Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,40 +1,38 @@
|
|
1 |
import gradio as gr
|
2 |
-
from transformers import AutoTokenizer, AutoModel
|
3 |
-
import
|
|
|
4 |
|
5 |
# Load the NASA-specific bi-encoder model and tokenizer
|
6 |
bi_encoder_model_name = "nasa-impact/nasa-smd-ibm-st-v2"
|
7 |
bi_tokenizer = AutoTokenizer.from_pretrained(bi_encoder_model_name)
|
8 |
bi_model = AutoModel.from_pretrained(bi_encoder_model_name)
|
9 |
|
10 |
-
#
|
11 |
-
|
12 |
-
gpt2_tokenizer = GPT2Tokenizer.from_pretrained(gpt2_model_name)
|
13 |
-
gpt2_model = GPT2LMHeadModel.from_pretrained(gpt2_model_name)
|
14 |
|
15 |
def encode_text(text):
|
16 |
inputs = bi_tokenizer(text, return_tensors='pt', padding=True, truncation=True, max_length=128)
|
17 |
outputs = bi_model(**inputs)
|
18 |
# Ensure the output is 2D by averaging the last hidden state along the sequence dimension
|
19 |
-
return outputs.last_hidden_state.mean(dim=1).detach().numpy()
|
20 |
|
21 |
def generate_response(user_input, context_embedding):
|
22 |
-
# Create a structured prompt for GPT-
|
23 |
-
|
|
|
24 |
|
25 |
-
# Generate a response using GPT-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
temperature=0.7,
|
32 |
top_p=0.9,
|
33 |
-
|
|
|
34 |
)
|
35 |
-
|
36 |
-
|
37 |
-
return generated_text
|
38 |
|
39 |
def chatbot(user_input, context=""):
|
40 |
context_embedding = encode_text(context) if context else ""
|
@@ -47,7 +45,7 @@ iface = gr.Interface(
|
|
47 |
inputs=[gr.Textbox(lines=2, placeholder="Enter your message here..."), gr.Textbox(lines=2, placeholder="Enter context here (optional)...")],
|
48 |
outputs="text",
|
49 |
title="Context-Aware Dynamic Response Chatbot",
|
50 |
-
description="A chatbot using a NASA-specific bi-encoder model to understand the input context and GPT-
|
51 |
)
|
52 |
|
53 |
# Launch the interface
|
@@ -57,3 +55,4 @@ iface.launch()
|
|
57 |
|
58 |
|
59 |
|
|
|
|
1 |
import gradio as gr
|
2 |
+
from transformers import AutoTokenizer, AutoModel
|
3 |
+
import openai
|
4 |
+
import os
|
5 |
|
6 |
# Load the NASA-specific bi-encoder model and tokenizer
|
7 |
bi_encoder_model_name = "nasa-impact/nasa-smd-ibm-st-v2"
|
8 |
bi_tokenizer = AutoTokenizer.from_pretrained(bi_encoder_model_name)
|
9 |
bi_model = AutoModel.from_pretrained(bi_encoder_model_name)
|
10 |
|
11 |
+
# Set up OpenAI API key
|
12 |
+
openai.api_key = os.getenv('OPENAI_API_KEY')
|
|
|
|
|
13 |
|
14 |
def encode_text(text):
|
15 |
inputs = bi_tokenizer(text, return_tensors='pt', padding=True, truncation=True, max_length=128)
|
16 |
outputs = bi_model(**inputs)
|
17 |
# Ensure the output is 2D by averaging the last hidden state along the sequence dimension
|
18 |
+
return outputs.last_hidden_state.mean(dim=1).detach().numpy().flatten()
|
19 |
|
20 |
def generate_response(user_input, context_embedding):
|
21 |
+
# Create a structured prompt for GPT-4
|
22 |
+
context_str = ' '.join(map(str, context_embedding)) # Convert context embedding to a string
|
23 |
+
combined_input = f"Question: {user_input}\nContext: {context_str}\nAnswer:"
|
24 |
|
25 |
+
# Generate a response using GPT-4
|
26 |
+
response = openai.Completion.create(
|
27 |
+
engine="gpt-4-turbo", # Use GPT-4 engine if available, otherwise use text-davinci-003
|
28 |
+
prompt=combined_input,
|
29 |
+
max_tokens=150,
|
30 |
+
temperature=0.5,
|
|
|
31 |
top_p=0.9,
|
32 |
+
frequency_penalty=0.5,
|
33 |
+
presence_penalty=0.0
|
34 |
)
|
35 |
+
return response.choices[0].text.strip()
|
|
|
|
|
36 |
|
37 |
def chatbot(user_input, context=""):
|
38 |
context_embedding = encode_text(context) if context else ""
|
|
|
45 |
inputs=[gr.Textbox(lines=2, placeholder="Enter your message here..."), gr.Textbox(lines=2, placeholder="Enter context here (optional)...")],
|
46 |
outputs="text",
|
47 |
title="Context-Aware Dynamic Response Chatbot",
|
48 |
+
description="A chatbot using a NASA-specific bi-encoder model to understand the input context and GPT-4 to generate dynamic responses."
|
49 |
)
|
50 |
|
51 |
# Launch the interface
|
|
|
55 |
|
56 |
|
57 |
|
58 |
+
|