update layout and bug fix for temp file name length
Browse files
app.py
CHANGED
|
@@ -17,7 +17,21 @@ pipe.enable_xformers_memory_efficient_attention()
|
|
| 17 |
pipe.unet.to(memory_format=torch.channels_last)
|
| 18 |
|
| 19 |
|
| 20 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 21 |
|
| 22 |
If you're not getting what you want, there may be a few reasons:
|
| 23 |
1. Is the image not changing enough? Your guidance_scale may be too low. It should be >1. Higher guidance scale encourages to generate images
|
|
@@ -27,15 +41,48 @@ be to the input. This pipeline requires a value of at least `1`. It's possible y
|
|
| 27 |
2. Alternatively, you can toggle image_guidance_scale. Image guidance scale is to push the generated image towards the inital image. Image guidance
|
| 28 |
scale is enabled by setting `image_guidance_scale > 1`. Higher image guidance scale encourages to generate images that are closely
|
| 29 |
linked to the source image `image`, usually at the expense of lower image quality.
|
| 30 |
-
|
| 31 |
3. I have observed that rephrasing the instruction sometimes improves results (e.g., "turn him into a dog" vs. "make him a dog" vs. "as a dog").
|
| 32 |
-
|
| 33 |
4. Increasing the number of steps sometimes improves results.
|
| 34 |
-
|
| 35 |
5. Do faces look weird? The Stable Diffusion autoencoder has a hard time with faces that are small in the image. Try:
|
| 36 |
* Cropping the image so the face takes up a larger portion of the frame.
|
| 37 |
"""
|
| 38 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 39 |
def previous(image):
|
| 40 |
return image
|
| 41 |
|
|
@@ -69,7 +116,7 @@ def chat(btn_upload, image_in, in_steps, in_guidance_scale, in_img_guidance_scal
|
|
| 69 |
edited_image = pipe(prompt, image=image_hid, num_inference_steps=int(in_steps), guidance_scale=float(in_guidance_scale), image_guidance_scale=float(in_img_guidance_scale)).images[0]
|
| 70 |
if os.path.exists(img_name):
|
| 71 |
os.remove(img_name)
|
| 72 |
-
temp_img_name = img_name[:-4]+str(int(time.time()))+'.png'
|
| 73 |
# Create a file-like object
|
| 74 |
with open(temp_img_name, "wb") as fp:
|
| 75 |
# Save the image to the file-like object
|
|
@@ -109,7 +156,7 @@ def chat(btn_upload, image_in, in_steps, in_guidance_scale, in_img_guidance_scal
|
|
| 109 |
edited_image = pipe(prompt, image=image_in, num_inference_steps=int(in_steps), guidance_scale=float(in_guidance_scale), image_guidance_scale=float(in_img_guidance_scale)).images[0]
|
| 110 |
if os.path.exists(img_name):
|
| 111 |
os.remove(img_name)
|
| 112 |
-
temp_img_name = img_name[:-4]+str(int(time.time()))+'.png'
|
| 113 |
with open(temp_img_name, "wb") as fp:
|
| 114 |
# Save the image to the file-like object
|
| 115 |
edited_image.save(fp)
|
|
@@ -123,25 +170,27 @@ def chat(btn_upload, image_in, in_steps, in_guidance_scale, in_img_guidance_scal
|
|
| 123 |
|
| 124 |
|
| 125 |
#Blocks layout
|
| 126 |
-
with gr.Blocks() as demo:
|
| 127 |
-
gr.
|
| 128 |
-
|
| 129 |
-
|
| 130 |
-
|
| 131 |
-
|
| 132 |
-
|
| 133 |
-
|
| 134 |
-
|
| 135 |
-
|
| 136 |
-
|
| 137 |
-
|
| 138 |
-
|
| 139 |
-
|
| 140 |
-
|
| 141 |
-
|
| 142 |
-
|
| 143 |
-
|
| 144 |
-
|
|
|
|
|
|
|
| 145 |
chatbot = gr.Chatbot()
|
| 146 |
state_in = gr.State()
|
| 147 |
with gr.Row():
|
|
|
|
| 17 |
pipe.unet.to(memory_format=torch.channels_last)
|
| 18 |
|
| 19 |
|
| 20 |
+
|
| 21 |
+
help_text = """
|
| 22 |
+
**Note: Please be advised that a safety checker has been implemented in this public space.
|
| 23 |
+
Any attempts to generate inappropriate or NSFW images will result in the display of a black screen
|
| 24 |
+
as a precautionary measure to protect all users. We appreciate your cooperation in
|
| 25 |
+
maintaining a safe and appropriate environment for all members of our community.**
|
| 26 |
+
|
| 27 |
+
New features and bug-fixes:
|
| 28 |
+
|
| 29 |
+
1. Chat style interface
|
| 30 |
+
2. Now use **'reverse'** as prompt to get back the previous image after an unwanted edit
|
| 31 |
+
3. Use **'restart'** as prompt to get back to original image and start over!
|
| 32 |
+
4. Now you can load larger image files (~5 mb) as well
|
| 33 |
+
|
| 34 |
+
Some notes from the official [instruct-pix2pix](https://huggingface.co/spaces/timbrooks/instruct-pix2pix) Space by the authors and from the official [Diffusers docs](https://huggingface.co/docs/diffusers/main/en/api/pipelines/stable_diffusion/pix2pix) -
|
| 35 |
|
| 36 |
If you're not getting what you want, there may be a few reasons:
|
| 37 |
1. Is the image not changing enough? Your guidance_scale may be too low. It should be >1. Higher guidance scale encourages to generate images
|
|
|
|
| 41 |
2. Alternatively, you can toggle image_guidance_scale. Image guidance scale is to push the generated image towards the inital image. Image guidance
|
| 42 |
scale is enabled by setting `image_guidance_scale > 1`. Higher image guidance scale encourages to generate images that are closely
|
| 43 |
linked to the source image `image`, usually at the expense of lower image quality.
