flashtool 0.0.9

A few ready-to use python tools for machine learning

FlashPythonToolbox

A few ready-to use python tools for machine learning

Install

pip install flashtool

Logger (to file)

For every print function, we print a copy to file without changing the whole file.

Logger(filepath = "./log.txt", mode = "w", stdout = None)

• filepath: output path. Default: "./log.txt"
• mode: write mode, e.g., w,w+,a,a+. Default: "w"
• stdout: capture which print source. Default: "sys.stdout"

Example:

from flashtool import Logger
import sys
sys.stdout = Logger("path/to/log/file/log.txt")
sys.stdout = Logger("path/to/log/file/log.txt","w")
sys.stdout = Logger("path/to/log/file/log.txt","w", sys.stdout)

Email Module

Default Sender: [e-mail]. You need a password for this.

# Function api
def send_email(port = 587,password = None,
        sender_email = "[e-mail]",
        smtp_server="smtp.gmail.com",
        receiver_email = None,
        subject="Subject",
        message="Hello! ",
        attachment=None):
    ...

# Example
from flashtool import send_email
send_email(receiver_email="[e-mail]", subject="Hi", message="first email")

Track Job

Track your job. Send you an email when it is done.

Tracking by PID

• Step 1: find your job pid by ps, e.g., ps aux | grep python
• Step 2:

from flashtool import trackpid
trackpid(12345,5,"[e-mail]")
trackpid([1357,2468],5,"[e-mail]")

API:

def trackpid(pid, checktime=1, receiver_email=None, sender_email="[e-mail]", msg=""):
  ...

• pid: pid or pids
• checktime: check pid every checktime seconds
• msg: additional message

Pytorch Utils

GPU Usage

import flashtool as flash; flash.check_torch_memory(brief=True)

• brief: short message (default: Flase)

Auto Script Running

import flashtool as flash;
basescript='script/run.sh'
config={
    "DATA": "subset5"
}
flash.run_script(basescript, config)