ltui 0.1.0

Terminal UI for PyTorch Lightning CSVLogger metrics.

ltui

Terminal UI for live and posthoc visualization of PyTorch Lightning CSVLogger metrics.

ltui is built for SSH and tmux workflows: no browser, no server process, no TensorBoard dependency. Point it at a log root and it recursively finds Lightning metrics.csv files, lets you switch runs and metrics from the keyboard, and renders a single focused terminal plot.

Features

• Recursively discovers Lightning metrics.csv files in nested experiment folders
• Live refresh for growing CSV files, with pause and manual rescan
• Compare multiple runs in one plot
• Groups train_ and val_ metric families by default
• Step/epoch x-axis switching with integer ticks
• Optional EMA smoothing and log scaling
• Fuzzy selectors for runs, metrics, and YAML configs
• Per-root UI state under ~/.local/state/ltui/

Installation

After the first PyPI release:

pip install ltui

From source

Create a dedicated micromamba environment and install the package editable:

git clone https://github.com/erik-ayari/ltui.git
cd ltui

micromamba create -n ltui python=3.11 -y
micromamba activate ltui
pip install -e ".[dev]"

This keeps runtime and development dependencies inside the ltui environment.

From any project directory

Activate the environment, then run ltui against the log root you want to inspect:

micromamba activate ltui
ltui /path/to/log/root

Optional shell helper:

ltui() {
  micromamba run -n ltui ltui "$@"
}

With that function in your shell config, ltui /path/to/log/root works from any directory without manually activating the environment.

Usage

ltui ./lightning_logs
ltui /data/experiments/my_model
ltui ~/runs/my_experiment

On startup, ltui selects the latest modified run and chooses a default loss metric when one is available.

Supported Log Layouts

ltui targets PyTorch Lightning CSVLogger output. Every discovered metrics.csv is treated as one selectable run/version.

Supported examples:

lightning_logs/version_0/metrics.csv
run_a/version_0/metrics.csv
run_a/lightning_logs/version_0/metrics.csv
experiments/group_1/run_a/lightning_logs/version_3/metrics.csv

Display names come from paths relative to the scanned root.

Keybindings

Key	Action
m	Open metric selector
r	Open run/version selector
c	Open config viewer for runs with a unique YAML config
/	Fuzzy search inside selector
arrow keys	Navigate selector
space	Toggle selection in selector, pause/resume on main screen
enter	Apply selector
n	Next selected metric/family
p	Previous selected metric/family
a	Toggle x-axis between step and epoch
d	Toggle dark/light plot theme
s	Toggle smoothing
x	Toggle log-x
y	Toggle log-y
R	Force rescan
q	Quit

Metric Handling

Metric columns are numeric columns except step and epoch. Rows with NaN for the selected metric are dropped, and points are sorted by x-axis before plotting.

Grouped mode is enabled by default. train_loss_step, train_loss_epoch, and val_loss appear as family loss; selecting loss plots whichever train/val sides exist. Use a to switch between step and epoch. On the step axis, validation epoch metrics use the step value from their CSV row, which places validation at the training step where it was logged.

Development

micromamba activate ltui
pytest
python -m build
twine check dist/*

The CLI entrypoint is:

ltui /path/to/log/root