
      output_steps: 10,
      sampling_mode: "DPM-2",
      clip_guidance_scale: 1000
    }
  }
);
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 nightmareai/k-diffusion using Replicate’s API. Check out the model's schema for an overview of inputs and outputs.

output = replicate.run(
    "nightmareai/k-diffusion:1535edc6074b8d68da1c1ff92786ba1465bf062c46adc2dfa5161e5d6f4a18ac",
    input={
        "cutn": 16,
        "churn": 50,
        "cut_pow": 0.5,
        "n_steps": 200,
        "init_scale": 1000,
        "sigma_start": 10,
        "text_prompt": "A mysterious orb by Ernst Fuchs",
        "latent_scale": 5,
        "output_steps": 10,
        "sampling_mode": "DPM-2",
        "clip_guidance_scale": 1000
    }
)

# The nightmareai/k-diffusion model can stream output as it's running.
# The predict method returns an iterator, and you can iterate over that output.
for item in output:
    # https://replicate.com/nightmareai/k-diffusion/api#output-schema
    print(item)

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 nightmareai/k-diffusion 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": "1535edc6074b8d68da1c1ff92786ba1465bf062c46adc2dfa5161e5d6f4a18ac",
    "input": {
      "cutn": 16,
      "churn": 50,
      "cut_pow": 0.5,
      "n_steps": 200,
      "init_scale": 1000,
      "sigma_start": 10,
      "text_prompt": "A mysterious orb by Ernst Fuchs",
      "latent_scale": 5,
      "output_steps": 10,
      "sampling_mode": "DPM-2",
      "clip_guidance_scale": 1000
    }
  }' \
  https://api.replicate.com/v1/predictions

