Spaces:
Running
Running
Input image rescaling added
Browse files
app.py
CHANGED
|
@@ -8,6 +8,8 @@ from PIL import Image
|
|
| 8 |
|
| 9 |
print("google-generativeai:", genai.__version__)
|
| 10 |
|
|
|
|
|
|
|
| 11 |
TITLE = """<h1 align="center">Gemini Playground 💬</h1>"""
|
| 12 |
SUBTITLE = """<h2 align="center">Play with Gemini Pro and Gemini Pro Vision API</h2>"""
|
| 13 |
DUPLICATE = """
|
|
@@ -20,13 +22,13 @@ DUPLICATE = """
|
|
| 20 |
</span>
|
| 21 |
</div>
|
| 22 |
"""
|
|
|
|
| 23 |
AVATAR_IMAGES = (
|
| 24 |
None,
|
| 25 |
"https://media.roboflow.com/spaces/gemini-icon.png"
|
| 26 |
)
|
| 27 |
|
| 28 |
-
|
| 29 |
-
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
|
| 30 |
|
| 31 |
|
| 32 |
def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
|
|
@@ -35,6 +37,11 @@ def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
|
|
| 35 |
return [sequence.strip() for sequence in stop_sequences.split(",")]
|
| 36 |
|
| 37 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 38 |
def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
|
| 39 |
return "", chatbot + [[text_prompt, None]]
|
| 40 |
|
|
@@ -42,7 +49,6 @@ def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
|
|
| 42 |
def bot(
|
| 43 |
google_key: str,
|
| 44 |
image_prompt: Optional[Image.Image],
|
| 45 |
-
image_prompt_2: Optional[Image.Image],
|
| 46 |
temperature: float,
|
| 47 |
max_output_tokens: int,
|
| 48 |
stop_sequences: str,
|
|
@@ -65,7 +71,7 @@ def bot(
|
|
| 65 |
top_k=top_k,
|
| 66 |
top_p=top_p)
|
| 67 |
|
| 68 |
-
if image_prompt is None
|
| 69 |
model = genai.GenerativeModel('gemini-pro')
|
| 70 |
response = model.generate_content(
|
| 71 |
text_prompt,
|
|
@@ -73,11 +79,10 @@ def bot(
|
|
| 73 |
generation_config=generation_config)
|
| 74 |
response.resolve()
|
| 75 |
else:
|
| 76 |
-
|
| 77 |
-
contents = [content for content in contents if content is not None]
|
| 78 |
model = genai.GenerativeModel('gemini-pro-vision')
|
| 79 |
response = model.generate_content(
|
| 80 |
-
contents=
|
| 81 |
stream=True,
|
| 82 |
generation_config=generation_config)
|
| 83 |
response.resolve()
|
|
@@ -101,8 +106,7 @@ google_key_component = gr.Textbox(
|
|
| 101 |
visible=GOOGLE_API_KEY is None
|
| 102 |
)
|
| 103 |
|
| 104 |
-
image_prompt_component = gr.Image(type="pil", label="Image")
|
| 105 |
-
image_prompt_2_component = gr.Image(type="pil", label="Image")
|
| 106 |
chatbot_component = gr.Chatbot(
|
| 107 |
label='Gemini',
|
| 108 |
bubble_full_width=False,
|
|
@@ -180,7 +184,6 @@ user_inputs = [
|
|
| 180 |
bot_inputs = [
|
| 181 |
google_key_component,
|
| 182 |
image_prompt_component,
|
| 183 |
-
image_prompt_2_component,
|
| 184 |
temperature_component,
|
| 185 |
max_output_tokens_component,
|
| 186 |
stop_sequences_component,
|
|
@@ -196,10 +199,7 @@ with gr.Blocks() as demo:
|
|
| 196 |
with gr.Column():
|
| 197 |
google_key_component.render()
|
| 198 |
with gr.Row():
|
| 199 |
-
|
