Spaces:
Sleeping
Sleeping
huntingcarlisle
commited on
Commit
·
d3e0e2f
1
Parent(s):
4bab302
Update app.py
Browse files
app.py
CHANGED
@@ -7,6 +7,7 @@ from io import BytesIO
|
|
7 |
# from IPython.display import display
|
8 |
import base64
|
9 |
import time
|
|
|
10 |
|
11 |
|
12 |
|
@@ -21,6 +22,10 @@ def display_image(image=None,width=500,height=500):
|
|
21 |
img = image.resize((width, height))
|
22 |
return img
|
23 |
|
|
|
|
|
|
|
|
|
24 |
# API Gateway endpoint URL
|
25 |
api_url = 'https://a02q342s5b.execute-api.us-east-2.amazonaws.com/reinvent-demo-inf2-sm-20231114'
|
26 |
|
@@ -45,18 +50,50 @@ api_url = 'https://a02q342s5b.execute-api.us-east-2.amazonaws.com/reinvent-demo-
|
|
45 |
|
46 |
|
47 |
# Creating Tabs
|
48 |
-
tab1, tab2, tab3 = st.tabs(["Image Generation",
|
49 |
|
50 |
with tab1:
|
51 |
# Create two columns for layout
|
52 |
left_column, right_column = st.columns(2)
|
|
|
|
|
|
|
|
|
53 |
# ===========
|
54 |
with left_column:
|
55 |
# Define Streamlit UI elements
|
56 |
-
st.title('Stable Diffusion XL Image Generation with AWS Inferentia')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
58 |
prompt_one = st.text_area("Enter your prompt:",
|
59 |
-
|
|
|
|
|
|
|
60 |
|
61 |
# Number of inference steps
|
62 |
num_inference_steps_one = st.slider("Number of Inference Steps",
|
@@ -76,15 +113,8 @@ with tab1:
|
|
76 |
negative_prompt_one = st.text_area("Enter your negative prompt:",
|
77 |
"cartoon, graphic, text, painting, crayon, graphite, abstract glitch, blurry")
|
78 |
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
if st.button('Generate Image'):
|
86 |
-
with st.spinner(f'Generating Image with {num_inference_steps_one} iterations'):
|
87 |
-
with right_column:
|
88 |
start_time = time.time()
|
89 |
# ===============
|
90 |
# Example input data
|
@@ -94,7 +124,8 @@ with tab1:
|
|
94 |
"num_inference_steps": num_inference_steps_one,
|
95 |
"seed": seed_one,
|
96 |
"negative_prompt": negative_prompt_one
|
97 |
-
}
|
|
|
98 |
}
|
99 |
|
100 |
# Make API request
|
@@ -105,20 +136,93 @@ with tab1:
|
|
105 |
result_one = response_one.json()
|
106 |
# st.success(f"Prediction result: {result}")
|
107 |
image_one = display_image(decode_base64_image(result_one["generated_images"][0]))
|
108 |
-
|
109 |
caption=f"{prompt_one}")
|
110 |
end_time = time.time()
|
111 |
total_time = round(end_time - start_time, 2)
|
112 |
-
|
113 |
-
|
114 |
-
|
115 |
-
|
116 |
# Calculate and display the time per iteration in milliseconds
|
117 |
time_per_iteration_ms = (total_time / num_inference_steps_one)
|
118 |
-
|
119 |
else:
|
120 |
st.error(f"Error: {response_one.text}")
|
121 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
122 |
|
123 |
with tab2:
|
124 |
# ===========
|
|
|
7 |
# from IPython.display import display
|
8 |
import base64
|
9 |
import time
|
10 |
+
import random
|
11 |
|
12 |
|
13 |
|
|
|
22 |
img = image.resize((width, height))
|
23 |
return img
|
24 |
|
25 |
+
def pretty_print(messages):
|
26 |
+
for message in messages:
|
27 |
+
return f"{message['role']}: {message['content']}"
|
28 |
+
|
29 |
# API Gateway endpoint URL
|
30 |
api_url = 'https://a02q342s5b.execute-api.us-east-2.amazonaws.com/reinvent-demo-inf2-sm-20231114'
|
31 |
|
|
|
50 |
|
51 |
|
52 |
# Creating Tabs
|
53 |
+
tab1, tab2, tab3 = st.tabs(["Image Generation", "Architecture", "Code"])
|
54 |
|
55 |
with tab1:
|
56 |
# Create two columns for layout
|
57 |
left_column, right_column = st.columns(2)
