Metrics and Value Types

Dalva supports different value types that are rendered differently in the web interface.

Supported Value Types

Python Type	attribute_type	Storage Column	Notes
float	float	float_value	Summary metrics (step=None)
int	int	int_value	Summary metrics (step=None)
str	string	string_value	Summary metrics (step=None)
bool	bool	bool_value	Summary metrics (step=None)
float	float_series	float_value	Time-series metrics (step=int)
int	int_series	int_value	Time-series metrics (step=int)
str	string_series	string_value	Time-series metrics (step=int)
bool	bool_series	bool_value	Time-series metrics (step=int)

Series vs Scalar Types

The step parameter determines whether a metric is stored as a scalar or series:

• step=None → Scalar types (float, int, string, bool)
• Single values stored without a step index
• Displayed as a single value card in the UI

• Use for: final metrics, summary statistics, best model performance

• step=<int> → Series types (float_series, int_series, string_series, bool_series)

• Values stored with a step index (e.g., step=0, step=1, step=2)
• Displayed as an interactive chart in the UI
• Use for: training curves, epoch-level metrics, iteration metrics

How Step Determines Type

The type is automatically determined by the step parameter:

# step=None → scalar type
run.log({"best_accuracy": 0.95})  # attribute_type="float"

# step=0 → series type
run.log({"accuracy": 0.85}, step=0)  # attribute_type="float_series"
run.log({"accuracy": 0.87}, step=1)  # attribute_type="float_series"
run.log({"accuracy": 0.89}, step=2)  # attribute_type="float_series"

Example: Logging Training Metrics

import dalva

run = dalva.init(project="my-project", name="experiment-1")

# Log training metrics at each epoch (series type → chart)
for epoch in range(100):
    train_loss = 1.0 / (epoch + 1)
    train_acc = min(0.99, epoch / 100)
    run.log({
        "train/loss": train_loss,
        "train/accuracy": train_acc
    }, step=epoch)

# Log summary metrics (scalar type → value card)
run.log({
    "best_model/accuracy": 0.95,
    "best_model/epoch": 87,
    "training_completed": True
})  # step=None by default

run.finish()

Type Enforcement

Once a metric key is used for a run, it must remain consistent:

run = dalva.init(project="my-project", name="experiment")

# First log with step=0 → float_series
run.log({"loss": 0.5}, step=0)

# Later, log without step → float (ERROR! Type mismatch)
run.log({"loss": 0.1})  # Raises ValueError: attribute_type mismatch

This prevents accidental type mismatches where you forget to add step.

Frontend Rendering

Scalar Metrics (step=None)

Scalar metrics are displayed as value cards in the run's Overview tab:

Metric	Value
best_model/accuracy	0.9523
best_model/hter	0.0821

Series Metrics (step=int)

Series metrics are displayed as interactive charts in the run's Metrics tab. The chart shows the metric value over steps, with hover tooltips for exact values.

Common Patterns

Pattern 1: Training Loop with Summary

run = dalva.init(project="my-project", name="training")

best_val_acc = 0.0
for epoch in range(100):
    # Training
    train_loss = compute_loss()
    run.log({"train/loss": train_loss}, step=epoch)

    # Validation
    val_acc = evaluate()
    run.log({"val/accuracy": val_acc}, step=epoch)

    if val_acc > best_val_acc:
        best_val_acc = val_acc

# Summary metrics (no step)
run.log({
    "best_val_accuracy": best_val_acc,
    "final_epoch": epoch
})

run.finish()

Pattern 2: Hyperparameters as Config

run = dalva.init(
    project="my-project",
    name="experiment-1",
    config={
        "model": "resnet50",
        "lr": 0.001,
        "batch_size": 32,
        "optimizer": "adam"
    }
)
# Config is stored separately from metrics and displayed in the "Config" tab

Metric Naming Conventions

While not required, a common convention uses path-like names:

Prefix	Usage	Example
train/	Training metrics	train/loss, train/accuracy
val/	Validation metrics	val/loss, val/accuracy
test/	Test metrics	test/accuracy, test/f1
best_model/	Final/best model metrics	best_model/accuracy
sys/	System info	sys/hostname, sys/running_time