Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,5 +1,3 @@
|
|
| 1 |
-
# Write the Streamlit app script
|
| 2 |
-
# Write the Streamlit app script
|
| 3 |
import streamlit as st
|
| 4 |
import pdfplumber
|
| 5 |
import torch
|
|
@@ -13,13 +11,8 @@ import os
|
|
| 13 |
|
| 14 |
print(os.listdir('.'))
|
| 15 |
|
| 16 |
-
|
| 17 |
-
# Download the 'punkt' package
|
| 18 |
nltk.download('punkt')
|
| 19 |
|
| 20 |
-
#openai.api_key = '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'
|
| 21 |
-
|
| 22 |
-
# Define your model architecture
|
| 23 |
class Bert_model(nn.Module):
|
| 24 |
def __init__(self, hidden_size, dropout_rate):
|
| 25 |
super(Bert_model, self).__init__()
|
|
@@ -38,22 +31,19 @@ class Bert_model(nn.Module):
|
|
| 38 |
logits = self.cls_final(pooled_output)
|
| 39 |
return logits
|
| 40 |
|
| 41 |
-
|
| 42 |
-
model_path = "model.pt" # Replace with your actual model path
|
| 43 |
state_dict = torch.load(model_path)
|
| 44 |
-
device = torch.device("cuda:0")
|
| 45 |
|
| 46 |
-
|
| 47 |
-
model = Bert_model(hidden_size=768, dropout_rate=0.1) # Adjust the hidden size to match the saved model
|
| 48 |
model = nn.DataParallel(model)
|
| 49 |
model.load_state_dict(state_dict)
|
| 50 |
model = model.to(device)
|
| 51 |
model.eval()
|
| 52 |
|
| 53 |
-
# Load the tokenizer
|
| 54 |
tokenizer = RobertaTokenizer.from_pretrained('deepset/roberta-base-squad2')
|
| 55 |
|
| 56 |
-
|
| 57 |
def preprocess_pdf(pdf_path, tokenizer):
|
| 58 |
with pdfplumber.open(pdf_path) as pdf:
|
| 59 |
text = " ".join([page.extract_text() for page in pdf.pages[2:]])
|
|
@@ -80,7 +70,7 @@ def translate_text(text, target_language):
|
|
| 80 |
|
| 81 |
def explain_term(term):
|
| 82 |
response = openai.ChatCompletion.create(
|
| 83 |
-
model="gpt-4",
|
| 84 |
messages=[
|
| 85 |
{
|
| 86 |
"role": "system",
|
|
@@ -95,7 +85,8 @@ def explain_term(term):
|
|
| 95 |
return response['choices'][0]['message']['content']
|
| 96 |
|
| 97 |
# Streamlit code to upload file
|
| 98 |
-
|
|
|
|
| 99 |
|
| 100 |
api_key = st.text_input("Enter your OpenAI API key:", type="password")
|
| 101 |
|
|
@@ -103,16 +94,14 @@ if api_key:
|
|
| 103 |
try:
|
| 104 |
openai.api_key = api_key
|
| 105 |
|
| 106 |
-
# Test the API key by making a small request
|
| 107 |
openai.ChatCompletion.create(
|
| 108 |
-
model="gpt-4",
|
| 109 |
messages=[
|
| 110 |
{"role": "system", "content": "You are a helpful assistant."},
|
| 111 |
{"role": "user", "content": "Hello"},
|
| 112 |
],
|
| 113 |
)
|
| 114 |
|
| 115 |
-
# If the above code doesn't raise an exception, the API key is valid
|
| 116 |
st.success("API key is valid!")
|
| 117 |
|
| 118 |
except Exception as e:
|
|
@@ -122,7 +111,6 @@ else:
|
|
| 122 |
|
| 123 |
uploaded_file = st.file_uploader("Choose a PDF file", type="pdf")
|
| 124 |
|
| 125 |
-
# Select language
|
| 126 |
language = st.selectbox('Select your language', ['English', 'French','Chinese','Korean','Spanish','German','Japanese'])
|
| 127 |
|
| 128 |
if uploaded_file is not None:
|
|
@@ -131,7 +119,6 @@ if uploaded_file is not None:
|
|
| 131 |
input_ids, attention_mask, text = preprocess_pdf("temp.pdf", tokenizer)
|
| 132 |
st.write('File successfully uploaded and processed')
|
| 133 |
|
| 134 |
-
# Ask a question
|
| 135 |
question = st.text_input("Enter your question:")
|
| 136 |
|
| 137 |
if question:
|
|
@@ -153,18 +140,10 @@ if uploaded_file is not None:
|
|
| 153 |
predictions.sort(key=lambda pair: pair[1], reverse=True)
|
| 154 |
top_5_sentences = predictions[:13]
|
| 155 |
|
| 156 |
-
#st.write("Top 5 Relevant Sentences:")
|
| 157 |
-
#for sentence, prediction, probabilities in top_5_sentences:
|
| 158 |
-
#st.write(f"Sentence: {sentence}, Prediction: {prediction}, Probability: {probabilities[prediction]}")
|
| 159 |
-
|
| 160 |
-
# Prepare the chat history with the top 3 sentences
|
| 161 |
chat_history = "\n".join([sentence[0] for sentence in top_5_sentences])
|
| 162 |
|
| 163 |
-
# Ask the question using OpenAI API
|
| 164 |
-
#openai.api_key = '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' # Replace with your actual OpenAI API key
|
| 165 |
-
|
| 166 |
response = openai.ChatCompletion.create(
|
| 167 |
-
model="gpt-4",
|
| 168 |
messages=[
|
| 169 |
{"role": "system", "content": "You are a helpful generator which read the short paragraphs and answer the question."},
|
| 170 |
{"role": "user", "content": chat_history},
|
|
@@ -182,7 +161,6 @@ if uploaded_file is not None:
|
|
| 182 |
term = st.text_input("Enter a term you want to define:")
|
| 183 |
|
| 184 |
if term:
|
| 185 |
-
# Define the term using OpenAI API
|
| 186 |
definition = explain_term(term)
|
| 187 |
|
| 188 |
if language != 'English':
|
|
|
|
|
|
|
|
|
|
| 1 |
import streamlit as st
|
| 2 |
import pdfplumber
|
| 3 |
import torch
|
|
|
|
| 11 |
|
| 12 |
print(os.listdir('.'))