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

Output

{
  "completed_at": "2022-07-30T15:28:42.627066Z",
  "created_at": "2022-07-30T15:19:03.742794Z",
  "data_removed": false,
  "error": null,
  "id": "m5allb7a6ff75elmv5y74pf5j4",
  "input": {
    "cutn": 16,
    "churn": 50,
    "cut_pow": 0.5,
    "n_steps": "200",
    "init_scale": 1000,
    "sigma_start": 10,
    "text_prompt": "A mysterious orb by Ernst Fuchs",
    "latent_scale": 5,
    "output_steps": 10,
    "sampling_mode": "DPM-2",
    "clip_guidance_scale": 1000
  },
  "logs": "LatentDiffusion: Running in eps-prediction mode\nDiffusionWrapper has 872.30 M params.\nmaking attention of type 'vanilla' with 512 in_channels\nWorking with z of shape (1, 4, 32, 32) = 4096 dimensions.\nmaking attention of type 'vanilla' with 512 in_channels\n\nDownloading:   0%|          | 0.00/28.0 [00:00<?, ?B/s]\nDownloading: 100%|██████████| 28.0/28.0 [00:00<00:00, 27.9kB/s]\n\nDownloading:   0%|          | 0.00/226k [00:00<?, ?B/s]\nDownloading:  65%|██████▌   | 148k/226k [00:00<00:00, 1.29MB/s]\nDownloading: 100%|██████████| 226k/226k [00:00<00:00, 1.94MB/s]\n\nDownloading:   0%|          | 0.00/455k [00:00<?, ?B/s]\nDownloading:  27%|██▋       | 124k/455k [00:00<00:00, 1.19MB/s]\nDownloading: 100%|██████████| 455k/455k [00:00<00:00, 3.39MB/s]\n\nDownloading:   0%|          | 0.00/570 [00:00<?, ?B/s]\nDownloading: 100%|██████████| 570/570 [00:00<00:00, 557kB/s]\nusing DPM-2 sampler\nStep 0 of 200:\nStep 10 of 200:\nStep 20 of 200:\nStep 30 of 200:\nStep 40 of 200:\nStep 50 of 200:\nStep 60 of 200:\nStep 70 of 200:\nStep 80 of 200:\nStep 90 of 200:\nStep 100 of 200:\nStep 110 of 200:\nStep 120 of 200:\nStep 130 of 200:\nStep 140 of 200:\nStep 150 of 200:\nStep 160 of 200:\nStep 170 of 200:\nStep 180 of 200:\nStep 190 of 200:\nDone!",
  "metrics": {
    "predict_time": 441.825611,
    "total_time": 578.884272
  },
  "output": [
    "https://replicate.delivery/mgxm/1830a3a8-688d-4650-9c52-a29622c830b2/step_0.png",
    "https://replicate.delivery/mgxm/00d79348-ceb0-45b8-b18e-1dbc729227a7/step_0.png",
    "https://replicate.delivery/mgxm/72d3f566-a2cb-4094-b317-852133a9f5ce/step_0.png",
    "https://replicate.delivery/mgxm/62ff3f5b-0373-497c-81de-778df95f197b/step_0.png",
    "https://replicate.delivery/mgxm/661947a9-1857-4f91-bb85-d75497a3fcc3/step_0.png",
    "https://replicate.delivery/mgxm/427a27b0-e8cd-4d69-aff7-61f5e0bd575a/step_0.png",
    "https://replicate.delivery/mgxm/35b56771-53f8-48a1-8083-c9c8df99ef57/step_0.png",
    "https://replicate.delivery/mgxm/0e3373a1-9d2b-4835-945a-d48e3a3cba5e/step_0.png",
    "https://replicate.delivery/mgxm/d83cc16f-3f83-41b5-9cd4-072669be6491/step_0.png",
    "https://replicate.delivery/mgxm/283c68f1-6fab-404b-a4e7-0ef6a12457fc/step_0.png",
    "https://replicate.delivery/mgxm/626445da-5ac0-4b2e-84ef-df6088cd0e4e/step_0.png",
    "https://replicate.delivery/mgxm/89fb93f7-2c3f-41e7-9f8c-52bbeec75a1e/step_0.png",
    "https://replicate.delivery/mgxm/e4ac0fb6-3c58-40eb-99f1-88c2b2818939/step_0.png",
    "https://replicate.delivery/mgxm/49201cf1-42f8-48ab-827d-299135c53f67/step_0.png",
    "https://replicate.delivery/mgxm/576c186d-9f9b-4b97-b210-fa042b4f1f81/step_0.png",
    "https://replicate.delivery/mgxm/f9093db3-f327-48aa-bb8b-8ed824e8d637/step_0.png",
    "https://replicate.delivery/mgxm/e60cee94-8587-46ce-a965-46649b588b29/step_0.png",
    "https://replicate.delivery/mgxm/07506ae1-10f5-441a-bc81-ad776c31e0fb/step_0.png",
    "https://replicate.delivery/mgxm/bad03f80-c031-44bf-833e-00a5b66853e0/step_0.png",
    "https://replicate.delivery/mgxm/e9d96b2d-1ec9-414e-94a5-cb364f35cf92/step_0.png",
    "https://replicate.delivery/mgxm/122c091b-5a94-4702-bbc8-3bcaeaebdd08/out_0.png"
  ],
  "started_at": "2022-07-30T15:21:20.801455Z",
  "status": "succeeded",
  "urls": {
    "get": "https://api.replicate.com/v1/predictions/m5allb7a6ff75elmv5y74pf5j4",
    "cancel": "https://api.replicate.com/v1/predictions/m5allb7a6ff75elmv5y74pf5j4/cancel"
  },
  "version": "4422635ba2ddbb82255480abb99165ff8106743587c3b8753dd8c06c5e1f84c4"
}

Generated in

7 minutes 22 seconds

Tweak itReport

This example was created by a different version, nightmareai/k-diffusion:4422635b.

Examples

View more examples

Run time and cost

This model costs approximately $0.15 to run on Replicate, or 6 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 12 minutes. The predict time for this model varies significantly based on the inputs.

Readme

This demo is based a simplified diffusion codebase implemented by RiversHaveWings (Katherine Crowson).

Currently it implements CLIP guidance on Jack’s finetuned latent diffusion using OpenCLIP ViT32-LAION2b. This runs slowly due to parameters in the model requiring gradient calculation.

This is a testbed for future developments and will change often and possibly break. You have been warned.