Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,6 @@
|
|
1 |
import os
|
2 |
from threading import Thread
|
3 |
from typing import Iterator
|
4 |
-
|
5 |
import gradio as gr
|
6 |
import spaces
|
7 |
import torch
|
@@ -9,8 +8,36 @@ import json
|
|
9 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
10 |
|
11 |
DESCRIPTION = """\
|
12 |
-
Shakti
|
13 |
-
For more details, please check [here](https://arxiv.org/pdf/2410.11331v1)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
"""
|
15 |
|
16 |
MAX_MAX_NEW_TOKENS = 2048
|
@@ -43,14 +70,14 @@ def load_model(selected_model: str):
|
|
43 |
token=os.getenv("SHAKTI")
|
44 |
)
|
45 |
model.eval()
|
46 |
-
|
|
|
47 |
|
48 |
|
49 |
-
# Initial model load
|
50 |
load_model("Shakti-2.5B")
|
51 |
|
52 |
|
53 |
-
@spaces.GPU(duration=90)
|
54 |
def generate(
|
55 |
message: str,
|
56 |
chat_history: list[tuple[str, str]],
|
@@ -62,24 +89,19 @@ def generate(
|
|
62 |
) -> Iterator[str]:
|
63 |
conversation = []
|
64 |
|
65 |
-
# Conditional logic for adding prompt based on model
|
66 |
if current_model == "Shakti-2.5B":
|
67 |
for user, assistant in chat_history:
|
68 |
-
conversation.extend(
|
69 |
-
|
70 |
-
|
71 |
-
|
72 |
-
|
73 |
-
]
|
74 |
-
)
|
75 |
else:
|
76 |
for user, assistant in chat_history:
|
77 |
-
conversation.extend(
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
]
|
82 |
-
)
|
83 |
|
84 |
conversation.append({"role": "user", "content": message})
|
85 |
|
@@ -110,72 +132,167 @@ def generate(
|
|
110 |
yield "".join(outputs)
|
111 |
|
112 |
|
113 |
-
def
|
114 |
-
|
115 |
-
|
116 |
-
|
117 |
-
|
118 |
-
|
119 |
-
|
120 |
-
|
121 |
-
|
122 |
-
|
123 |
-
|
124 |
-
|
125 |
-
|
126 |
-
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
131 |
|
132 |
|
133 |
def on_model_select(selected_model):
|
134 |
load_model(selected_model) # Load the selected model
|
135 |
-
|
136 |
-
return gr.update(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
137 |
|
138 |
|
139 |
-
|
|
|
140 |
|
141 |
-
|
|
|
142 |
gr.Markdown(DESCRIPTION)
|
143 |
-
gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
|
144 |
-
|
145 |
-
# Dropdown for model selection
|
146 |
-
model_dropdown = gr.Dropdown(
|
147 |
-
label="Select Model",
|
148 |
-
choices=["Shakti-100M", "Shakti-250M", "Shakti-2.5B"],
|
149 |
-
value="Shakti-2.5B",
|
150 |
-
interactive=True,
|
151 |
-
)
|
152 |
|
153 |
-
|
154 |
-
|
155 |
-
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
-
)
|
161 |
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
-
|
167 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
168 |
)
|
169 |
|
170 |
-
|
171 |
-
|
172 |
-
inputs=[
|
173 |
-
|
174 |
-
outputs=chat_history,
|
175 |
-
live=True,
|
176 |
)
|
177 |
|
178 |
-
#
|
179 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
180 |
|
181 |
-
|
|
|
|
1 |
import os
|
2 |
from threading import Thread
|
3 |
from typing import Iterator
|
|
|
4 |
import gradio as gr
|
5 |
import spaces
|
6 |
import torch
|
|
|
8 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
9 |
|
10 |
DESCRIPTION = """\
|
11 |
+
Shakti LLMs (Large Language Models) are a group of compact language models specifically optimized for resource-constrained environments such as edge devices, including smartphones, wearables, and IoT (Internet of Things) systems. These models provide support for vernacular languages and domain-specific tasks, making them particularly suitable for industries such as healthcare, finance, and customer service.
|
12 |
+
For more details, please check [here](https://arxiv.org/pdf/2410.11331v1)
|
13 |
+
"""
|
14 |
+
|
15 |
+
|
16 |
+
# """\
|
17 |
+
# Shakti LLMs are a group of small language model specifically optimized for resource-constrained environments such as edge devices, including smartphones, wearables, and IoT systems. With support for vernacular languages and domain-specific tasks, Shakti excels in industries such as healthcare, finance, and customer service.
