Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,33 +1,32 @@
|
|
1 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer
|
2 |
-
import torch
|
3 |
-
import gradio as gr
|
4 |
-
|
5 |
-
# Load model and tokenizer
|
6 |
-
tokenizer = AutoTokenizer.from_pretrained("microsoft/DialoGPT-medium")
|
7 |
-
model = AutoModelForCausalLM.from_pretrained("microsoft/DialoGPT-medium")
|
8 |
-
|
9 |
-
chat_history_ids = None
|
10 |
-
|
11 |
def chat(user_input, history=[]):
|
12 |
global chat_history_ids
|
13 |
|
14 |
-
# Encode input
|
15 |
new_input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors='pt')
|
|
|
|
|
16 |
bot_input_ids = torch.cat([chat_history_ids, new_input_ids], dim=-1) if chat_history_ids is not None else new_input_ids
|
17 |
|
18 |
-
# Generate
|
19 |
-
chat_history_ids = model.generate(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
20 |
response = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
|
21 |
|
|
|
|
|
|
|
|
|
22 |
history.append((user_input, response))
|
23 |
return history, history
|
24 |
|
25 |
-
# Chat UI
|
26 |
-
chatbot_ui = gr.ChatInterface(
|
27 |
-
fn=chat,
|
28 |
-
title="Teen Mental Health Chatbot 🤖💬",
|
29 |
-
description="Talk to a supportive AI. Not a replacement for professional help.",
|
30 |
-
)
|
31 |
-
|
32 |
-
chatbot_ui.launch()
|
33 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
def chat(user_input, history=[]):
|
2 |
global chat_history_ids
|
3 |
|
4 |
+
# Encode user input
|
5 |
new_input_ids = tokenizer.encode(user_input + tokenizer.eos_token, return_tensors='pt')
|
6 |
+
|
7 |
+
# Append to chat history or start new
|
8 |
bot_input_ids = torch.cat([chat_history_ids, new_input_ids], dim=-1) if chat_history_ids is not None else new_input_ids
|
9 |
|
10 |
+
# Generate response (LIMIT max_length and num_return_sequences)
|
11 |
+
chat_history_ids = model.generate(
|
12 |
+
bot_input_ids,
|
13 |
+
max_length=1000, # Can reduce to 500 if needed
|
14 |
+
pad_token_id=tokenizer.eos_token_id,
|
15 |
+
do_sample=True,
|
16 |
+
top_k=50,
|
17 |
+
top_p=0.95,
|
18 |
+
temperature=0.7,
|
19 |
+
num_return_sequences=1
|
20 |
+
)
|
21 |
+
|
22 |
+
# Decode only the new response part
|
23 |
response = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
|
24 |
|
25 |
+
# Truncate response if it’s too long (hard limit)
|
26 |
+
if len(response) > 1000:
|
27 |
+
response = response[:1000] + "..."
|
28 |
+
|
29 |
history.append((user_input, response))
|
30 |
return history, history
|
31 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
32 |
|