Spaces:
Runtime error
Runtime error
Commit
·
7c6ef59
1
Parent(s):
a771b6a
add ref image
Browse files
app.py
CHANGED
|
@@ -11,19 +11,30 @@ device="cuda"
|
|
| 11 |
model_id = "IDEA-CCNL/Taiyi-Stable-Diffusion-1B-Chinese-v0.1"
|
| 12 |
|
| 13 |
pipe_text2img = StableDiffusionPipeline.from_pretrained(model_id).to(device)
|
|
|
|
|
|
|
|
|
|
| 14 |
pipe_img2img = StableDiffusionImg2ImgPipeline(**pipe_text2img.components)
|
| 15 |
-
pipe_inpaint = StableDiffusionInpaintPipeline(**pipe_text2img.components)
|
| 16 |
|
| 17 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
|
| 19 |
-
def
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
return img
|
| 23 |
|
| 24 |
-
|
| 25 |
-
|
| 26 |
-
|
|
|
|
| 27 |
image = output.images[0]
|
| 28 |
return image
|
| 29 |
|
|
@@ -35,20 +46,29 @@ with gr.Blocks() as demo:
|
|
| 35 |
["女孩背影, 日落, 唯美插画"],
|
| 36 |
]
|
| 37 |
with gr.Row():
|
| 38 |
-
with gr.Column(scale=2, ):
|
| 39 |
-
output = gr.Image(label = '输出(output)')
|
| 40 |
-
|
| 41 |
with gr.Column(scale=1, ):
|
| 42 |
-
guide = gr.Slider(2, 15, value = 7, label = '文本引导强度(guidance scale)')
|
| 43 |
-
steps = gr.Slider(10, 30, value = 20, step = 1, label = '迭代次数(inference steps)')
|
|
|
|
| 44 |
width = gr.Slider(256, 768, value = 512, step = 64, label = '宽度(width)')
|
| 45 |
height = gr.Slider(256, 768, value = 512, step = 64, label = '高度(height)')
|
| 46 |
-
|
| 47 |
prompt = gr.Textbox(label = '提示词(prompt)')
|
| 48 |
-
submit_btn = gr.Button("
|
| 49 |
-
|
| 50 |
-
ex = gr.Examples(examples, fn=infer, inputs=[prompt, guide, steps, width, height], outputs=output)
|
| 51 |
-
|
| 52 |
-
submit_btn.click(fn = infer, inputs = [prompt, guide, steps, width, height], outputs = output)
|
| 53 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
demo.queue(concurrency_count=10).launch()
|
|
|
|
| 11 |
model_id = "IDEA-CCNL/Taiyi-Stable-Diffusion-1B-Chinese-v0.1"
|
| 12 |
|
| 13 |
pipe_text2img = StableDiffusionPipeline.from_pretrained(model_id).to(device)
|
| 14 |
+
|
| 15 |
+
# pipe_inpaint = StableDiffusionInpaintPipeline.from_pretrained(model_id).to(device) # work
|
| 16 |
+
# pipe_inpaint = StableDiffusionInpaintPipeline(**pipe_text2img.components) # not work
|
| 17 |
pipe_img2img = StableDiffusionImg2ImgPipeline(**pipe_text2img.components)
|
|
|
|
| 18 |
|
| 19 |
+
def infer_text2img(prompt, width, height):
|
| 20 |
+
output = pipe_text2img(prompt, width=width, height=height, guidance_scale=7.5, num_inference_steps=20,)
|
| 21 |
+
image = output.images[0]
|
| 22 |
+
return image
|
| 23 |
+
|
| 24 |
+
def infer_img2img(prompt, width, height, image_in, strength ):
|
| 25 |
+
init_image = image_in.convert("RGB").resize((width, height))
|
| 26 |
+
output = pipe_img2img(prompt, init_image=init_image, strength=strength, width=width, height=height, guidance_scale=7.5, num_inference_steps=20)
|
| 27 |
+
image = output.images[0]
|
| 28 |
+
return image
|
| 29 |
|
| 30 |
+
def infer_inpaint(prompt, width, height, image_in):
|
| 31 |
+
init_image = image_in["image"].convert("RGB").resize((width, height))
|
| 32 |
+
mask = image_in["mask"].convert("RGB").resize((width, height))
|
|
|
|
| 33 |
|
| 34 |
+
output = pipe_inpaint(prompt, \
|
| 35 |
+
init_image=init_image, mask_image=mask, \
|
| 36 |
+
width=width, height=height, \
|
| 37 |
+
guidance_scale=7.5, num_inference_steps=20)
|
| 38 |
image = output.images[0]
|
| 39 |
return image
|
| 40 |
|
|
|
|
| 46 |
["女孩背影, 日落, 唯美插画"],
|
| 47 |
]
|
| 48 |
with gr.Row():
|
|
|
|
|
|
|
|
|
|
| 49 |
with gr.Column(scale=1, ):
|
| 50 |
+
# guide = gr.Slider(2, 15, value = 7, label = '文本引导强度(guidance scale)')
|
| 51 |
+
# steps = gr.Slider(10, 30, value = 20, step = 1, label = '迭代次数(inference steps)')
|
| 52 |
+
image_in = gr.Image(source='upload', elem_id="image_upload", type="pil", label="参考图")
|
| 53 |
width = gr.Slider(256, 768, value = 512, step = 64, label = '宽度(width)')
|
| 54 |
height = gr.Slider(256, 768, value = 512, step = 64, label = '高度(height)')
|
| 55 |
+
strength = gr.Slider(0, 1.0, value = 0.8, step = 0.1, label = '参考图改变程度(strength)')
|
| 56 |
prompt = gr.Textbox(label = '提示词(prompt)')
|
| 57 |
+
submit_btn = gr.Button("生成图像(Generate)")
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
+
with gr.Column(scale=1, ):
|
| 60 |
+
image_out = gr.Image(label = '输出(output)')
|
| 61 |
+
ex = gr.Examples(examples, fn=infer_text2img, inputs=[prompt, width, height], outputs=image_out)
|
| 62 |
+
# with gr.Column(scale=1, ):
|
| 63 |
+
# image_in = gr.Image(source='upload', tool='sketch', elem_id="image_upload", type="pil", label="Upload")
|
| 64 |
+
# inpaint_prompt = gr.Textbox(label = '提示词(prompt)')
|
| 65 |
+
# inpaint_btn = gr.Button("图像编辑(Inpaint)")
|
| 66 |
+
# img2img_prompt = gr.Textbox(label = '提示词(prompt)')
|
| 67 |
+
# img2img_btn = gr.Button("图像编辑(Inpaint)")
|
| 68 |
+
if isinstance(image_in, Image.Image):
|
| 69 |
+
submit_btn.click(fn = infer_img2img, inputs = [prompt, width, height, image_in, strength], outputs = image_out)
|
| 70 |
+
else:
|
| 71 |
+
submit_btn.click(fn = infer_text2img, inputs = [prompt, width, height], outputs = image_out)
|
| 72 |
+
# inpaint_btn.click(fn = infer_inpaint, inputs = [inpaint_prompt, width, height, image_in], outputs = image_out)
|
| 73 |
+
# img2img_btn.click(fn = infer_img2img, inputs = [img2img_prompt, width, height, image_in], outputs = image_out)
|
| 74 |
demo.queue(concurrency_count=10).launch()
|