Spaces:
Running
on
Zero
Running
on
Zero
Update app.py
Browse files
app.py
CHANGED
@@ -1,26 +1,18 @@
|
|
1 |
import gradio as gr
|
|
|
2 |
import torch
|
3 |
-
from transformers import AutoTokenizer,
|
4 |
|
5 |
# Load the model and tokenizer
|
6 |
model_name = "akjindal53244/Llama-3.1-Storm-8B"
|
7 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
8 |
-
|
9 |
-
|
10 |
-
model=model_name,
|
11 |
torch_dtype=torch.bfloat16,
|
12 |
device_map="auto"
|
13 |
)
|
14 |
|
15 |
-
|
16 |
-
HTML_CONTENT = """
|
17 |
-
<h1>Llama-3.1-Storm-8B Text Generation</h1>
|
18 |
-
<p>Generate text using the powerful Llama-3.1-Storm-8B model. Enter a prompt and let the AI create!</p>
|
19 |
-
<div class="llama-image">
|
20 |
-
<img src="https://cdn-uploads.huggingface.co/production/uploads/64c75c1237333ccfef30a602/tmOlbERGKP7JSODa6T06J.jpeg" alt="Llama" style="width:200px; border-radius:10px;">
|
21 |
-
</div>
|
22 |
-
"""
|
23 |
-
|
24 |
def generate_text(prompt, max_length, temperature):
|
25 |
messages = [
|
26 |
{"role": "system", "content": "You are a helpful assistant."},
|
@@ -28,8 +20,10 @@ def generate_text(prompt, max_length, temperature):
|
|
28 |
]
|
29 |
formatted_prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
|
30 |
|
31 |
-
|
32 |
-
|
|
|
|
|
33 |
max_new_tokens=max_length,
|
34 |
do_sample=True,
|
35 |
temperature=temperature,
|
@@ -37,20 +31,116 @@ def generate_text(prompt, max_length, temperature):
|
|
37 |
top_p=0.95,
|
38 |
)
|
39 |
|
40 |
-
return outputs[0]['
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
41 |
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
temperature = gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature")
|
49 |
-
submit_button = gr.Button("Generate")
|
50 |
-
with gr.Column(scale=2):
|
51 |
-
output = gr.Textbox(label="Generated Text", lines=10)
|
52 |
|
53 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
54 |
|
55 |
-
|
56 |
-
|
|
|
1 |
import gradio as gr
|
2 |
+
import spaces
|
3 |
import torch
|
4 |
+
from transformers import AutoTokenizer, AutoModelForCausalLM
|
5 |
|
6 |
# Load the model and tokenizer
|
7 |
model_name = "akjindal53244/Llama-3.1-Storm-8B"
|
8 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
9 |
+
model = AutoModelForCausalLM.from_pretrained(
|
10 |
+
model_name,
|
|
|
11 |
torch_dtype=torch.bfloat16,
|
12 |
device_map="auto"
|
13 |
)
|
14 |
|
15 |
+
@spaces.GPU(duration=120)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
16 |
def generate_text(prompt, max_length, temperature):
|
17 |
messages = [
|
18 |
{"role": "system", "content": "You are a helpful assistant."},
|
|
|
20 |
]
|
21 |
formatted_prompt = tokenizer.apply_chat_template(messages, add_generation_prompt=True, tokenize=False)
|
22 |
|
23 |
+
inputs = tokenizer(formatted_prompt, return_tensors="pt").to(model.device)
|
24 |
+
|
25 |
+
outputs = model.generate(
|
26 |
+
**inputs,
|
27 |
max_new_tokens=max_length,
|
28 |
do_sample=True,
|
29 |
temperature=temperature,
|
|
|
31 |
top_p=0.95,
|
32 |
)
|
33 |
|
34 |
+
return tokenizer.decode(outputs[0][inputs['input_ids'].shape[1]:], skip_special_tokens=True)
|
35 |
+
|
36 |
+
# Custom CSS
|
37 |
+
css = """
|
38 |
+
body {
|
39 |
+
background: linear-gradient(135deg, #f5f7fa, #c3cfe2);
|
40 |
+
font-family: Arial, sans-serif;
|
