Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,4 +1,3 @@
|
|
1 |
-
import streamlit as st
|
2 |
import subprocess
|
3 |
import os
|
4 |
from io import StringIO
|
@@ -6,6 +5,19 @@ import sys
|
|
6 |
import black
|
7 |
from pylint import lint
|
8 |
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
|
10 |
# Global state to manage communication between Tool Box and Workspace Chat App
|
11 |
if 'chat_history' not in st.session_state:
|
@@ -91,12 +103,12 @@ def chat_interface_with_agent(input_text, agent_name):
|
|
91 |
combined_input = f"{agent_prompt}\n\nUser: {input_text}\nAgent:"
|
92 |
|
93 |
# Truncate input text to avoid exceeding the model's maximum length
|
94 |
-
max_input_length =
|
95 |
input_ids = tokenizer.encode(combined_input, return_tensors="pt")
|
96 |
if input_ids.shape[1] > max_input_length:
|
97 |
input_ids = input_ids[:, :max_input_length]
|
98 |
|
99 |
-
outputs = model.generate(input_ids, max_length=
|
100 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
101 |
return response
|
102 |
|
@@ -127,12 +139,12 @@ def chat_interface(input_text):
|
|
127 |
|
128 |
|
129 |
# Truncate input text to avoid exceeding the model's maximum length
|
130 |
-
max_input_length =
|
131 |
input_ids = tokenizer.encode(input_text, return_tensors="pt")
|
132 |
if input_ids.shape[1] > max_input_length:
|
133 |
input_ids = input_ids[:, :max_input_length]
|
134 |
|
135 |
-
outputs = model.generate(input_ids, max_length=
|
136 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
137 |
return response
|
138 |
|
@@ -264,7 +276,7 @@ def summarize_text(text):
|
|
264 |
return f"Error loading model: {e}"
|
265 |
|
266 |
# Truncate input text to avoid exceeding the model's maximum length
|
267 |
-
max_input_length =
|
268 |
inputs = text
|
269 |
if len(text) > max_input_length:
|
270 |
inputs = text[:max_input_length]
|
@@ -346,7 +358,7 @@ def generate_code(idea):
|
|
346 |
input_ids = tokenizer.encode(input_text, return_tensors="pt")
|
347 |
output_sequences = model.generate(
|
348 |
input_ids=input_ids,
|
349 |
-
max_length=
|
350 |
num_return_sequences=1,
|
351 |
no_repeat_ngram_size=2,
|
352 |
early_stopping=True,
|
|
|
|
|
1 |
import subprocess
|
2 |
import os
|
3 |
from io import StringIO
|
|
|
5 |
import black
|
6 |
from pylint import lint
|
7 |
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
8 |
+
# Initialize chat_history in the session state
|
9 |
+
|
10 |
+
if 'chat_history' not in st.session_state:
|
11 |
+
st.session_state['chat_history'] = []
|
12 |
+
|
13 |
+
# Access and update chat_history
|
14 |
+
chat_history = st.session_state['chat_history']
|
15 |
+
chat_history.append("New message")
|
16 |
+
|
17 |
+
# Display chat history
|
18 |
+
st.write("Chat History:")
|
19 |
+
for message in chat_history:
|
20 |
+
st.write(message)
|
21 |
|
22 |
# Global state to manage communication between Tool Box and Workspace Chat App
|
23 |
if 'chat_history' not in st.session_state:
|
|
|
103 |
combined_input = f"{agent_prompt}\n\nUser: {input_text}\nAgent:"
|
104 |
|
105 |
# Truncate input text to avoid exceeding the model's maximum length
|
106 |
+
max_input_length = max_input_length
|
107 |
input_ids = tokenizer.encode(combined_input, return_tensors="pt")
|
108 |
if input_ids.shape[1] > max_input_length:
|
109 |
input_ids = input_ids[:, :max_input_length]
|
110 |
|
111 |
+
outputs = model.generate(input_ids, max_length=max_input_length, do_sample=True)
|
112 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
113 |
return response
|
114 |
|
|
|
139 |
|
140 |
|
141 |
# Truncate input text to avoid exceeding the model's maximum length
|
142 |
+
max_input_length = max_input_length
|
143 |
input_ids = tokenizer.encode(input_text, return_tensors="pt")
|
144 |
if input_ids.shape[1] > max_input_length:
|
145 |
input_ids = input_ids[:, :max_input_length]
|
146 |
|
147 |
+
outputs = model.generate(input_ids, max_length=max, do_sample=True)
|
148 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
149 |
return response
|
150 |
|
|
|
276 |
return f"Error loading model: {e}"
|
277 |
|
278 |
# Truncate input text to avoid exceeding the model's maximum length
|
279 |
+
max_input_length = max_input_length
|
280 |
inputs = text
|
281 |
if len(text) > max_input_length:
|
282 |
inputs = text[:max_input_length]
|
|
|
358 |
input_ids = tokenizer.encode(input_text, return_tensors="pt")
|
359 |
output_sequences = model.generate(
|
360 |
input_ids=input_ids,
|
361 |
+
max_length=max_length,
|
362 |
num_return_sequences=1,
|
363 |
no_repeat_ngram_size=2,
|
364 |
early_stopping=True,
|