Spaces:
Sleeping
Sleeping
Create app2.py
Browse files
app2.py
ADDED
@@ -0,0 +1,61 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
from ctransformers import AutoModelForCausalLM
|
3 |
+
from transformers import AutoTokenizer, pipeline
|
4 |
+
import torch
|
5 |
+
import re
|
6 |
+
|
7 |
+
# Initialize the model
|
8 |
+
model = AutoModelForCausalLM.from_pretrained("Detsutut/Igea-1B-v0.0.1-Q4_K_M-GGUF")
|
9 |
+
tokenizer = AutoTokenizer.from_pretrained( "Detsutut/Igea-350M-v0.0.1")
|
10 |
+
|
11 |
+
|
12 |
+
gen_pipeline = pipeline(
|
13 |
+
"text-generation",
|
14 |
+
model=model,
|
15 |
+
tokenizer=tokenizer
|
16 |
+
)
|
17 |
+
|
18 |
+
# Define the function to generate text
|
19 |
+
def generate_text(input_text, max_new_tokens, temperature, top_p, split_output):
|
20 |
+
if split_output:
|
21 |
+
max_new_tokens=30
|
22 |
+
top_p=0.95
|
23 |
+
output = gen_pipeline(
|
24 |
+
input_text,
|
25 |
+
max_new_tokens=max_new_tokens,
|
26 |
+
temperature=temperature,
|
27 |
+
top_p=top_p,
|
28 |
+
return_full_text = False
|
29 |
+
)
|
30 |
+
generated_text = output[0]['generated_text']
|
31 |
+
if split_output:
|
32 |
+
sentences = re.split('(?<!\w\.\w.)(?<![A-Z][a-z]\.)(?<=\.|\?)\s', generated_text)
|
33 |
+
if sentences:
|
34 |
+
generated_text = sentences[0]
|
35 |
+
return f"<span>{input_text}</span><b style='color: blue;'>{generated_text}</b>"
|
36 |
+
|
37 |
+
# Create the Gradio interface
|
38 |
+
input_text = gr.Textbox(lines=2, placeholder="Enter your text here...", label="Input Text")
|
39 |
+
|
40 |
+
max_new_tokens = gr.Slider(minimum=1, maximum=200, value=30, step=1, label="Max New Tokens")
|
41 |
+
temperature = gr.Slider(minimum=0.1, maximum=2.0, value=1.0, step=0.1, label="Temperature")
|
42 |
+
top_p = gr.Slider(minimum=0.0, maximum=1.0, value=0.95, step=0.01, label="Top-p")
|
43 |
+
split_output = gr.Checkbox(label="Quick single-sentence output", value=True)
|
44 |
+
|
45 |
+
with gr.Blocks(css="#outbox { border-radius: 8px !important; border: 1px solid #e5e7eb !important; padding: 8px !important; text-align:center !important;}") as iface:
|
46 |
+
gr.Markdown("# Igea Text Generation Interface ⚕️🩺")
|
47 |
+
gr.Markdown("⚠️ 🐢💬 This model runs on a **hardware-limited**, free-tier HuggingFace space, resulting in a **low output token throughput** (approx. 1 token/s)")
|
48 |
+
input_text.render()
|
49 |
+
with gr.Accordion("Advanced Options", open=False):
|
50 |
+
max_new_tokens.render()
|
51 |
+
temperature.render()
|
52 |
+
top_p.render()
|
53 |
+
split_output.render()
|
54 |
+
output = gr.HTML(label="Generated Text",elem_id="outbox")
|
55 |
+
|
56 |
+
btn = gr.Button("Generate")
|
57 |
+
btn.click(generate_text, [input_text, max_new_tokens, temperature, top_p, split_output], output)
|
58 |
+
|
59 |
+
# Launch the interface
|
60 |
+
if __name__ == "__main__":
|
61 |
+
iface.launch(inline=True)
|