Update app.py
Browse files
app.py
CHANGED
@@ -1,3 +1,35 @@
|
|
1 |
import gradio as gr
|
|
|
|
|
2 |
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
+
import jax
|
3 |
+
from diffusers import FlaxStableDiffusionPipeline
|
4 |
|
5 |
+
pipeline, pipeline_params = FlaxStableDiffusionPipeline.from_pretrained(
|
6 |
+
"bguisard/stable-diffusion-nano",
|
7 |
+
)
|
8 |
+
|
9 |
+
prng_seed = jax.random.PRNGKey(0)
|
10 |
+
inference_steps = 50
|
11 |
+
|
12 |
+
|
13 |
+
def generate_image(prompt: str):
|
14 |
+
prompt_ids = pipeline.prepare_inputs(prompt)
|
15 |
+
images = pipeline(
|
16 |
+
prompt_ids=prompt_ids,
|
17 |
+
params=pipeline_params,
|
18 |
+
prng_seed=prng_seed,
|
19 |
+
height=128,
|
20 |
+
width=128,
|
21 |
+
num_inference_steps=inference_steps,
|
22 |
+
jit=False,
|
23 |
+
).images
|
24 |
+
pil_imgs = pipeline.numpy_to_pil(images)
|
25 |
+
return pil_imgs[0]
|
26 |
+
|
27 |
+
|
28 |
+
app = gr.Interface(
|
29 |
+
fn=generate_image,
|
30 |
+
inputs="text",
|
31 |
+
outputs=gr.Image(shape=(128, 128)),
|
32 |
+
examples=[["A watercolor painting of a bird"]],
|
33 |
+
)
|
34 |
+
|
35 |
+
app.launch()
|