|
58 |
+
|
59 |
+
with right_column:
|
60 |
+
cont = st.container()
|
61 |
+
|
62 |
# ===========
|
63 |
with left_column:
|
64 |
# Define Streamlit UI elements
|
65 |
+
st.title('Stable Diffusion XL Image Generation with AWS Inferentia 2')
|
66 |
+
|
67 |
+
sample_prompts = [
|
68 |
+
"A futuristic cityscape at sunset, cyberpunk",
|
69 |
+
"A serene landscape with mountains and a river, photorealistic style",
|
70 |
+
"An astronaut riding a horse, artistic and surreal",
|
71 |
+
"A robot playing chess in a medieval setting, high detail",
|
72 |
+
"An underwater scene with colorful coral reefs and fish, vibrant colors",
|
73 |
+
"Raccoon astronaut in space, sci-fi, future, cold color palette, muted colors, detailed, 8k",
|
74 |
+
"A lost city rediscovered in the Amazon jungle, overgrown with plants, in the style of a vintage travel poster",
|
75 |
+
"A steampunk train emitting clouds of steam as it races through a mountain pass, digital art",
|
76 |
+
"An enchanted forest with bioluminescent trees and fairies dancing, in a Studio Ghibli style",
|
77 |
+
"A portrait of an elegant alien empress with a detailed headdress, reminiscent of Art Nouveau",
|
78 |
+
"A post-apocalyptic Tokyo with nature reclaiming skyscrapers, in the style of a concept art",
|
79 |
+
"A mythical phoenix rising from ashes, vibrant colors, with a nebula in the background",
|
80 |
+
"A cybernetic wolf in a neon-lit city, cyberpunk theme, rain-drenched streets",
|
81 |
+
"A high fantasy battle scene with dragons in the sky and knights on the ground, epic scale",
|
82 |
+
"An ice castle on a lonely mountain peak, under the northern lights, fantasy illustration",
|
83 |
+
"A surreal landscape where giant flowers bloom in the desert, with a distant thunderstorm, hyperrealism"
|
84 |
+
]
|
85 |
+
|
86 |
+
def set_random_prompt():
|
87 |
+
# This function will be called when the button is clicked
|
88 |
+
random_prompt = random.choice(sample_prompts)
|
89 |
+
# Update the session state for the input field
|
90 |
+
st.session_state.prompt_one = random_prompt
|
91 |
|
92 |
prompt_one = st.text_area("Enter your prompt:",
|
93 |
+
|
94 |
+
key="prompt_one")
|
95 |
+
|
96 |
+
st.button('Random Prompt', on_click=set_random_prompt)
|
97 |
|
98 |
# Number of inference steps
|
99 |
num_inference_steps_one = st.slider("Number of Inference Steps",
|
|
|
113 |
negative_prompt_one = st.text_area("Enter your negative prompt:",
|
114 |
"cartoon, graphic, text, painting, crayon, graphite, abstract glitch, blurry")
|
115 |
|
116 |
+
if st.button('Generate Image'):
|
117 |
+
with st.spinner(f'Generating Image with {num_inference_steps_one} iterations'):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
118 |
start_time = time.time()
|
119 |
# ===============
|
120 |
# Example input data
|
|
|
124 |
"num_inference_steps": num_inference_steps_one,
|
125 |
"seed": seed_one,
|
126 |
"negative_prompt": negative_prompt_one
|
127 |
+
},
|
128 |
+
"endpoint": "huggingface-pytorch-inference-neuronx-2023-11-14-21-22-10-388"
|
129 |
}
|
130 |
|
131 |
# Make API request
|
|
|
136 |
result_one = response_one.json()
|
137 |
# st.success(f"Prediction result: {result}")
|
138 |
image_one = display_image(decode_base64_image(result_one["generated_images"][0]))
|
139 |
+
cont.image(image_one,
|
140 |
caption=f"{prompt_one}")
|
141 |
end_time = time.time()
|
142 |
total_time = round(end_time - start_time, 2)
|
143 |
+
cont.text(f"Prompt: {prompt_one}")
