Update model loading to use AutoModelForCausalLM

app.py

@@ -1,4 +1,45 @@
 import streamlit as st
+import pandas as pd
+import asyncio
+from llama_models import process_text
+from dotenv import load_dotenv
+import os
 
-x = st.slider('Select a value')
-st.write(x, 'squared is', x * x)
+# Load environment variables from .env file
+load_dotenv()
+
+async def process_csv(file):
+    df = pd.read_csv(file, header=None)  # Read the CSV file without a header
+    descriptions = df[0].tolist()  # Access the first column directly
+    SAMPLE_SIZE = min(5, len(descriptions))  # Adjust sample size as needed
+    descriptions = descriptions[:SAMPLE_SIZE]
+
+    model_name = "instruction-pretrain/finance-Llama3-8B"  # Ensure this is the correct model name
+
+    results = []
+    for desc in descriptions:
+        result = await process_text(model_name, desc)
+        results.append(result)
+    
+    df['predictions'] = results
+    return df
+
+st.title("Finance Model Deployment")
+
+st.write("""
+### Upload a CSV file with company descriptions to extract key products, geographies, and important keywords:
+""")
+
+uploaded_file = st.file_uploader("Choose a CSV file", type="csv")
+
+if uploaded_file is not None:
+    if st.button("Predict"):
+        with st.spinner("Processing..."):
+            df = asyncio.run(process_csv(uploaded_file))
+            st.write(df)
+            st.download_button(
+                label="Download Predictions as CSV",
+                data=df.to_csv(index=False).encode('utf-8'),
+                file_name='predictions.csv',
+                mime='text/csv'
+            )

llama_models.py

@@ -1,12 +1,12 @@
 import os
-from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import aiohttp
 
 HUGGINGFACE_API_KEY = os.getenv("HUGGINGFACE_API_KEY")
 
 def load_model(model_name):
     tokenizer = AutoTokenizer.from_pretrained(model_name)
-    model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
+    model = AutoModelForCausalLM.from_pretrained(model_name)  # Use AutoModelForCausalLM for Llama
     return tokenizer, model
 
 async def process_text(model_name, text):