Spaces:
Sleeping
Sleeping
Update app.py (#1)
Browse files- Update app.py (52d69de8d339328ef19de7ed98bf3787ed107748)
app.py
CHANGED
@@ -1,26 +1,21 @@
|
|
1 |
import json
|
2 |
-
from huggingface_hub import InferenceClient
|
3 |
import gradio as gr
|
4 |
import random
|
|
|
5 |
|
6 |
API_URL = "https://api-inference.huggingface.co/models/"
|
7 |
|
8 |
-
client = InferenceClient(
|
9 |
-
"mistralai/Mistral-7B-Instruct-v0.1"
|
10 |
-
)
|
11 |
|
12 |
def format_prompt(message, history):
|
13 |
prompt = "You're a helpful assistant."
|
14 |
for user_prompt, bot_response in history:
|
15 |
-
prompt += f"[INST] {user_prompt} [/INST]"
|
16 |
-
|
17 |
-
prompt += f"[INST] {message} [/INST]"
|
18 |
return prompt
|
19 |
|
20 |
def generate(prompt, history, temperature=0.9, max_new_tokens=2048, top_p=0.95, repetition_penalty=1.0):
|
21 |
-
temperature = float(temperature)
|
22 |
-
if temperature < 1e-2:
|
23 |
-
temperature = 1e-2
|
24 |
top_p = float(top_p)
|
25 |
|
26 |
generate_kwargs = dict(
|
@@ -40,47 +35,35 @@ def generate(prompt, history, temperature=0.9, max_new_tokens=2048, top_p=0.95,
|
|
40 |
for response in stream:
|
41 |
output += response.token.text
|
42 |
yield output
|
43 |
-
return output
|
44 |
|
45 |
def load_database():
|
46 |
try:
|
47 |
-
# Attempt to load the database from JSON
|
48 |
with open("database.json", "r", encoding="utf-8") as f:
|
49 |
return json.load(f)
|
50 |
except (FileNotFoundError, json.JSONDecodeError):
|
51 |
-
# Handle potential errors gracefully
|
52 |
print("Error loading database: File not found or invalid format. Creating an empty database.")
|
53 |
-
return []
|
54 |
|
55 |
def save_database(data):
|
56 |
try:
|
57 |
-
# Save the updated database to JSON
|
58 |
with open("database.json", "w", encoding="utf-8") as f:
|
59 |
json.dump(data, f, indent=4)
|
60 |
except (IOError, json.JSONEncodeError):
|
61 |
-
# Handle potential errors gracefully
|
62 |
print("Error saving database: Encountered an issue while saving.")
|
63 |
|
64 |
def chat_interface(message):
|
65 |
database = load_database()
|
66 |
|
67 |
-
# Check if the question already exists in the database
|
68 |
if (message, None) not in database:
|
69 |
-
|
70 |
-
response = generate(message, history=[])
|
71 |
database.append((message, response))
|
72 |
save_database(database)
|
73 |
else:
|
74 |
-
# If it does, retrieve the stored response
|
75 |
_, stored_response = next(item for item in database if item[0] == message)
|
76 |
response = stored_response
|
77 |
|
78 |
return response
|
79 |
|
80 |
-
with gr.
|
81 |
-
|
82 |
-
input_textbox = gr.Textbox(label="Your question")
|
83 |
-
output_textbox = gr.Textbox(label="Assistant's response", value="", editable=False)
|
84 |
|
85 |
-
# Use demo.launch instead of demo.queue().launch()
|
86 |
-
demo.launch(fn=chat_interface, inputs=input_textbox, outputs=output_textbox)
|
|
|
1 |
import json
|
|
|
2 |
import gradio as gr
|
3 |
import random
|
4 |
+
from huggingface_hub import InferenceClient
|
5 |
|
6 |
API_URL = "https://api-inference.huggingface.co/models/"
|
7 |
|
8 |
+
client = InferenceClient("mistralai/Mistral-7B-Instruct-v0.1")
|
|
|
|
|
9 |
|
10 |
def format_prompt(message, history):
|
11 |
prompt = "You're a helpful assistant."
|
12 |
for user_prompt, bot_response in history:
|
13 |
+
prompt += f" [INST] {user_prompt} [/INST] {bot_response}</s> "
|
14 |
+
prompt += f" [INST] {message} [/INST]"
|
|
|
15 |
return prompt
|
16 |
|
17 |
def generate(prompt, history, temperature=0.9, max_new_tokens=2048, top_p=0.95, repetition_penalty=1.0):
|
18 |
+
temperature = float(temperature) if temperature > 0 else 0.01
|
|
|
|
|
19 |
top_p = float(top_p)
|
20 |
|
21 |
generate_kwargs = dict(
|
|
|
35 |
for response in stream:
|
36 |
output += response.token.text
|
37 |
yield output
|
|
|
38 |
|
39 |
def load_database():
|
40 |
try:
|
|
|
41 |
with open("database.json", "r", encoding="utf-8") as f:
|
42 |
return json.load(f)
|
43 |
except (FileNotFoundError, json.JSONDecodeError):
|
|
|
44 |
print("Error loading database: File not found or invalid format. Creating an empty database.")
|
45 |
+
return []
|
46 |
|
47 |
def save_database(data):
|
48 |
try:
|
|
|
49 |
with open("database.json", "w", encoding="utf-8") as f:
|
50 |
json.dump(data, f, indent=4)
|
51 |
except (IOError, json.JSONEncodeError):
|
|
|
52 |
print("Error saving database: Encountered an issue while saving.")
|
53 |
|
54 |
def chat_interface(message):
|
55 |
database = load_database()
|
56 |
|
|
|
57 |
if (message, None) not in database:
|
58 |
+
response = next(generate(message, history=[]))
|
|
|
59 |
database.append((message, response))
|
60 |
save_database(database)
|
61 |
else:
|
|
|
62 |
_, stored_response = next(item for item in database if item[0] == message)
|
63 |
response = stored_response
|
64 |
|
65 |
return response
|
66 |
|
67 |
+
with gr.Interface(fn=chat_interface, inputs="textbox", outputs="textbox", title="Chat Interface") as iface:
|
68 |
+
iface.launch()
|
|
|
|
|
69 |
|
|
|
|