multimodalart HF Staff commited on
Commit
3ccb734
·
1 Parent(s): d3c463a

Update app description to highlight MoE architecture and speed

Browse files
Files changed (1) hide show
  1. app.py +5 -2
app.py CHANGED
@@ -90,11 +90,14 @@ with gr.Blocks(css=CSS, theme=gr.themes.Soft()) as demo:
90
  with gr.Column(elem_id="col-container"):
91
  gr.Markdown(
92
  """
93
- # 🖼️ Nucleus-Image (ZeroGPU)
94
- Generate images with [`NucleusAI/Nucleus-Image`](https://huggingface.co/NucleusAI/Nucleus-Image).
 
95
  """
96
  )
97
 
 
 
98
  prompt = gr.Textbox(
99
  label="Prompt",
100
  placeholder="Describe the image you want to generate...",
 
90
  with gr.Column(elem_id="col-container"):
91
  gr.Markdown(
92
  """
93
+ # Nucleus-Image
94
+ **17B sparse MoE diffusion transformer** that activates only ~2B parameters per forward pass — fast, high-quality image generation.
95
+ Fully open-source: weights, training code, and dataset. [[model]](https://huggingface.co/NucleusAI/Nucleus-Image) [[paper]](https://arxiv.org/abs/2604.12163) [[code]](https://github.com/NucleusAI/Nucleus-Image)
96
  """
97
  )
98
 
99
+
100
+
101
  prompt = gr.Textbox(
102
  label="Prompt",
103
  placeholder="Describe the image you want to generate...",