41 |
+
}
|
42 |
+
#custom-header {
|
43 |
+
text-align: center;
|
44 |
+
background: rgba(255, 255, 255, 0.8);
|
45 |
+
padding: 20px;
|
46 |
+
border-radius: 10px;
|
47 |
+
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
|
48 |
+
position: relative;
|
49 |
+
max-width: 800px;
|
50 |
+
margin: 20px auto;
|
51 |
+
}
|
52 |
+
#custom-header h1 {
|
53 |
+
color: #4A90E2;
|
54 |
+
font-size: 2em;
|
55 |
+
margin-bottom: 10px;
|
56 |
+
}
|
57 |
+
.llama-image {
|
58 |
+
position: relative;
|
59 |
+
transition: transform 0.3s;
|
60 |
+
display: inline-block;
|
61 |
+
margin-top: 20px;
|
62 |
+
}
|
63 |
+
.llama-image:hover {
|
64 |
+
transform: scale(1.05);
|
65 |
+
}
|
66 |
+
.llama-image img {
|
67 |
+
width: 200px;
|
68 |
+
border-radius: 10px;
|
69 |
+
box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
|
70 |
+
}
|
71 |
+
.llama-description {
|
72 |
+
position: absolute;
|
73 |
+
bottom: -30px;
|
74 |
+
left: 50%;
|
75 |
+
transform: translateX(-50%);
|
76 |
+
background-color: #4A90E2;
|
77 |
+
color: white;
|
78 |
+
padding: 5px 10px;
|
79 |
+
border-radius: 5px;
|
80 |
+
opacity: 0;
|
81 |
+
transition: opacity 0.3s;
|
82 |
+
white-space: nowrap;
|
83 |
+
}
|
84 |
+
.llama-image:hover .llama-description {
|
85 |
+
opacity: 1;
|
86 |
+
}
|
87 |
+
.artifact {
|
88 |
+
position: absolute;
|
89 |
+
background: rgba(74, 144, 226, 0.1);
|
90 |
+
border-radius: 50%;
|
91 |
+
}
|
92 |
+
.artifact.large {
|
93 |
+
width: 300px;
|
94 |
+
height: 300px;
|
95 |
+
top: -50px;
|
96 |
+
left: -150px;
|
97 |
+
}
|
98 |
+
.artifact.medium {
|
99 |
+
width: 200px;
|
100 |
+
height: 200px;
|
101 |
+
bottom: -50px;
|
102 |
+
right: -100px;
|
103 |
+
}
|
104 |
+
.artifact.small {
|
105 |
+
width: 100px;
|
106 |
+
height: 100px;
|
107 |
+
top: 50%;
|
108 |
+
left: 50%;
|
109 |
+
transform: translate(-50%, -50%);
|
110 |
+
}
|
111 |
+
"""
|
112 |
|
113 |
+
# Custom HTML
|
114 |
+
custom_html = """
|
115 |
+
<div id="custom-header">
|
116 |
+
<div class="artifact large"></div>
|
117 |
+
<div class="artifact medium"></div>
|
118 |
+
<div class="artifact small"></div>
|
|
|
|
|
|
|
|
|
119 |
|
120 |
+
<h1>Llama-3.1-Storm-8B Text Generation</h1>
|
121 |
+
<p>Generate text using the powerful Llama-3.1-Storm-8B model. Enter a prompt and let the AI create!</p>
|
122 |
+
|
123 |
+
<div class="llama-image">
|
124 |
+
<img src="https://cdn-uploads.huggingface.co/production/uploads/64c75c1237333ccfef30a602/tmOlbERGKP7JSODa6T06J.jpeg" alt="Llama">
|
125 |
+
<div class="llama-description">Llama-3.1-Storm-8B Model</div>
|
126 |
+
</div>
|
127 |
+
</div>
|
128 |
+
"""
|
129 |
+
|
130 |
+
# Create Gradio interface
|
131 |
+
iface = gr.Interface(
|
132 |
+
fn=generate_text,
|
133 |
+
inputs=[
|
134 |
+
gr.Textbox(lines=5, label="Prompt"),
|
135 |
+
gr.Slider(minimum=1, maximum=500, value=128, step=1, label="Max Length"),
|
136 |
+
gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.1, label="Temperature"),
|
137 |
+
],
|
138 |
+
outputs=gr.Textbox(lines=10, label="Generated Text"),
|
139 |
+
title="Llama-3.1-Storm-8B Text Generation",
|
140 |
+
description="Enter a prompt to generate text using the Llama-3.1-Storm-8B model.",
|
141 |
+
css=css,
|
142 |
+
article=custom_html
|
143 |
+
)
|
144 |
|
145 |
+
# Launch the app
|
146 |
+
iface.launch()
|