This video installs Polars and demonstrates how to integrate with Ollama local models to do data analysis.
Code:
conda create -n polarsdf python=3.11 -y && conda activate polarsdf
conda install jupyter -y
jupyter notebook
!pip install polars pyarrow pandas ollama
import polars as pl
from datetime import datetime
df = pl.DataFrame(
{
"date": [
datetime(2024, 7, 1),
datetime(2024, 7, 2),
datetime(2024, 7, 3),
datetime(2024, 7, 4),
datetime(2024, 7, 5),
],
"revenue": [1000.0, 1200.0, 1100.0, 1300.0, 1400.0],
"expenses": [500, 600, 550, 650, 700]
}
)
print(df)
df.write_csv("dummyfinance.csv")
df_csv = pl.read_csv("dummyfinance.csv")
print(df_csv)
df.select(pl.col("*"))
df.select(pl.col("revenue", "date"))
df.filter(
pl.col("date").is_between(datetime(2024, 7, 2), datetime(2024, 7, 4)),
)
import ollama
prompt = "Analyze the financial data: "
for row in df.itertuples(index=True):
prompt += f"Date: {row.date.strftime('%Y-%m-%d')}, Revenue: ${row.revenue}, Expenses: ${row.expenses}; "
prompt += "Predict the Revenue and Expenses for the next date."
response = ollama.generate(model='llama3', prompt=prompt)
predicted_revenue = response['response'].split('Revenue: $')[-1].split('\n')[0]
predicted_expenses = response['response'].split('Expenses: $')[-1].split('\n')[0]
print(predicted_revenue)
print(predicted_expenses)