|
|
|
|
| 44 |
3. I have observed that rephrasing the instruction sometimes improves results (e.g., "turn him into a dog" vs. "make him a dog" vs. "as a dog").
|
|
|
|
| 45 |
4. Increasing the number of steps sometimes improves results.
|
|
|
|
| 46 |
5. Do faces look weird? The Stable Diffusion autoencoder has a hard time with faces that are small in the image. Try:
|
| 47 |
* Cropping the image so the face takes up a larger portion of the frame.
|
| 48 |
"""
|
| 49 |
|
| 50 |
+
css = """
|
| 51 |
+
#col-container {max-width: 580px; margin-left: auto; margin-right: auto;}
|
| 52 |
+
a {text-decoration-line: underline; font-weight: 600;}
|
| 53 |
+
.footer {
|
| 54 |
+
margin-bottom: 45px;
|
| 55 |
+
margin-top: 10px;
|
| 56 |
+
text-align: center;
|
| 57 |
+
border-bottom: 1px solid #e5e5e5;
|
| 58 |
+
}
|
| 59 |
+
.footer>p {
|
| 60 |
+
font-size: .8rem;
|
| 61 |
+
display: inline-block;
|
| 62 |
+
padding: 0 10px;
|
| 63 |
+
transform: translateY(10px);
|
| 64 |
+
background: white;
|
| 65 |
+
}
|
| 66 |
+
.dark .footer {
|
| 67 |
+
border-color: #303030;
|
| 68 |
+
}
|
| 69 |
+
.dark .footer>p {
|
| 70 |
+
background: #0b0f19;
|
| 71 |
+
}
|
| 72 |
+
.animate-spin {
|
| 73 |
+
animation: spin 1s linear infinite;
|
| 74 |
+
}
|
| 75 |
+
@keyframes spin {
|
| 76 |
+
from {
|
| 77 |
+
transform: rotate(0deg);
|
| 78 |
+
}
|
| 79 |
+
to {
|
| 80 |
+
transform: rotate(360deg);
|
| 81 |
+
}
|
| 82 |
+
}
|
| 83 |
+
"""
|
| 84 |
+
|
| 85 |
+
|
| 86 |
def previous(image):
|
| 87 |
return image
|
| 88 |
|
|
|
|
| 116 |
edited_image = pipe(prompt, image=image_hid, num_inference_steps=int(in_steps), guidance_scale=float(in_guidance_scale), image_guidance_scale=float(in_img_guidance_scale)).images[0]
|
| 117 |
if os.path.exists(img_name):
|
| 118 |
os.remove(img_name)
|
| 119 |
+
temp_img_name = img_name[:-4]+str(int(time.time()))[-4:]+'.png'
|
| 120 |
# Create a file-like object
|
| 121 |
with open(temp_img_name, "wb") as fp:
|
| 122 |
# Save the image to the file-like object
|
|
|
|
| 156 |
edited_image = pipe(prompt, image=image_in, num_inference_steps=int(in_steps), guidance_scale=float(in_guidance_scale), image_guidance_scale=float(in_img_guidance_scale)).images[0]
|
| 157 |
if os.path.exists(img_name):
|
| 158 |
os.remove(img_name)
|
| 159 |
+
temp_img_name = img_name[:-4]+str(int(time.time()))[-4:]+'.png'
|
| 160 |
with open(temp_img_name, "wb") as fp:
|
| 161 |
# Save the image to the file-like object
|
| 162 |
edited_image.save(fp)
|
|
|
|
| 170 |
|
| 171 |
|
| 172 |
#Blocks layout
|
| 173 |
+
with gr.Blocks(css="style.css") as demo:
|
| 174 |
+
with gr.Column(elem_id="col-container"):
|
| 175 |
+
gr.HTML("""<div style="text-align: center; max-width: 700px; margin: 0 auto;">
|
| 176 |
+
<div
|
| 177 |
+
style="
|
| 178 |
+
display: inline-flex;
|
| 179 |
+
align-items: center;
|
| 180 |
+
gap: 0.8rem;
|
| 181 |
+
font-size: 1.75rem;
|
| 182 |
+
"
|
| 183 |
+
>
|
| 184 |
+
<h1 style="font-weight: 900; margin-bottom: 7px; margin-top: 5px;">
|
| 185 |
+
ChatPix2Pix: Image Editing by Instructions
|
| 186 |
+
</h1>
|
| 187 |
+
</div>
|
| 188 |
+
<p style="margin-bottom: 10px; font-size: 94%">
|
| 189 |
+
For faster inference without waiting in the queue, you may duplicate the space and upgrade to GPU in settings <a href="https://huggingface.co/spaces/ysharma/InstructPix2Pix_Chatbot?duplicate=true"><img style="margin-top: 0em; margin-bottom: 0em" src="https://bit.ly/3gLdBN6" alt="Duplicate Space"></a>
|
| 190 |
+
<a href="https://huggingface.co/timbrooks/instruct-pix2pix" target="_blank">Diffusers implementation of instruct-pix2pix</a> - InstructPix2Pix: Learning to Follow Image Editing Instructions!
|
| 191 |
+
</p>
|
| 192 |
+
</div>""")
|
| 193 |
+
#gr.Markdown("""<h1><center>dummy</h1></center> """)
|
| 194 |
chatbot = gr.Chatbot()
|
| 195 |
state_in = gr.State()
|
| 196 |
with gr.Row():
|