jyoung105/flash-sdxl 🔢📝 → 🖼️

▶️ 39 runs 📅 Sep 2024 ⚙️ Cog 0.13.2 🔗 GitHub 📄 Paper ⚖️ License
fast-generation text-to-image

About

Flash Diffusion: Accelerating Any Conditional Diffusion Model for Few Steps Image Generation

Example Output

Prompt:

"A man with hoodie on, illustration"

Output

Example output

Performance Metrics

2.96s Prediction Time
215.20s Total Time
All Input Parameters 
{
  "eta": 0,
  "steps": 4,
  "width": 1024,
  "height": 1024,
  "prompt": "A man with hoodie on, illustration",
  "clip_skip": 0,
  "num_images": 1,
  "guidance_scale": 0
}
Input Parameters
eta Type: numberDefault: 0Range: 0 - 1
Stochastic parameter to control the randomness.
seed Type: integer
Random seed. Leave blank to randomize the seed.
steps Type: integerDefault: 4Range: 1 - 50
Number of denoising steps.
width Type: integerDefault: 1024Range: 1 - 2048
Width of the output image.
height Type: integerDefault: 1024Range: 1 - 2048
Height of the output image.
prompt Type: string
Input prompt, text of what you want to generate.
clip_skip Type: integerDefault: 0
Number of the layers to skip in CLIP.
num_images Type: integerDefault: 1Range: 1 - 4
Number of output images.
guidance_scale Type: numberDefault: 0Range: 0 - 20
Scale for classifier-free guidance.
negative_prompt Type: string
Input negative prompt, text of what you don't want to generate.
Output Schema

Output

Type: array • Items Type: string • Items Format: uri

Example Execution Logs 
DEVICE: cuda
DTYPE: torch.float16
Using seed: 24843
Finish setup in 0.0004892349243164062 secs.
[Debug] Prompt: A man with hoodie on, illustration, best quality, high detail, sharp focus
  0%|          | 0/4 [00:00<?, ?it/s]
 25%|██▌       | 1/4 [00:00<00:01,  2.50it/s]
 50%|█████     | 2/4 [00:00<00:00,  4.46it/s]
 75%|███████▌  | 3/4 [00:00<00:00,  5.93it/s]
100%|██████████| 4/4 [00:00<00:00,  7.01it/s]
100%|██████████| 4/4 [00:00<00:00,  5.66it/s]
Finish generation in 2.17893385887146 secs.
Version Details
Version ID
5b026f92273c153de127f7e164354d51f748989118443a6fd9a29802bbd85f36
Version Created
November 22, 2024
Run on Replicate →