Skip to content

Malav-P/mlobs

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

mlc

ML training monitor. Wraps your training script, streams output, and gives some charts. Quick and easy. No Wandb, no changes to your code, just reading from terminal output.

Quick Start

Install first:

curl -fsSL https://raw.githubusercontent.com/Malav-P/mlobs/main/install.sh | sh

Then run your training script:

mlc run python train.py --lr 1e-3
# OR
mlc run python main.py
# OR
mlc run <whatever your command is>

Info

  • Run data is stored at ~/.mlc/runs
  • To guarantee that your metrics are captured, print in jsonl format:
        print(json.dumps({
                            "step":      global_step,
                            "epoch":     epoch,
                            "loss":      round(loss, 6),
                            "acc":       round(acc, 6),
                            "lr":        round(lr, 6),
                            "grad_norm": round(float(np.linalg.norm(grad)), 6),
                        }), 
             flush=True)

Uninstall

sudo rm /usr/local/bin/mlc
rm -rf ~/.mlc/runs

About

machine learning observability

Resources

Stars

Watchers

Forks

Packages

 
 
 

Contributors