|
| 13 |
|
|
|
|
|
|
|
| 14 |
nltk.download('punkt')
|
| 15 |
|
|
|
|
|
|
|
|
|
|
| 16 |
class Bert_model(nn.Module):
|
| 17 |
def __init__(self, hidden_size, dropout_rate):
|
| 18 |
super(Bert_model, self).__init__()
|
|
|
|
| 31 |
logits = self.cls_final(pooled_output)
|
| 32 |
return logits
|
| 33 |
|
| 34 |
+
model_path = "model.pt"
|
|
|
|
| 35 |
state_dict = torch.load(model_path)
|
| 36 |
+
device = torch.device("cuda:0")
|
| 37 |
|
| 38 |
+
model = Bert_model(hidden_size=768, dropout_rate=0.1)
|
|
|
|
| 39 |
model = nn.DataParallel(model)
|
| 40 |
model.load_state_dict(state_dict)
|
| 41 |
model = model.to(device)
|
| 42 |
model.eval()
|
| 43 |
|
|
|
|
| 44 |
tokenizer = RobertaTokenizer.from_pretrained('deepset/roberta-base-squad2')
|
| 45 |
|
| 46 |
+
|
| 47 |
def preprocess_pdf(pdf_path, tokenizer):
|
| 48 |
with pdfplumber.open(pdf_path) as pdf:
|
| 49 |
text = " ".join([page.extract_text() for page in pdf.pages[2:]])
|
|
|
|
| 70 |
|
| 71 |
def explain_term(term):
|
| 72 |
response = openai.ChatCompletion.create(
|
| 73 |
+
model="gpt-4.5-turbo",
|
| 74 |
messages=[
|
| 75 |
{
|
| 76 |
"role": "system",
|
|
|
|
| 85 |
return response['choices'][0]['message']['content']
|
| 86 |
|
| 87 |
# Streamlit code to upload file
|
| 88 |
+
|
| 89 |
+
st.title('FinChat')
|
| 90 |
|
| 91 |
api_key = st.text_input("Enter your OpenAI API key:", type="password")
|
| 92 |
|
|
|
|
| 94 |
try:
|
| 95 |
openai.api_key = api_key
|
| 96 |
|
|
|
|
| 97 |
openai.ChatCompletion.create(
|
| 98 |
+
model="gpt-4.5-turbo",
|
| 99 |
messages=[
|
| 100 |
{"role": "system", "content": "You are a helpful assistant."},
|
| 101 |
{"role": "user", "content": "Hello"},
|
| 102 |
],
|
| 103 |
)
|
| 104 |
|
|
|
|
| 105 |
st.success("API key is valid!")
|
| 106 |
|
| 107 |
except Exception as e:
|
|
|
|
| 111 |
|
| 112 |
uploaded_file = st.file_uploader("Choose a PDF file", type="pdf")
|
| 113 |
|
|
|
|
| 114 |
language = st.selectbox('Select your language', ['English', 'French','Chinese','Korean','Spanish','German','Japanese'])
|
| 115 |
|
| 116 |
if uploaded_file is not None:
|
|
|
|
| 119 |
input_ids, attention_mask, text = preprocess_pdf("temp.pdf", tokenizer)
|
| 120 |
st.write('File successfully uploaded and processed')
|
| 121 |
|
|
|
|
| 122 |
question = st.text_input("Enter your question:")
|
| 123 |
|
| 124 |
if question:
|
|
|
|
| 140 |
predictions.sort(key=lambda pair: pair[1], reverse=True)
|
| 141 |
top_5_sentences = predictions[:13]
|
| 142 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 143 |
chat_history = "\n".join([sentence[0] for sentence in top_5_sentences])
|
| 144 |
|
|
|
|
|
|
|
|
|
|
| 145 |
response = openai.ChatCompletion.create(
|
| 146 |
+
model="gpt-4.5-turbo",
|
| 147 |
messages=[
|
| 148 |
{"role": "system", "content": "You are a helpful generator which read the short paragraphs and answer the question."},
|
| 149 |
{"role": "user", "content": chat_history},
|
|
|
|
| 161 |
term = st.text_input("Enter a term you want to define:")
|
| 162 |
|
| 163 |
if term:
|
|
|
|
| 164 |
definition = explain_term(term)
|
| 165 |
|
| 166 |
if language != 'English':
|