|
144 |
+
cont.text(f"Number of Iterations: {num_inference_steps_one}")
|
145 |
+
cont.text(f"Random Seed: {seed_one}")
|
146 |
+
cont.text(f'Total time taken: {total_time} seconds')
|
147 |
# Calculate and display the time per iteration in milliseconds
|
148 |
time_per_iteration_ms = (total_time / num_inference_steps_one)
|
149 |
+
cont.text(f'Time per iteration: {time_per_iteration_ms:.2f} seconds')
|
150 |
else:
|
151 |
st.error(f"Error: {response_one.text}")
|
152 |
|
153 |
+
# with tab2:
|
154 |
+
|
155 |
+
# st.title('Llama 2 7B Text Generation with AWS Inferentia 2')
|
156 |
+
|
157 |
+
# params = {
|
158 |
+
# "do_sample" : True,
|
159 |
+
# "top_p": 0.6,
|
160 |
+
# "temperature": 0.9,
|
161 |
+
# "top_k": 50,
|
162 |
+
# "max_new_tokens": 512,
|
163 |
+
# "repetition_penalty": 1.03,
|
164 |
+
# }
|
165 |
+
|
166 |
+
# if "messages" not in st.session_state:
|
167 |
+
# st.session_state.messages = [
|
168 |
+
# {"role": "system", "content": "You are a helpful Travel Planning Assistant. You respond with only 1-2 sentences."},
|
169 |
+
# {'role': 'user', 'content': 'Where can I travel in the fall for cloudy, rainy, and beautiful views?'},
|
170 |
+
# ]
|
171 |
+
|
172 |
+
# for message in st.session_state.messages:
|
173 |
+
# with st.chat_message(message["role"]):
|
174 |
+
# st.markdown(message["content"])
|
175 |
+
|
176 |
+
# with st.chat_message("assistant"):
|
177 |
+
# message_placeholder = st.empty()
|
178 |
+
# full_response = ""
|
179 |
+
# prompt_input_one = {
|
180 |
+
# "prompt": st.session_state.messages,
|
181 |
+
# "parameters": params,
|
182 |
+
# "endpoint": "huggingface-pytorch-inference-neuronx-2023-11-28-16-09-51-708"
|
183 |
+
# }
|
184 |
+
|
185 |
+
# response_one = requests.post(api_url, json=prompt_input_one)
|
186 |
+
|
187 |
+
# if response_one.status_code == 200:
|
188 |
+
# result_one = response_one.json()
|
189 |
+
# # st.success(f"Prediction result: {result}")
|
190 |
+
# full_response += result_one["generation"]
|
191 |
+
# else:
|
192 |
+
# st.error(f"Error: {response_one.text}")
|
193 |
+
|
194 |
+
# message_placeholder.markdown(full_response)
|
195 |
+
# st.session_state.messages.append({"role": "assistant", "content": full_response})
|
196 |
+
|
197 |
+
# if prompt := st.chat_input("What is up?"):
|
198 |
+
# st.session_state.messages.append({"role": "user", "content": prompt})
|
199 |
+
# print(st.session_state.messages)
|
200 |
+
# with st.chat_message("user"):
|
201 |
+
# st.markdown(prompt)
|
202 |
+
|
203 |
+
# with st.chat_message("assistant"):
|
204 |
+
# message_placeholder = st.empty()
|
205 |
+
# new_response = ""
|
206 |
+
# prompt_input_one = {
|
207 |
+
# "prompt": st.session_state.messages,
|
208 |
+
# "parameters": params,
|
209 |
+
# "endpoint": "huggingface-pytorch-inference-neuronx-2023-11-28-16-09-51-708"
|
210 |
+
# }
|
211 |
+
|
212 |
+
# response_one = requests.post(api_url, json=prompt_input_one)
|
213 |
+
|
214 |
+
# if response_one.status_code == 200:
|
215 |
+
# result_one = response_one.json()
|
216 |
+
# # st.success(f"Prediction result: {result}")
|
217 |
+
# new_response += result_one["generation"]
|
218 |
+
# else:
|
219 |
+
# st.error(f"Error: {response_one.text}")
|
220 |
+
|
221 |
+
# message_placeholder.markdown(new_response)
|
222 |
+
# st.session_state.messages.append({"role": "assistant", "content": new_response})
|
223 |
+
|
224 |
+
|
225 |
+
|
226 |
|
227 |
with tab2:
|
228 |
# ===========
|