gpu util
Browse files
app.py
CHANGED
@@ -1,17 +1,15 @@
|
|
1 |
import gradio as gr
|
2 |
from transformers import pipeline
|
3 |
-
import
|
4 |
-
|
5 |
-
# Check if a GPU is available
|
6 |
-
device = 0 if torch.cuda.is_available() else -1
|
7 |
-
print("Using GPU" if device == 0 else "Using CPU")
|
8 |
-
|
9 |
-
# Load the model on the GPU if available
|
10 |
-
model = pipeline("text-generation", model="gpt2", device=device)
|
11 |
|
|
|
|
|
12 |
def generate_text(prompt):
|
|
|
|
|
13 |
return model(prompt, max_length=50)[0]["generated_text"]
|
14 |
|
|
|
15 |
interface = gr.Interface(
|
16 |
fn=generate_text,
|
17 |
inputs=gr.Textbox(label="Enter your prompt here"),
|
|
|
1 |
import gradio as gr
|
2 |
from transformers import pipeline
|
3 |
+
from spaces import GPU # Import the GPU decorator for ZeroGPU
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
+
# Decorate the function to indicate it needs GPU resources
|
6 |
+
@GPU
|
7 |
def generate_text(prompt):
|
8 |
+
# Load the model within the function so that it only runs on GPU when the function is called
|
9 |
+
model = pipeline("text-generation", model="gpt2", device=0)
|
10 |
return model(prompt, max_length=50)[0]["generated_text"]
|
11 |
|
12 |
+
# Create the Gradio interface
|
13 |
interface = gr.Interface(
|
14 |
fn=generate_text,
|
15 |
inputs=gr.Textbox(label="Enter your prompt here"),
|