Update app.py
Browse files
app.py
CHANGED
|
@@ -5,40 +5,74 @@ from transformers import (
|
|
| 5 |
AutoModelForSeq2SeqLM,
|
| 6 |
)
|
| 7 |
import torch
|
|
|
|
| 8 |
|
| 9 |
-
# Define the model names and
|
| 10 |
MODEL_MAPPING = {
|
| 11 |
-
"text2shellcommands": "t5-small", # Example seq2seq model
|
| 12 |
-
"pentest_ai": "bert-base-uncased", # Example
|
| 13 |
}
|
| 14 |
|
| 15 |
-
#
|
| 16 |
def select_model():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
st.sidebar.header("Model Configuration")
|
| 18 |
-
|
|
|
|
| 19 |
|
| 20 |
|
| 21 |
-
#
|
| 22 |
@st.cache_resource
|
| 23 |
def load_model_and_tokenizer(model_name):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 24 |
try:
|
| 25 |
-
# Load the tokenizer
|
| 26 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
|
|
|
|
| 27 |
if "t5" in model_name or "seq2seq" in model_name:
|
|
|
|
| 28 |
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
| 29 |
else:
|
|
|
|
| 30 |
model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
| 31 |
|
| 32 |
return tokenizer, model
|
| 33 |
except Exception as e:
|
| 34 |
-
|
|
|
|
| 35 |
return None, None
|
| 36 |
|
| 37 |
|
| 38 |
-
#
|
| 39 |
def predict_with_model(user_input, model, tokenizer, model_choice):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 40 |
if model_choice == "text2shellcommands":
|
| 41 |
-
# Generate shell commands (
|
| 42 |
inputs = tokenizer(user_input, return_tensors="pt", padding=True, truncation=True)
|
| 43 |
with torch.no_grad():
|
| 44 |
outputs = model.generate(**inputs)
|
|
@@ -57,25 +91,85 @@ def predict_with_model(user_input, model, tokenizer, model_choice):
|
|
| 57 |
}
|
| 58 |
|
| 59 |
|
| 60 |
-
#
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
def main():
|
| 62 |
st.title("AI Model Inference Dashboard")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
|
| 64 |
# Model selection
|
| 65 |
model_choice = select_model()
|
| 66 |
model_name = MODEL_MAPPING.get(model_choice)
|
| 67 |
tokenizer, model = load_model_and_tokenizer(model_name)
|
| 68 |
|
| 69 |
-
# Input text
|
| 70 |
-
|
| 71 |
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
|
| 76 |
-
|
| 77 |
-
|
| 78 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 79 |
|
| 80 |
|
| 81 |
if __name__ == "__main__":
|
|
|
|
| 5 |
AutoModelForSeq2SeqLM,
|
| 6 |
)
|
| 7 |
import torch
|
| 8 |
+
import os
|
| 9 |
|
| 10 |
+
# Define the model names and their corresponding Hugging Face models
|
| 11 |
MODEL_MAPPING = {
|
| 12 |
+
"text2shellcommands": "t5-small", # Example seq2seq model for generating shell commands
|
| 13 |
+
"pentest_ai": "bert-base-uncased", # Example classification model for pentesting tasks
|
| 14 |
}
|
| 15 |
|
| 16 |
+
# Function to create a sidebar for model selection
|
| 17 |
def select_model():
|
| 18 |
+
"""
|
| 19 |
+
Adds a dropdown to the Streamlit sidebar for selecting a model.
|
| 20 |
+
Returns:
|
| 21 |
+
str: The selected model key from MODEL_MAPPING.
|
| 22 |
+
"""
|
| 23 |
st.sidebar.header("Model Configuration")
|
| 24 |
+
selected_model = st.sidebar.selectbox("Select a model", list(MODEL_MAPPING.keys()))
|
| 25 |
+
return selected_model
|
| 26 |
|
| 27 |
|
| 28 |
+
# Function to load the model and tokenizer with caching
|
| 29 |
@st.cache_resource
|
| 30 |
def load_model_and_tokenizer(model_name):
|
| 31 |
+
"""
|
| 32 |
+
Loads the tokenizer and model for the specified Hugging Face model name.
|
| 33 |
+
Uses caching to optimize performance.
|
| 34 |
+
|
| 35 |
+
Args:
|
| 36 |
+
model_name (str): The name of the Hugging Face model to load.
|
| 37 |
+
|
| 38 |
+
Returns:
|
| 39 |
+
tuple: A tokenizer and model instance.
|
| 40 |
+
"""
|
| 41 |
try:
|
| 42 |
+
# Load the tokenizer
|
| 43 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
| 44 |
+
|
| 45 |
+
# Determine the correct model class to use
|
| 46 |
if "t5" in model_name or "seq2seq" in model_name:
|
| 47 |
+
# Load a sequence-to-sequence model
|
| 48 |
model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
|
| 49 |
else:
|
| 50 |
+
# Load a sequence classification model
|
| 51 |
model = AutoModelForSequenceClassification.from_pretrained(model_name)
|
| 52 |
|
| 53 |
return tokenizer, model
|
| 54 |
except Exception as e:
|
| 55 |
+
# Display an error message in the Streamlit app
|
| 56 |
+
st.error(f"An error occurred while loading the model or tokenizer: {str(e)}")
|
| 57 |
return None, None
|
| 58 |
|
| 59 |
|
| 60 |
+
# Function to handle predictions based on the selected model
|
| 61 |
def predict_with_model(user_input, model, tokenizer, model_choice):
|
| 62 |
+
"""
|
| 63 |
+
Handles predictions using the loaded model and tokenizer.
