Add models.yaml as the single source of truth for leaderboard metadata

benchmarks/fev_bench/models.yaml

@@ -0,0 +1,247 @@
+# Leaderboard metadata for fev-bench (https://huggingface.co/spaces/autogluon/fev-leaderboard).
+# One entry per results CSV in `results/`, keyed by its file name without `.csv`.
+# Format: see `ModelMetadata` in scripts/fev_bench_metadata.py.
+
+# --- Pretrained models ---------------------------------------------------------------------------
+
+chronos-bolt:
+  organization: AWS
+  url: https://huggingface.co/amazon/chronos-bolt-base
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+chronos-2:
+  organization: AWS
+  url: https://huggingface.co/amazon/chronos-2
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+citras-fm:
+  organization: Hitachi
+  url: https://huggingface.co/hitachi-nlp/citras-fm
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: false
+  display_name: CITRAS-FM
+
+flowstate:
+  organization: IBM
+  url: https://huggingface.co/ibm-granite/granite-timeseries-flowstate-r1
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+moirai-2_0:
+  organization: Salesforce
+  url: https://huggingface.co/Salesforce/moirai-2.0-R-small
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: false
+
+sundial-base:
+  organization: Tsinghua University
+  url: https://huggingface.co/thuml/sundial-base-128m
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+t0-beta:
+  organization: The Forecasting Company
+  url: https://huggingface.co/theforecastingcompany/t0-beta
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+tabpfn-ts:
+  organization: Prior Labs
+  url: https://huggingface.co/Prior-Labs/TabPFN-v2-reg
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+tabpfn-ts-3:
+  organization: Prior Labs
+  url: https://huggingface.co/Prior-Labs/tabpfn_3
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: false
+
+timesfm-2_5:
+  organization: Google
+  url: https://huggingface.co/google/timesfm-2.5-200m-pytorch
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+timesfm-3:
+  organization: Google
+  url: https://huggingface.co/google/timesfm-3.0-pytorch
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: false
+
+tirex:
+  organization: NX-AI
+  url: https://huggingface.co/NX-AI/TiRex
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: false
+
+tirex-2:
+  organization: NX-AI
+  url: https://huggingface.co/NX-AI/TiRex-2-fevbench
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+toto-1_0:
+  organization: Datadog
+  url: https://huggingface.co/Datadog/Toto-Open-Base-1.0
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+toto-2_0-2_5b:
+  organization: Datadog
+  url: https://huggingface.co/Datadog/Toto-2.0-2.5B
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+toto-2_0-22m:
+  organization: Datadog
+  url: https://huggingface.co/Datadog/Toto-2.0-22m
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+# Redundant Toto-2.0 checkpoint sizes: kept in the repo for reference, not shown on the leaderboard.
+toto-2_0-1b:
+  organization: Datadog
+  url: https://huggingface.co/Datadog/Toto-2.0-1B
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+  hidden: true
+
+toto-2_0-313m:
+  organization: Datadog
+  url: https://huggingface.co/Datadog/Toto-2.0-313m
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+  hidden: true
+
+toto-2_0-4m:
+  organization: Datadog
+  url: https://huggingface.co/Datadog/Toto-2.0-4m
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+  hidden: true
+
+ts-icl:
+  organization: EDF-Lab
+  url: https://huggingface.co/taharnbl/TS-ICL
+  model_type: pretrained
+  zero_shot: true
+  commercial_use: true
+
+# --- Closed-weight / API models ------------------------------------------------------------------
+
+lingjiang2_api:
+  organization: Alibaba
+  url: TODO
+  model_type: closed-api
+  zero_shot: true
+  commercial_use: true
+
+# --- Task-specific models ------------------------------------------------------------------------
+
+deepar:
+  organization: "—"
+  url: https://github.com/awslabs/gluonts
+  model_type: task-specific
+  zero_shot: false
+  commercial_use: true
+
+patchtst:
+  organization: "—"
+  url: https://github.com/awslabs/gluonts
+  model_type: task-specific
+  zero_shot: false
+  commercial_use: true
+
+tft:
+  organization: "—"
+  url: https://github.com/awslabs/gluonts
+  model_type: task-specific
+  zero_shot: false
+  commercial_use: true
+
+catboost:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/mlforecast
+  model_type: task-specific
+  zero_shot: false
+  commercial_use: true
+
+lightgbm:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/mlforecast
+  model_type: task-specific
+  zero_shot: false
+  commercial_use: true
+
+# --- Statistical models --------------------------------------------------------------------------
+
+autoarima:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true
+
+autoets:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true
+
+autotheta:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true
+
+drift:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true
+
+naive:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true
+
+seasonal_naive:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true
+
+stat_ensemble:
+  organization: "—"
+  url: https://nixtlaverse.nixtla.io/statsforecast/
+  model_type: statistical
+  zero_shot: false
+  commercial_use: true