Spaces:
Running
on
T4
Running
on
T4
SandLogicTechnologies
commited on
Commit
•
d816a8a
1
Parent(s):
3080342
Update app.py
Browse files
app.py
CHANGED
@@ -8,9 +8,8 @@ import torch
|
|
8 |
import json
|
9 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
10 |
|
11 |
-
|
12 |
DESCRIPTION = """\
|
13 |
-
Shakti is a 2.5 billion parameter 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
|
14 |
For more details, please check [here](https://arxiv.org/pdf/2410.11331v1).
|
15 |
"""
|
16 |
|
@@ -20,17 +19,31 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "2048"))
|
|
20 |
|
21 |
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
22 |
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
model
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
|
|
|
|
34 |
|
35 |
@spaces.GPU(duration=90)
|
36 |
def generate(
|
@@ -79,6 +92,28 @@ def generate(
|
|
79 |
outputs.append(text)
|
80 |
yield "".join(outputs)
|
81 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
82 |
|
83 |
chat_interface = gr.ChatInterface(
|
84 |
fn=generate,
|
@@ -97,39 +132,28 @@ chat_interface = gr.ChatInterface(
|
|
97 |
step=0.1,
|
98 |
value=0.6,
|
99 |
),
|
100 |
-
# gr.Slider(
|
101 |
-
# label="Top-p (nucleus sampling)",
|
102 |
-
# minimum=0.05,
|
103 |
-
# maximum=1.0,
|
104 |
-
# step=0.05,
|
105 |
-
# value=0.9,
|
106 |
-
# ),
|
107 |
-
# gr.Slider(
|
108 |
-
# label="Top-k",
|
109 |
-
# minimum=1,
|
110 |
-
# maximum=1000,
|
111 |
-
# step=1,
|
112 |
-
# value=50,
|
113 |
-
# ),
|
114 |
-
# gr.Slider(
|
115 |
-
# label="Repetition penalty",
|
116 |
-
# minimum=1.0,
|
117 |
-
# maximum=2.0,
|
118 |
-
# step=0.05,
|
119 |
-
# value=1.2,
|
120 |
-
# ),
|
121 |
],
|
122 |
stop_btn=None,
|
123 |
-
examples=
|
124 |
-
["Tell me a story"], ["write a short poem which is hard to sing"], ['मुझे भारतीय इतिहास के बारे में बताएं']
|
125 |
-
],
|
126 |
cache_examples=False,
|
127 |
)
|
128 |
|
129 |
with gr.Blocks(css="style.css", fill_height=True) as demo:
|
130 |
gr.Markdown(DESCRIPTION)
|
131 |
gr.DuplicateButton(value="Duplicate Space for private use", elem_id="duplicate-button")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
132 |
chat_interface.render()
|
133 |
|
134 |
if __name__ == "__main__":
|
135 |
-
demo.queue(max_size=20).launch()
|
|
|
8 |
import json
|
9 |
from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
|
10 |
|
|
|
11 |
DESCRIPTION = """\
|
12 |
+
Shakti is a 2.5 billion parameter 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.
|
13 |
For more details, please check [here](https://arxiv.org/pdf/2410.11331v1).
|
14 |
"""
|
15 |
|
|
|
19 |
|
20 |
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
|
21 |
|
22 |
+
# Model configurations
|
23 |
+
model_options = {
|
24 |
+
"Shakti-100M": "SandLogicTechnologies/Shakti-100M",
|
25 |
+
"Shakti-250M": "SandLogicTechnologies/Shakti-250M",
|
26 |
+
"Shakti-2.5B": "SandLogicTechnologies/Shakti-2.5B"
|
27 |
+
}
|
28 |
+
|
29 |
+
# Initialize tokenizer and model variables
|
30 |
+
tokenizer = None
|
31 |
+
model = None
|
32 |
+
|
33 |
+
def load_model(selected_model: str):
|
34 |
+
global tokenizer, model
|
35 |
+
model_id = model_options[selected_model]
|
36 |
+
tokenizer = AutoTokenizer.from_pretrained(model_id, token=os.getenv("SHAKTI"))
|
37 |
+
model = AutoModelForCausalLM.from_pretrained(
|
38 |
+
model_id,
|
39 |
+
device_map="auto",
|
40 |
+
torch_dtype=torch.bfloat16,
|
41 |
+
token=os.getenv("SHAKTI")
|
42 |
+
)
|
43 |
+
model.eval()
|
44 |
|
45 |
+
# Initial model load (default to 2.5B)
|
46 |
+
load_model("Shakti-2.5B")
|
47 |
|
48 |
@spaces.GPU(duration=90)
|
49 |
def generate(
|
|
|
92 |
outputs.append(text)
|
93 |
yield "".join(outputs)
|
94 |
|
95 |
+
def update_examples(selected_model):
|
96 |
+
if selected_model == "Shakti-100M":
|
97 |
+
return [["Tell me a story"],
|
98 |
+
["Write a short poem on Rose"],
|
99 |
+
["What are computers"]]
|
100 |
+
elif selected_model == "Shakti-250M":
|
101 |
+
return [["Can you explain the pathophysiology of hypertension and its impact on the cardiovascular system?"],
|
102 |
+
["What are the potential side effects of beta-blockers in the treatment of arrhythmias?"],
|
103 |
+
["What foods are good for boosting the immune system?"],
|
104 |
+
["What is the difference between a stock and a bond?"],
|
105 |
+
["How can I start saving for retirement?"],
|
106 |
+
["What are some low-risk investment options?"],
|
107 |
+
["What is a power of attorney and when is it used?"],
|
108 |
+
["What are the key differences between a will and a trust?"],
|
109 |
+
["How do I legally protect my business name?"]]
|
110 |
+
else:
|
111 |
+
return [["Tell me a story"], ["write a short poem which is hard to sing"], ['मुझे भारतीय इतिहास के बारे में बताएं']]
|
112 |
+
|
113 |
+
def on_model_select(selected_model):
|
114 |
+
load_model(selected_model) # Load the selected model
|
115 |
+
return update_examples(selected_model) # Return new examples based on the selected model
|
116 |
+
|
117 |
|
118 |
chat_interface = gr.ChatInterface(
|
119 |
fn=generate,
|
|
|
132 |
step=0.1,
|
133 |
value=0.6,
|
134 |
),
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
135 |
],
|
136 |
stop_btn=None,
|
137 |
+
examples=update_examples("Shakti-2.5B"), # Set initial examples for 2.5B model
|
|
|
|
|
138 |
cache_examples=False,
|
139 |
)
|
140 |
|
141 |
with gr.Blocks(css="style.css", fill_height=True) as demo:
|
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 |
+
# Function to handle model change and update examples dynamically
|
154 |
+
model_dropdown.change(on_model_select, inputs=model_dropdown, outputs=[chat_interface])
|
155 |
+
|
156 |
chat_interface.render()
|
157 |
|
158 |
if __name__ == "__main__":
|
159 |
+
demo.queue(max_size=20).launch()
|