Spaces:
Running
Running
Commit
·
4667b7d
1
Parent(s):
b6ed54a
feat: Implement async model responses with real-time progress
Browse files- Make model calls run concurrently instead of sequentially
- Add real-time progress indicator with 0.1s precision timer
- Display responses immediately as they arrive
- Improve error handling and loading states
app.py
CHANGED
@@ -1,7 +1,9 @@
|
|
1 |
import os
|
2 |
import gradio as gr
|
|
|
3 |
from dotenv import load_dotenv
|
4 |
from huggingface_hub import InferenceClient
|
|
|
5 |
|
6 |
# Load environment variables
|
7 |
load_dotenv()
|
@@ -22,7 +24,7 @@ AVAILABLE_MODELS = [
|
|
22 |
# Initialize inference client
|
23 |
inference_client = InferenceClient(token=HF_TOKEN)
|
24 |
|
25 |
-
def get_model_response(prompt, model_name, temperature_value, do_sample):
|
26 |
"""Get response from a Hugging Face model."""
|
27 |
try:
|
28 |
# Build kwargs dynamically
|
@@ -38,31 +40,80 @@ def get_model_response(prompt, model_name, temperature_value, do_sample):
|
|
38 |
if do_sample and temperature_value > 0:
|
39 |
generation_args["temperature"] = temperature_value
|
40 |
|
41 |
-
|
|
|
|
|
|
|
|
|
|
|
42 |
return response
|
43 |
|
44 |
except Exception as e:
|
45 |
return f"Error: {str(e)}"
|
46 |
|
47 |
-
def
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
"""Compare outputs from two selected models."""
|
49 |
if not prompt.strip():
|
50 |
-
|
51 |
-
|
52 |
-
|
53 |
-
gr.update(interactive=True)
|
54 |
-
)
|
55 |
|
56 |
-
|
57 |
-
|
|
|
58 |
|
59 |
-
#
|
60 |
-
|
61 |
-
|
62 |
-
|
63 |
|
64 |
-
|
65 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
66 |
|
67 |
# Update temperature slider interactivity based on sampling checkbox
|
68 |
def update_slider_state(enabled):
|
@@ -79,7 +130,7 @@ with gr.Blocks(css="""
|
|
79 |
.disabled-slider { opacity: 0.5; pointer-events: none; }
|
80 |
""") as demo:
|
81 |
gr.Markdown("# LLM Comparison Tool")
|
82 |
-
gr.Markdown("
|
83 |
|
84 |
with gr.Row():
|
85 |
prompt = gr.Textbox(
|
@@ -117,7 +168,6 @@ with gr.Blocks(css="""
|
|
117 |
height=300,
|
118 |
type="messages"
|
119 |
)
|
120 |
-
|
121 |
|
122 |
with gr.Column():
|
123 |
model2_dropdown = gr.Dropdown(
|
@@ -157,11 +207,10 @@ with gr.Blocks(css="""
|
|
157 |
).then(
|
158 |
fn=compare_models,
|
159 |
inputs=[prompt, model1_dropdown, model2_dropdown, temp1, temp2, do_sample1, do_sample2],
|
160 |
-
outputs=[chatbot1, chatbot2, submit_btn]
|
|
|
161 |
)
|
162 |
|
163 |
-
|
164 |
-
|
165 |
do_sample1.change(
|
166 |
fn=update_slider_state,
|
167 |
inputs=[do_sample1],
|
@@ -175,5 +224,4 @@ with gr.Blocks(css="""
|
|
175 |
)
|
176 |
|
177 |
if __name__ == "__main__":
|
178 |
-
demo.launch()
|
179 |
-
# demo.launch(share=True)
|
|
|
1 |
import os
|
2 |
import gradio as gr
|
3 |
+
import asyncio
|
4 |
from dotenv import load_dotenv
|
5 |
from huggingface_hub import InferenceClient
|
6 |
+
from functools import partial
|
7 |
|
8 |
# Load environment variables
|
9 |
load_dotenv()
|
|
|
24 |
# Initialize inference client
|
25 |
inference_client = InferenceClient(token=HF_TOKEN)
|
26 |
|
27 |
+
async def get_model_response(prompt, model_name, temperature_value, do_sample):
|
28 |
"""Get response from a Hugging Face model."""
|
29 |
try:
|
30 |
# Build kwargs dynamically
|
|
|
40 |
if do_sample and temperature_value > 0:
|
41 |
generation_args["temperature"] = temperature_value
|
42 |
|
43 |
+
# Run the inference in a thread pool to not block the event loop
|
44 |
+
loop = asyncio.get_event_loop()
|
45 |
+
response = await loop.run_in_executor(
|
46 |
+
None,
|
47 |
+
partial(inference_client.text_generation, **generation_args)
|
48 |
+
)
|
49 |
return response
|
50 |
|
51 |
except Exception as e:
|
52 |
return f"Error: {str(e)}"
|
53 |
|
54 |
+
async def process_single_response(prompt, model_name, temp, do_sample, chatbot):
|
55 |
+
"""Process a single model response and update its chatbot."""
