My python code is running very slow on millions of records

[shantanu97]

I want to process data through a python that has 2 million rows and more than 100 columns. My code takes 20 minutes to create an output file. I don't know if there is something else that make my code faster, or if I can change something to make it faster. Any help would be greatly appreciated!

df2 = pd.DataFrame()
    for fn in csv_files:  # Looping Over CSV Files
        all_dfs = pd.read_csv(fn, header=None)

        # Finding non-null columns
        non_null_columns = [col for col in all_dfs.columns if all_dfs.loc[:, col].notna().any()]

        # print(non_null_columns)
        for i in range(0, len(all_dfs)):  # Row Loop
            SourceFile = ""
            RowNumber = ""
            ColumnNumber = ""
            Value = ""
            for j in range(0, len(non_null_columns)):  # Column Loop
                SourceFile = Path(fn.name)
                RowNumber = i+1
                ColumnNumber = j+1
                Value = all_dfs.iloc[i, j]
                df2 = df2.append(pd.DataFrame({
                    "SourceFile": [SourceFile],
                    "RowNumber": [RowNumber],
                    "ColumnNumber": [ColumnNumber],
                    "Value": [Value]
                }), ignore_index=True)
                # print(df2)
    df2['Value'].replace('', np.nan, inplace=True)  # Removing Null Value
    df2.dropna(subset=['Value'], inplace=True)
    df2.to_csv(os.path.join(path_save, f"Compiled.csv"), index=False)
    print("Output: Compiled.csv")

Attach python code.