|
18 |
+
# For more details, please check [here](https://arxiv.org/pdf/2410.11331v1).
|
19 |
+
# """
|
20 |
+
|
21 |
+
|
22 |
+
# Custom CSS for the send button
|
23 |
+
CUSTOM_CSS = """
|
24 |
+
.send-btn {
|
25 |
+
padding: 0.5rem !important;
|
26 |
+
width: 55px !important;
|
27 |
+
height: 55px !important;
|
28 |
+
border-radius: 50% !important;
|
29 |
+
margin-top: 1rem;
|
30 |
+
cursor: pointer;
|
31 |
+
}
|
32 |
+
|
33 |
+
.send-btn svg {
|
34 |
+
width: 20px !important;
|
35 |
+
height: 20px !important;
|
36 |
+
position: absolute;
|
37 |
+
top: 50%;
|
38 |
+
left: 50%;
|
39 |
+
transform: translate(-50%, -50%);
|
40 |
+
}
|
41 |
"""
|
42 |
|
43 |
MAX_MAX_NEW_TOKENS = 2048
|
|
|
70 |
token=os.getenv("SHAKTI")
|
71 |
)
|
72 |
model.eval()
|
73 |
+
print("Selected Model: ", selected_model)
|
74 |
+
current_model = selected_model
|
75 |
|
76 |
|
77 |
+
# Initial model load
|
78 |
load_model("Shakti-2.5B")
|
79 |
|
80 |
|
|
|
81 |
def generate(
|
82 |
message: str,
|
83 |
chat_history: list[tuple[str, str]],
|
|
|
89 |
) -> Iterator[str]:
|
90 |
conversation = []
|
91 |
|
|
|
92 |
if current_model == "Shakti-2.5B":
|
93 |
for user, assistant in chat_history:
|
94 |
+
conversation.extend([
|
95 |
+
json.loads(os.getenv("PROMPT")),
|
96 |
+
{"role": "user", "content": user},
|
97 |
+
{"role": "assistant", "content": assistant},
|
98 |
+
])
|
|
|
|
|
99 |
else:
|
100 |
for user, assistant in chat_history:
|
101 |
+
conversation.extend([
|
102 |
+
{"role": "user", "content": user},
|
103 |
+
{"role": "assistant", "content": assistant},
|
104 |
+
])
|
|
|
|
|
105 |
|
106 |
conversation.append({"role": "user", "content": message})
|
107 |
|
|
|
132 |
yield "".join(outputs)
|
133 |
|
134 |
|
135 |
+
def respond(message, chat_history, max_new_tokens, temperature):
|
136 |
+
bot_message = ""
|
137 |
+
for chunk in generate(message, chat_history, max_new_tokens, temperature):
|
138 |
+
bot_message += chunk
|
139 |
+
chat_history.append((message, bot_message))
|
140 |
+
return "", chat_history
|
141 |
+
|
142 |
+
|
143 |
+
def get_examples(selected_model):
|
144 |
+
examples = {
|
145 |
+
"Shakti-100M": [
|
146 |
+
["Tell me a story"],
|
147 |
+
["Write a short poem on Rose"],
|
148 |
+
["What are computers"]
|
149 |
+
],
|
150 |
+
"Shakti-250M": [
|
151 |
+
["Can you explain the pathophysiology of hypertension and its impact on the cardiovascular system?"],
|
152 |
+
["What are the potential side effects of beta-blockers in the treatment of arrhythmias?"],
|
153 |
+
["What foods are good for boosting the immune system?"],
|
154 |
+
["What is the difference between a stock and a bond?"],
|
155 |
+
["How can I start saving for retirement?"],
|
156 |
+
["What are some low-risk investment options?"]
