plot.py

import argparse
import os
import pandas as pd
import numpy as np
import socket
import matplotlib.pyplot as plt
from pathlib import Path


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--path', type=str, default='./results', help="Specify the path to load the results")
    parser.add_argument('-t','--timestamp', type=str, default=None, help="Specify the timestamp when the experiement started")
    args = parser.parse_args()
    
    if args.timestamp is None:
        log = args.path+'/'+np.sort(os.listdir(args.path))[-1]
    else:
        log = args.path+'/'+args.timestamp

    log_path = Path(log)
    csv_files = sorted([p for p in log_path.glob("*.csv")])

    # Separate USAGE.csv (per‑user) from per‑device CSVs (CPU.csv, GPU_*.csv)
    usage_file = next((p for p in csv_files if p.name.upper() == "USAGE.CSV"), None)
    device_files = [p for p in csv_files if p is not usage_file]

    # ---- Plot per‑device utilisation/memory (original behaviour) ----
    if device_files:
        dfs = [pd.read_csv(fp) for fp in device_files]
        devices = [fp.stem for fp in device_files]

        fig1, ax1 = plt.subplots(2, figsize=(12, 10))
        fig1.suptitle(f"{socket.gethostname()} – Device metrics")

        for df, d in zip(dfs, devices):
            df['Time'] = pd.to_datetime(df['Time'], format="%H:%M-%d-%m-%Y")
            ax1[0].plot(df['Time'], df['Util'], label=d)
            ax1[1].plot(df['Time'], df['Mem'],  label=d)

        ax1[0].legend();  ax1[0].set_ylabel("Utilisation %")
        ax1[1].legend();  ax1[1].set_ylabel("Memory %")
        fig1.tight_layout()
        fig1.savefig(log_path / "device_plot.pdf")
        print(f"Device plots saved to {log_path/'device_plot.pdf'}")

    # ---- Plot per‑user CPU / GPU usage from USAGE.csv ----
    if usage_file is not None and usage_file.exists():
        df_u = pd.read_csv(usage_file)
        df_u['Time'] = pd.to_datetime(df_u['Time'], format="%H:%M-%d-%m-%Y")
        users = df_u['User'].unique()

        fig2, ax2 = plt.subplots(3, figsize=(12, 12))
        fig2.suptitle(f"{socket.gethostname()} – Per‑user metrics")

        for u in users:
            sub = df_u[df_u['User'] == u]
            ax2[0].plot(sub['Time'], sub['CPU_Util'], label=u)
            ax2[1].plot(sub['Time'], sub['CPU_Mem'], label=u)
            ax2[2].plot(sub['Time'], sub['GPU_Mem_GB'], label=u)

        ax2[0].set_ylabel("CPU Util %")
        ax2[1].set_ylabel("CPU Mem %")
        ax2[2].set_ylabel("GPU Mem (GB)")
        for a in ax2: a.legend()
        fig2.tight_layout()
        fig2.savefig(log_path / "usage_plot.pdf")
        print(f"Per‑user plots saved to {log_path/'usage_plot.pdf'}")