Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -4,20 +4,14 @@ import torch
|
|
| 4 |
from PIL import Image
|
| 5 |
from transformers import AutoProcessor, AutoModelForCausalLM
|
| 6 |
|
| 7 |
-
#
|
| 8 |
-
#
|
| 9 |
-
# from gradio_client import Client
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
| 13 |
|
| 14 |
# Initialize Florence model
|
| 15 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 16 |
florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True).to(device).eval()
|
| 17 |
florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True)
|
| 18 |
|
| 19 |
-
# api_key = os.getenv("HF_READ_TOKEN")
|
| 20 |
-
|
| 21 |
def generate_caption(image):
|
| 22 |
if not isinstance(image, Image.Image):
|
| 23 |
image = Image.fromarray(image)
|
|
@@ -37,32 +31,22 @@ def generate_caption(image):
|
|
| 37 |
task="<MORE_DETAILED_CAPTION>",
|
| 38 |
image_size=(image.width, image.height)
|
| 39 |
)
|
| 40 |
-
prompt =
|
| 41 |
-
print("\n\nGeneration completed!:"+ prompt)
|
| 42 |
return prompt
|
| 43 |
-
# yield prompt, None
|
| 44 |
-
# image_path = generate_image(prompt,random.randint(0, 4294967296))
|
| 45 |
-
# yield prompt, image_path
|
| 46 |
|
| 47 |
-
#
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
#
|
| 51 |
-
|
| 52 |
-
|
| 53 |
-
#
|
| 54 |
-
|
| 55 |
-
|
| 56 |
-
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
|
| 62 |
-
io = gr.Interface(generate_caption,
|
| 63 |
-
inputs=[gr.Image(label="Input Image")],
|
| 64 |
-
outputs = [gr.Textbox(label="Output Prompt", lines=2, show_copy_button = True),
|
| 65 |
-
# gr.Image(label="Output Image")
|
| 66 |
-
]
|
| 67 |
-
)
|
| 68 |
io.launch(debug=True)
|
|
|
|
| 4 |
from PIL import Image
|
| 5 |
from transformers import AutoProcessor, AutoModelForCausalLM
|
| 6 |
|
| 7 |
+
# subprocess to install flash-attn if necessary (commented for now)
|
| 8 |
+
# subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 9 |
|
| 10 |
# Initialize Florence model
|
| 11 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 12 |
florence_model = AutoModelForCausalLM.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True).to(device).eval()
|
| 13 |
florence_processor = AutoProcessor.from_pretrained('microsoft/Florence-2-base', trust_remote_code=True)
|
| 14 |
|
|
|
|
|
|
|
| 15 |
def generate_caption(image):
|
| 16 |
if not isinstance(image, Image.Image):
|
| 17 |
image = Image.fromarray(image)
|
|
|
|
| 31 |
task="<MORE_DETAILED_CAPTION>",
|
| 32 |
image_size=(image.width, image.height)
|
| 33 |
)
|
| 34 |
+
prompt = parsed_answer["<MORE_DETAILED_CAPTION>"]
|
|
|
|
| 35 |
return prompt
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
+
# Custom CSS for pink background and title styling
|
| 38 |
+
css = """
|
| 39 |
+
body {background-color: #FFC0CB;}
|
| 40 |
+
h1 {color: #800080; text-align: center; font-size: 32px; font-family: 'Comic Sans MS', cursive, sans-serif;}
|
| 41 |
+
"""
|
| 42 |
+
|
| 43 |
+
# Interface with pink background and a welcome message
|
| 44 |
+
io = gr.Interface(
|
| 45 |
+
fn=generate_caption,
|
| 46 |
+
inputs=[gr.Image(label="Input Image")],
|
| 47 |
+
outputs=[gr.Textbox(label="Output Prompt", lines=2, show_copy_button=True)],
|
| 48 |
+
title="歡迎來到我的魔法世界✨",
|
| 49 |
+
css=css
|
| 50 |
+
)
|
| 51 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
io.launch(debug=True)
|