| 200 |
-
image_prompt_component.render()
|
| 201 |
-
with gr.Accordion("Multi Image", open=False):
|
| 202 |
-
image_prompt_2_component.render()
|
| 203 |
chatbot_component.render()
|
| 204 |
text_prompt_component.render()
|
| 205 |
run_button_component.render()
|
|
|
|
| 8 |
|
| 9 |
print("google-generativeai:", genai.__version__)
|
| 10 |
|
| 11 |
+
GOOGLE_API_KEY = os.environ.get("GOOGLE_API_KEY")
|
| 12 |
+
|
| 13 |
TITLE = """<h1 align="center">Gemini Playground 💬</h1>"""
|
| 14 |
SUBTITLE = """<h2 align="center">Play with Gemini Pro and Gemini Pro Vision API</h2>"""
|
| 15 |
DUPLICATE = """
|
|
|
|
| 22 |
</span>
|
| 23 |
</div>
|
| 24 |
"""
|
| 25 |
+
|
| 26 |
AVATAR_IMAGES = (
|
| 27 |
None,
|
| 28 |
"https://media.roboflow.com/spaces/gemini-icon.png"
|
| 29 |
)
|
| 30 |
|
| 31 |
+
IMAGE_WIDTH = 512
|
|
|
|
| 32 |
|
| 33 |
|
| 34 |
def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
|
|
|
|
| 37 |
return [sequence.strip() for sequence in stop_sequences.split(",")]
|
| 38 |
|
| 39 |
|
| 40 |
+
def preprocess_image(image: Image.Image) -> Optional[Image.Image]:
|
| 41 |
+
image_height = int(image.height * IMAGE_WIDTH / image.width)
|
| 42 |
+
return image.resize((IMAGE_WIDTH, image_height))
|
| 43 |
+
|
| 44 |
+
|
| 45 |
def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
|
| 46 |
return "", chatbot + [[text_prompt, None]]
|
| 47 |
|
|
|
|
| 49 |
def bot(
|
| 50 |
google_key: str,
|
| 51 |
image_prompt: Optional[Image.Image],
|
|
|
|
| 52 |
temperature: float,
|
| 53 |
max_output_tokens: int,
|
| 54 |
stop_sequences: str,
|
|
|
|
| 71 |
top_k=top_k,
|
| 72 |
top_p=top_p)
|
| 73 |
|
| 74 |
+
if image_prompt is None:
|
| 75 |
model = genai.GenerativeModel('gemini-pro')
|
| 76 |
response = model.generate_content(
|
| 77 |
text_prompt,
|
|
|
|
| 79 |
generation_config=generation_config)
|
| 80 |
response.resolve()
|
| 81 |
else:
|
| 82 |
+
image_prompt = preprocess_image(image_prompt)
|
|
|
|
| 83 |
model = genai.GenerativeModel('gemini-pro-vision')
|
| 84 |
response = model.generate_content(
|
| 85 |
+
contents=[text_prompt, image_prompt],
|
| 86 |
stream=True,
|
| 87 |
generation_config=generation_config)
|
| 88 |
response.resolve()
|
|
|
|
| 106 |
visible=GOOGLE_API_KEY is None
|
| 107 |
)
|
| 108 |
|
| 109 |
+
image_prompt_component = gr.Image(type="pil", label="Image", scale=1)
|
|
|
|
| 110 |
chatbot_component = gr.Chatbot(
|
| 111 |
label='Gemini',
|
| 112 |
bubble_full_width=False,
|
|
|
|
| 184 |
bot_inputs = [
|
| 185 |
google_key_component,
|
| 186 |
image_prompt_component,
|
|
|
|
| 187 |
temperature_component,
|
| 188 |
max_output_tokens_component,
|
| 189 |
stop_sequences_component,
|
|
|
|
| 199 |
with gr.Column():
|
| 200 |
google_key_component.render()
|
| 201 |
with gr.Row():
|
| 202 |
+
image_prompt_component.render()
|
|
|
|
|
|
|
|
|
|
| 203 |
chatbot_component.render()
|
| 204 |
text_prompt_component.render()
|
| 205 |
run_button_component.render()
|