Readers: Accessing Run Data

Readers: Accessing Run Data#

notata provides two main classes for accessing run data:

  • LogReader: For accessing data and metadata from a single run directory.

  • ExperimentReader: For accessing multiple runs within an experiment directory.

LogReader#

LogReader allows you to access parameters, metadata, arrays, artifacts, and plots from a single run.

Example:

from notata.reader import LogReader

reader = LogReader("outputs/log_my_run")
print("Run ID:", reader.run_id)
print("Metadata:", reader.meta)
print("Parameters:", reader.params)
print("Available arrays:", reader.arrays)

To load an array:

energy = reader.load_array("energy")
print("Energy array:", energy)

To load a JSON artifact:

metrics = reader.load_json("metrics")
print("Metrics:", metrics)

ExperimentReader#

ExperimentReader allows you to iterate over multiple runs in an experiment directory.

Example:

from notata.reader import ExperimentReader

exp = ExperimentReader("outputs/my_experiment")
print("Number of runs:", len(exp))

for run in exp:
    print("Run ID:", run.run_id)
    print("Metadata:", run.meta)

To access a specific run:

run = exp["1"]
print("Run ID:", run.run_id)
print("Parameters:", run.params)

To access parameters and metadata for all runs:

print("Parameters:", exp.params)
print("Metadata:", exp.meta)

Best Practices#

  • Use LogReader for detailed inspection of a single run.

  • Use ExperimentReader for batch processing or querying multiple runs.

  • Always check meta for the status and runtime of each run.

  • Use arrays and artifacts to explore available data and outputs.

Contents