|
| 64 |
+
|
| 65 |
+
Args:
|
| 66 |
+
user_input (str): Text input from the user.
|
| 67 |
+
model: Loaded Hugging Face model.
|
| 68 |
+
tokenizer: Loaded Hugging Face tokenizer.
|
| 69 |
+
model_choice (str): Selected model key from MODEL_MAPPING.
|
| 70 |
+
|
| 71 |
+
Returns:
|
| 72 |
+
dict: A dictionary containing the prediction results.
|
| 73 |
+
"""
|
| 74 |
if model_choice == "text2shellcommands":
|
| 75 |
+
# Generate shell commands (Seq2Seq task)
|
| 76 |
inputs = tokenizer(user_input, return_tensors="pt", padding=True, truncation=True)
|
| 77 |
with torch.no_grad():
|
| 78 |
outputs = model.generate(**inputs)
|
|
|
|
| 91 |
}
|
| 92 |
|
| 93 |
|
| 94 |
+
# Function to process uploaded files
|
| 95 |
+
def process_uploaded_file(uploaded_file):
|
| 96 |
+
"""
|
| 97 |
+
Reads and processes the uploaded file. Supports text and CSV files.
|
| 98 |
+
|
| 99 |
+
Args:
|
| 100 |
+
uploaded_file: The uploaded file.
|
| 101 |
+
|
| 102 |
+
Returns:
|
| 103 |
+
str: The content of the file as a string.
|
| 104 |
+
"""
|
| 105 |
+
try:
|
| 106 |
+
if uploaded_file is not None:
|
| 107 |
+
file_type = uploaded_file.type
|
| 108 |
+
|
| 109 |
+
# Text file processing
|
| 110 |
+
if "text" in file_type:
|
| 111 |
+
content = uploaded_file.read().decode("utf-8")
|
| 112 |
+
return content
|
| 113 |
+
# CSV file processing
|
| 114 |
+
elif "csv" in file_type:
|
| 115 |
+
import pandas as pd
|
| 116 |
+
df = pd.read_csv(uploaded_file)
|
| 117 |
+
return df.to_string() # Convert the dataframe to string
|
| 118 |
+
else:
|
| 119 |
+
st.error("Unsupported file type. Please upload a text or CSV file.")
|
| 120 |
+
return None
|
| 121 |
+
except Exception as e:
|
| 122 |
+
st.error(f"Error processing file: {e}")
|
| 123 |
+
return None
|
| 124 |
+
|
| 125 |
+
|
| 126 |
+
# Main function to define the Streamlit app
|
| 127 |
def main():
|
| 128 |
st.title("AI Model Inference Dashboard")
|
| 129 |
+
st.markdown(
|
| 130 |
+
"""
|
| 131 |
+
This dashboard allows you to interact with different AI models for inference tasks,
|
| 132 |
+
such as generating shell commands or performing text classification.
|
| 133 |
+
"""
|
| 134 |
+
)
|
| 135 |
|
| 136 |
# Model selection
|
| 137 |
model_choice = select_model()
|
| 138 |
model_name = MODEL_MAPPING.get(model_choice)
|
| 139 |
tokenizer, model = load_model_and_tokenizer(model_name)
|
| 140 |
|
| 141 |
+
# Input text area or file upload
|
| 142 |
+
input_choice = st.radio("Choose Input Method", ("Text Input", "Upload File"))
|
| 143 |
|
| 144 |
+
if input_choice == "Text Input":
|
| 145 |
+
user_input = st.text_area("Enter your text input:", placeholder="Type your text here...")
|
| 146 |
+
|
| 147 |
+
# Handle prediction after submit
|
| 148 |
+
submit_button = st.button("Submit")
|
| 149 |
+
|
| 150 |
+
if submit_button and user_input:
|
| 151 |
+
st.write("### Prediction Results:")
|
| 152 |
+
result = predict_with_model(user_input, model, tokenizer, model_choice)
|
| 153 |
+
for key, value in result.items():
|
| 154 |
+
st.write(f"**{key}:** {value}")
|
| 155 |
+
|
| 156 |
+
elif input_choice == "Upload File":
|
| 157 |
+
uploaded_file = st.file_uploader("Choose a text or CSV file", type=["txt", "csv"])
|
| 158 |
+
|
| 159 |
+
# Handle prediction after submit
|
| 160 |
+
submit_button = st.button("Submit")
|
| 161 |
+
|
| 162 |
+
if submit_button and uploaded_file:
|
| 163 |
+
file_content = process_uploaded_file(uploaded_file)
|
| 164 |
+
if file_content:
|
| 165 |
+
st.write("### File Content:")
|
| 166 |
+
st.write(file_content)
|
| 167 |
+
result = predict_with_model(file_content, model, tokenizer, model_choice)
|
| 168 |
+
st.write("### Prediction Results:")
|
| 169 |
+
for key, value in result.items():
|
| 170 |
+
st.write(f"**{key}:** {value}")
|
| 171 |
+
else:
|
| 172 |
+
st.info("No valid content found in the file.")
|
| 173 |
|
| 174 |
|
| 175 |
if __name__ == "__main__":
|