Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,22 +1,34 @@
|
|
1 |
import gradio as gr
|
|
|
2 |
|
3 |
-
#
|
4 |
-
def
|
5 |
-
print(f"
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
13 |
-
|
14 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
|
16 |
-
# Load the models (
|
17 |
-
deepseek_r1_distill =
|
18 |
-
deepseek_r1 =
|
19 |
-
deepseek_r1_zero =
|
20 |
|
21 |
# --- Chatbot function ---
|
22 |
def chatbot(input_text, history, model_choice, system_message, max_new_tokens, temperature, top_p):
|
@@ -24,22 +36,29 @@ def chatbot(input_text, history, model_choice, system_message, max_new_tokens, t
|
|
24 |
print(f"Input: {input_text}, History: {history}, Model: {model_choice}")
|
25 |
|
26 |
# Choose the model based on user selection
|
27 |
-
if model_choice == "DeepSeek-R1-Distill-Qwen-32B":
|
28 |
-
|
29 |
-
elif model_choice == "DeepSeek-R1":
|
30 |
-
|
31 |
-
elif model_choice == "DeepSeek-R1-Zero":
|
32 |
-
|
33 |
else:
|
34 |
-
|
|
|
|
|
|
|
|
|
|
|
35 |
|
36 |
-
#
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
|
|
|
|
43 |
|
44 |
# --- Gradio Interface ---
|
45 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
@@ -63,7 +82,7 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
63 |
|
64 |
# Options moved below the chat interface
|
65 |
with gr.Row():
|
66 |
-
with gr.Accordion("Options", open=True):
|
67 |
model_choice = gr.Radio(
|
68 |
choices=["DeepSeek-R1-Distill-Qwen-32B", "DeepSeek-R1", "DeepSeek-R1-Zero"],
|
69 |
label="Choose a Model",
|
@@ -84,7 +103,6 @@ with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
84 |
top_p = gr.Slider(
|
85 |
minimum=0.10, maximum=1.00, value=0.90, label="Top-p (nucleus sampling)"
|
86 |
)
|
87 |
-
|
88 |
|
89 |
# Maintain chat history
|
90 |
chat_history = gr.State([])
|
|
|
1 |
import gradio as gr
|
2 |
+
import subprocess
|
3 |
|
4 |
+
# Function to load a model using Hugging Face Spaces and enable GPU
|
5 |
+
def load_model_with_gpu(model_name):
|
6 |
+
print(f"Attempting to load {model_name} with GPU enabled...")
|
7 |
+
try:
|
8 |
+
# Use subprocess to run hf.space_info and get GPU setting
|
9 |
+
result = subprocess.run(
|
10 |
+
["python", "-c", f"from huggingface_hub import space_info; print(space_info('{model_name}').hardware)"],
|
11 |
+
capture_output=True,
|
12 |
+
text=True,
|
13 |
+
check=True
|
14 |
+
)
|
15 |
+
hardware = result.stdout.strip()
|
16 |
+
print(f"Hardware for {model_name}: {hardware}")
|
17 |
+
|
18 |
+
demo = gr.load(name=model_name, src="spaces")
|
19 |
+
|
20 |
+
# Return the loaded model demo
|
21 |
+
print(f"Successfully loaded {model_name}")
|
22 |
+
return demo
|
23 |
+
|
24 |
+
except Exception as e:
|
25 |
+
print(f"Error loading model {model_name}: {e}")
|
26 |
+
return None
|
27 |
|
28 |
+
# Load the models with GPU enabled (if available)
|
29 |
+
deepseek_r1_distill = load_model_with_gpu("deepseek-ai/DeepSeek-R1-Distill-Qwen-32B")
|
30 |
+
deepseek_r1 = load_model_with_gpu("deepseek-ai/DeepSeek-R1")
|
31 |
+
deepseek_r1_zero = load_model_with_gpu("deepseek-ai/DeepSeek-R1-Zero")
|
32 |
|
33 |
# --- Chatbot function ---
|
34 |
def chatbot(input_text, history, model_choice, system_message, max_new_tokens, temperature, top_p):
|
|
|
36 |
print(f"Input: {input_text}, History: {history}, Model: {model_choice}")
|
37 |
|
38 |
# Choose the model based on user selection
|
39 |
+
if model_choice == "DeepSeek-R1-Distill-Qwen-32B" and deepseek_r1_distill:
|
40 |
+
model_demo = deepseek_r1_distill
|
41 |
+
elif model_choice == "DeepSeek-R1" and deepseek_r1:
|
42 |
+
model_demo = deepseek_r1
|
43 |
+
elif model_choice == "DeepSeek-R1-Zero" and deepseek_r1_zero:
|
44 |
+
model_demo = deepseek_r1_zero
|
45 |
else:
|
46 |
+
default_response = "Model not selected or could not be loaded."
|
47 |
+
history.append((input_text, default_response))
|
48 |
+
return history, history, "", model_choice, system_message, max_new_tokens, temperature, top_p
|
49 |
+
|
50 |
+
# Adjust the call to the model, remove default_value if not applicable
|
51 |
+
model_output = model_demo(input_text, history, max_new_tokens, temperature, top_p, system_message)
|
52 |
|
53 |
+
# Check if model_output is iterable and has expected number of elements
|
54 |
+
if not isinstance(model_output, (list, tuple)) or len(model_output) < 2:
|
55 |
+
error_message = "Model output does not have the expected format."
|
56 |
+
history.append((input_text, error_message))
|
57 |
+
return history, history, "", model_choice, system_message, max_new_tokens, temperature, top_p
|
58 |
+
|
59 |
+
response = model_output[-1][1] if model_output[-1][1] else "Model did not return a response."
|
60 |
+
history.append((input_text, response))
|
61 |
+
return history, history, "", model_choice, system_message, max_new_tokens, temperature, top_p
|
62 |
|
63 |
# --- Gradio Interface ---
|
64 |
with gr.Blocks(theme=gr.themes.Soft()) as demo:
|
|
|
82 |
|
83 |
# Options moved below the chat interface
|
84 |
with gr.Row():
|
85 |
+
with gr.Accordion("Options", open=True):
|
86 |
model_choice = gr.Radio(
|
87 |
choices=["DeepSeek-R1-Distill-Qwen-32B", "DeepSeek-R1", "DeepSeek-R1-Zero"],
|
88 |
label="Choose a Model",
|
|
|
103 |
top_p = gr.Slider(
|
104 |
minimum=0.10, maximum=1.00, value=0.90, label="Top-p (nucleus sampling)"
|
105 |
)
|
|
|
106 |
|
107 |
# Maintain chat history
|
108 |
chat_history = gr.State([])
|