|
56 |
+
response = await get_model_response(prompt, model_name, temp, do_sample)
|
57 |
+
chat_history = [{"role": "user", "content": prompt}, {"role": "assistant", "content": response}]
|
58 |
+
return chat_history
|
59 |
+
|
60 |
+
async def compare_models(prompt, model1, model2, temp1, temp2, do_sample1, do_sample2):
|
61 |
"""Compare outputs from two selected models."""
|
62 |
if not prompt.strip():
|
63 |
+
empty_response = [{"role": "user", "content": prompt}, {"role": "assistant", "content": "Please enter a prompt"}]
|
64 |
+
yield empty_response, empty_response, gr.update(interactive=True)
|
65 |
+
return # Exit the generator
|
|
|
|
|
66 |
|
67 |
+
# Initialize with "Generating..." messages
|
68 |
+
initial_message = [{"role": "user", "content": prompt}, {"role": "assistant", "content": "Generating..."}]
|
69 |
+
yield initial_message, initial_message, gr.update(interactive=False)
|
70 |
|
71 |
+
# Create tasks for both model responses
|
72 |
+
task1 = asyncio.create_task(process_single_response(prompt, model1, temp1, do_sample1, "chatbot1"))
|
73 |
+
task2 = asyncio.create_task(process_single_response(prompt, model2, temp2, do_sample2, "chatbot2"))
|
|
|
74 |
|
75 |
+
chat1 = chat2 = initial_message
|
76 |
+
start_time = asyncio.get_event_loop().time()
|
77 |
+
|
78 |
+
try:
|
79 |
+
while not (task1.done() and task2.done()):
|
80 |
+
# Update the messages with elapsed time
|
81 |
+
elapsed = round(asyncio.get_event_loop().time() - start_time, 1)
|
82 |
+
chat1_content = chat1[1]["content"]
|
83 |
+
chat2_content = chat2[1]["content"]
|
84 |
+
|
85 |
+
if not task1.done():
|
86 |
+
chat1 = [{"role": "user", "content": prompt},
|
87 |
+
{"role": "assistant", "content": f"Generating... ({elapsed:.1f}s)"}]
|
88 |
+
if not task2.done():
|
89 |
+
chat2 = [{"role": "user", "content": prompt},
|
90 |
+
{"role": "assistant", "content": f"Generating... ({elapsed:.1f}s)"}]
|
91 |
+
|
92 |
+
# Check if any task completed
|
93 |
+
done, pending = await asyncio.wait([t for t in [task1, task2] if not t.done()],
|
94 |
+
timeout=0.1,
|
95 |
+
return_when=asyncio.FIRST_COMPLETED)
|
96 |
+
|
97 |
+
for task in done:
|
98 |
+
if task == task1:
|
99 |
+
chat1 = await task1
|
100 |
+
else:
|
101 |
+
chat2 = await task2
|
102 |
+
|
103 |
+
yield chat1, chat2, gr.update(interactive=False)
|
104 |
+
|
105 |
+
# Ensure we have both final results
|
106 |
+
if not task1.done():
|
107 |
+
chat1 = await task1
|
108 |
+
if not task2.done():
|
109 |
+
chat2 = await task2
|
110 |
+
|
111 |
+
# Final yield with both results
|
112 |
+
yield chat1, chat2, gr.update(interactive=True)
|
113 |
+
|
114 |
+
except Exception as e:
|
115 |
+
error_message = [{"role": "user", "content": prompt}, {"role": "assistant", "content": f"Error: {str(e)}"}]
|
116 |
+
yield error_message, error_message, gr.update(interactive=True)
|
117 |
|
118 |
# Update temperature slider interactivity based on sampling checkbox
|
119 |
def update_slider_state(enabled):
|
|
|
130 |
.disabled-slider { opacity: 0.5; pointer-events: none; }
|
131 |
""") as demo:
|
132 |
gr.Markdown("# LLM Comparison Tool")
|
133 |
+
gr.Markdown("Using HuggingFace's Inference API, compare outputs from different `text-generation` models side by side.")
|
134 |
|
135 |
with gr.Row():
|
136 |
prompt = gr.Textbox(
|
|
|
168 |
height=300,
|
169 |
type="messages"
|
170 |
)
|
|
|
171 |
|
172 |
with gr.Column():
|
173 |
model2_dropdown = gr.Dropdown(
|
|
|
207 |
).then(
|
208 |
fn=compare_models,
|
209 |
inputs=[prompt, model1_dropdown, model2_dropdown, temp1, temp2, do_sample1, do_sample2],
|
210 |
+
outputs=[chatbot1, chatbot2, submit_btn],
|
211 |
+
queue=True # Enable queuing for streaming updates
|
212 |
)
|
213 |
|
|
|
|
|
214 |
do_sample1.change(
|
215 |
fn=update_slider_state,
|
216 |
inputs=[do_sample1],
|
|
|
224 |
)
|
225 |
|
226 |
if __name__ == "__main__":
|
227 |
+
demo.queue().launch()
|
|