|
157 |
+
],
|
158 |
+
"Shakti-2.5B": [
|
159 |
+
["Tell me a story"],
|
160 |
+
["write a short poem which is hard to sing"],
|
161 |
+
['मुझे भारतीय इतिहास के बारे में बताएं']
|
162 |
+
]
|
163 |
+
}
|
164 |
+
return examples.get(selected_model, [])
|
165 |
|
166 |
|
167 |
def on_model_select(selected_model):
|
168 |
load_model(selected_model) # Load the selected model
|
169 |
+
# Return the message and chat history updates
|
170 |
+
return gr.update(value=""), gr.update(value=[]) # Clear message and chat history
|
171 |
+
|
172 |
+
|
173 |
+
def update_examples_visibility(selected_model):
|
174 |
+
# Return individual updates for each example section
|
175 |
+
return (
|
176 |
+
gr.update(visible=selected_model == "Shakti-100M"),
|
177 |
+
gr.update(visible=selected_model == "Shakti-250M"),
|
178 |
+
gr.update(visible=selected_model == "Shakti-2.5B")
|
179 |
+
)
|
180 |
|
181 |
|
182 |
+
def example_selector(example):
|
183 |
+
return example
|
184 |
|
185 |
+
|
186 |
+
with gr.Blocks(css=CUSTOM_CSS) as demo:
|
187 |
gr.Markdown(DESCRIPTION)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
188 |
|
189 |
+
with gr.Row():
|
190 |
+
model_dropdown = gr.Dropdown(
|
191 |
+
label="Select Model",
|
192 |
+
choices=list(model_options.keys()),
|
193 |
+
value="Shakti-2.5B",
|
194 |
+
interactive=True
|
195 |
+
)
|
|
|
196 |
|
197 |
+
chatbot = gr.Chatbot()
|
198 |
+
|
199 |
+
with gr.Row():
|
200 |
+
with gr.Column(scale=20):
|
201 |
+
msg = gr.Textbox(
|
202 |
+
label="Message",
|
203 |
+
placeholder="Enter your message here",
|
204 |
+
lines=2,
|
205 |
+
show_label=False
|
206 |
+
)
|
207 |
+
with gr.Column(scale=1, min_width=50):
|
208 |
+
send_btn = gr.Button(
|
209 |
+
value="➤",
|
210 |
+
variant="primary",
|
211 |
+
elem_classes=["send-btn"]
|
212 |
+
)
|
213 |
+
|
214 |
+
with gr.Accordion("Parameters", open=False):
|
215 |
+
max_tokens_slider = gr.Slider(
|
216 |
+
label="Max new tokens",
|
217 |
+
minimum=1,
|
218 |
+
maximum=MAX_MAX_NEW_TOKENS,
|
219 |
+
step=1,
|
220 |
+
value=DEFAULT_MAX_NEW_TOKENS,
|
221 |
+
)
|
222 |
+
temperature_slider = gr.Slider(
|
223 |
+
label="Temperature",
|
224 |
+
minimum=0.1,
|
225 |
+
maximum=4.0,
|
226 |
+
step=0.1,
|
227 |
+
value=0.6,
|
228 |
+
)
|
229 |
+
|
230 |
+
# Add submit action handlers
|
231 |
+
submit_click = send_btn.click(
|
232 |
+
respond,
|
233 |
+
inputs=[msg, chatbot, max_tokens_slider, temperature_slider],
|
234 |
+
outputs=[msg, chatbot]
|
235 |
)
|
236 |
|
237 |
+
submit_enter = msg.submit(
|
238 |
+
respond,
|
239 |
+
inputs=[msg, chatbot, max_tokens_slider, temperature_slider],
|
240 |
+
outputs=[msg, chatbot]
|
|
|
|
|
241 |
)
|
242 |
|
243 |
+
# Create separate example sections for each model
|
244 |
+
with gr.Row():
|
245 |
+
with gr.Column(visible=False) as examples_100m:
|
246 |
+
gr.Examples(
|
247 |
+
examples=get_examples("Shakti-100M"),
|
248 |
+
inputs=msg,
|
249 |
+
label="Example prompts for Shakti-100M",
|
250 |
+
fn=example_selector
|
251 |
+
)
|
252 |
+
|
253 |
+
with gr.Column(visible=False) as examples_250m:
|
254 |
+
gr.Examples(
|
255 |
+
examples=get_examples("Shakti-250M"),
|
256 |
+
inputs=msg,
|
257 |
+
label="Example prompts for Shakti-250M",
|
258 |
+
fn=example_selector
|
259 |
+
)
|
260 |
+
|
261 |
+
with gr.Column(visible=True) as examples_2_5b:
|
262 |
+
gr.Examples(
|
263 |
+
examples=get_examples("Shakti-2.5B"),
|
264 |
+
inputs=msg,
|
265 |
+
label="Example prompts for Shakti-2.5B",
|
266 |
+
fn=example_selector
|
267 |
+
)
|
268 |
+
|
269 |
+
|
270 |
+
# Update model selection and examples visibility
|
271 |
+
def combined_update(selected_model):
|
272 |
+
msg_update, chat_update = on_model_select(selected_model)
|
273 |
+
examples_100m_update, examples_250m_update, examples_2_5b_update = update_examples_visibility(
|
274 |
+
selected_model)
|
275 |
+
return [
|
276 |
+
msg_update,
|
277 |
+
chat_update,
|
278 |
+
examples_100m_update,
|
279 |
+
examples_250m_update,
|
280 |
+
examples_2_5b_update
|
281 |
+
]
|
282 |
+
|
283 |
+
|
284 |
+
# Updated change event handler
|
285 |
+
model_dropdown.change(
|
286 |
+
combined_update,
|
287 |
+
inputs=[model_dropdown],
|
288 |
+
outputs=[
|
289 |
+
msg,
|
290 |
+
chatbot,
|
291 |
+
examples_100m,
|
292 |
+
examples_250m,
|
293 |
+
examples_2_5b
|
294 |
+
]
|
295 |
+
)
|
296 |
|
297 |
+
if __name__ == "__main__":
|
298 |
+
demo.queue(max_size=20).launch()
|