bfirsh/vqgan-clip 🔢📝🖼️ → ❓

▶️ 6.6K runs 📅 Jun 2021 ⚙️ Cog 0.1.3+shimmed 🔗 GitHub 📄 Paper

clip image-generation text-to-image vqgan

About

Generates images with VQGAN and CLIP

Example Output

Prompt:

"the first day of the waters"

Output

[object Object]

Performance Metrics

289.50s Total Time

All Input Parameters

{
  "prompt": "the first day of the waters",
  "iterations": 250
}

Input Parameters

cutn Type: integerDefault: 64
seed Type: integerDefault: 0: Random seed
prompt Type: string: Text prompt
cut_pow Type: numberDefault: 1
step_size Type: numberDefault: 0.05
iterations Type: integerDefault: 250Range: ∞ - 250: Number of iterations
image_prompt Type: stringDefault: null: Image prompt
initial_image Type: stringDefault: null: Image to start with
initial_weight Type: numberDefault: 0

Output Schema

Type: array • Items Type: object

Example Execution Logs

i: 0, loss: 0.909588, losses: 0.909588
i: 10, loss: 0.847548, losses: 0.847548
i: 20, loss: 0.818597, losses: 0.818597
i: 30, loss: 0.808365, losses: 0.808365
i: 40, loss: 0.801111, losses: 0.801111
i: 50, loss: 0.791895, losses: 0.791895
i: 60, loss: 0.786057, losses: 0.786057
i: 70, loss: 0.774552, losses: 0.774552
i: 80, loss: 0.770548, losses: 0.770548
i: 90, loss: 0.763699, losses: 0.763699
i: 100, loss: 0.759811, losses: 0.759811
i: 110, loss: 0.752506, losses: 0.752506
i: 120, loss: 0.749852, losses: 0.749852
i: 130, loss: 0.749731, losses: 0.749731
i: 140, loss: 0.749962, losses: 0.749962
i: 150, loss: 0.747008, losses: 0.747008
i: 160, loss: 0.742415, losses: 0.742415
i: 170, loss: 0.744414, losses: 0.744414
i: 180, loss: 0.738426, losses: 0.738426
i: 190, loss: 0.73867, losses: 0.73867
i: 200, loss: 0.737181, losses: 0.737181
i: 210, loss: 0.736317, losses: 0.736317
i: 220, loss: 0.734543, losses: 0.734543
i: 230, loss: 0.729937, losses: 0.729937
i: 240, loss: 0.729107, losses: 0.729107
i: 249, loss: 0.733397, losses: 0.733397

Version Details

Version ID: 2c07b8ebd0d0803e671966e634363b8f7ab4849e6504aac48b9c60320b1eba1a
Version Created: September 24, 2021

Run on Replicate →