dvc plots

Commands to visualize and compare plot data

Description

DVC plots commands generate interactive visualizations from your data files. These commands create HTML reports with charts and graphs to help you analyze model performance, training dynamics, and experiment results.

Subcommands

plots show

Generate plots from target files or from plots definitions in dvc.yaml.

All Plots

dvc plots show

Arguments

targets · list

Plots files or plot IDs from dvc.yaml to visualize. Shows all plots by default.

Options

-t, --template · string

Special JSON or HTML schema file to inject with the data. See plot templates.

-x · string

Field name for X axis.

-y · string

Field name for Y axis.

--no-header · flag

Provided CSV or TSV datafile does not have a header.

--title · string

Plot title.

--x-label · string

X axis label.

--y-label · string

Y axis label.

-o, --out · string

Directory to save plots to. Defaults to dvc_plots.

--show-vega · flag

Show output in Vega format.

--open · flag

Open plot file directly in the browser.

--html-template · string

Custom HTML template for VEGA visualization.

---

plots diff

Show multiple versions of a plot by overlaying them in a single image.

Compare Commits

dvc plots diff HEAD~1 HEAD

Arguments

revisions · list

Git commits to plot from. Compares workspace with HEAD by default.

Options

--targets · list

Specific plots to visualize. Accepts any file path or plot name from dvc.yaml file. Shows all tracked plots by default.

-t, --template · string

Special JSON or HTML schema file to inject with the data.

-x · string

Field name for X axis.

-y · string

Field name for Y axis.

--no-header · flag

Provided CSV or TSV datafile does not have a header.

--title · string

Plot title.

--x-label · string

X axis label.

--y-label · string

Y axis label.

-o, --out · string

Directory to save plots to.

--open · flag

Open plot file directly in the browser.

--html-template · string

Custom HTML template for VEGA visualization.

---

plots modify

Modify display properties of data-series plots defined in stages (has no effect on image plots).

Set X/Y Fields

dvc plots modify training.csv -x epoch -y accuracy

Arguments

target · string · required

Plots file to set properties for (defined at the stage level).

Options

-t, --template · string

Plot template to use.

-x · string

Field name for X axis.

-y · string

Field name for Y axis.

--title · string

Plot title.

--x-label · string

X axis label.

--y-label · string

Y axis label.

--unset · list

Unset one or more display properties.

---

plots templates

List built-in plots templates or show JSON specification for one.

List All Templates

dvc plots templates

Arguments

template · string

Template for which to show JSON specification. List all template names by default.

Examples

Visualizing Training Metrics

# Generate plots for training history
dvc plots show training_history.csv -x epoch -y loss --title "Training Loss"

This creates an HTML file with an interactive line chart showing loss over epochs.

Comparing Model Versions

# Compare plots across different model versions
dvc plots diff v1.0 v2.0 v3.0 --targets metrics/accuracy.csv

Creates overlaid plots showing how accuracy evolved across three model versions.

Custom Plot Configuration

# Create scatter plot with custom labels
dvc plots show results.csv \
  -x learning_rate \
  -y accuracy \
  --title "Hyperparameter Tuning Results" \
  --x-label "Learning Rate" \
  --y-label "Model Accuracy"

Auto-Open in Browser

# Generate and immediately view plots
dvc plots show --open

Enable auto-open permanently with: dvc config plots.auto_open true

Supported Data Formats

CSV

training.csv

epoch,loss,accuracy
1,0.8234,0.6543
2,0.5123,0.7654
3,0.3456,0.8123

JSON

metrics.json

[
  {"epoch": 1, "loss": 0.8234, "accuracy": 0.6543},
  {"epoch": 2, "loss": 0.5123, "accuracy": 0.7654},
  {"epoch": 3, "loss": 0.3456, "accuracy": 0.8123}
]

TSV

data.tsv

step	precision	recall
100	0.85	0.82
200	0.89	0.87
300	0.92	0.91

Images

DVC plots also supports image files (PNG, JPEG, SVG) for displaying confusion matrices, visualizations, and other graphics.

Built-in Templates

Linear

Line plot for time series data

Scatter

Scatter plot for correlation analysis

Smooth

Smoothed line plot

Confusion

Confusion matrix heatmap

Bar

Bar chart for categorical data

Custom

Bring your own Vega-Lite spec

Output Structure

When you run dvc plots show, DVC creates:

dvc_plots/
├── index.html          # Main visualization page
└── static/
    ├── data/           # Plot data files
    └── vega/           # Vega specifications

The output directory can be customized with -o, --out option or by setting plots.out_dir in DVC config.

Plot files must be tracked by DVC (in dvc.yaml or .dvc files) to be included in dvc plots diff.

Use Cases

Training Monitoring

Visualize loss curves, accuracy trends, and other training metrics over time.

Model Comparison

Compare performance metrics across different model architectures or hyperparameters.

A/B Testing

Visualize differences between control and experimental model versions.

Performance Analysis

Create confusion matrices, ROC curves, and precision-recall plots.

Related Commands

• dvc metrics - Display metrics in tabular format
• dvc params - Track parameter changes
• dvc exp show - Show experiments with metrics and params