tasks/README.md

tasks — benchmark catalog

Runtime data loading uses this tree (tempus_bench/tasks/).

Task CSV data is hosted on Hugging Face: Smlcrm/tempus_bench_tasks. On first benchmark run, missing CSVs are downloaded automatically.

Layout

tasks/
  univariate/<task_id>/
    task.yaml
    metadata.json
    <csv>
  multivariate/<task_id>/
    task.yaml
    metadata.json
    <csv>
  covariate/<task_id>/
    task.yaml
    metadata.json
    <csv>   # copy of sibling multivariate data (covariate_*.csv)
  datasets_catalog.csv
  all_tasks.yaml

CSV schema

Each task CSV has one row per variate:

column	description
variable_name	variate identifier (matches yaml lists)
variable_unit	unit string for the series
timestamps	JSON array of ISO 8601 UTC strings
values	JSON array of numbers (or null for missing)

There is no variable_type column. Targets and covariates are selected from task.yaml at load time.

Flat task.yaml schema

Univariate

task:
  context_window: 512
  forecast_horizon: 64
  handle_missing: interpolate
  normalization_method: standard   # standard | none
  file_name: <csv>
  target_variable_names: [name]

Multivariate (joint forecasting)

task:
  context_window: 32
  forecast_horizon: 8
  handle_missing: interpolate
  normalization_method: standard
  file_name: multivariate_<suffix>.csv
  target_variable_names: [all variates]

Covariate (single target + covariates)

task:
  context_window: 32
  forecast_horizon: 8
  handle_missing: interpolate
  normalization_method: standard
  file_name: covariate_<suffix>.csv
  target_variable_name: PM25
  covariate_variable_names: [remaining variates used as covariates]

Naming: multivariate_transport_monthly_foo pairs with covariate/covariate_transport_monthly_foo and covariate_transport_monthly_foo.csv.

Discovery

Wildcard discovery (*) registers every folder under univariate/, multivariate/, and covariate/ (30 tasks total: 10 + 10 + 10).

Examples:

• multivariate/multivariate_transport_monthly_airline_baggage_complaints — joint multivariate
• covariate/covariate_transport_monthly_airline_baggage_complaints — single target + covariates

Regenerating catalog artifacts

python tempus_bench/scripts/generate_tasks_metadata.py
python tempus_bench/scripts/generate_tasks_task_yaml.py

Uploading task data to Hugging Face

python hugging_face_upload/upload_dataset.py --repo-id Smlcrm/tempus_bench_tasks

See hugging_face_upload/README.md for details.