Spaces:
Runtime error
Runtime error
Deploy Space: update app.py
Browse files
app.py
CHANGED
|
@@ -1,3 +1,4 @@
|
|
|
|
|
| 1 |
import os
|
| 2 |
import sys
|
| 3 |
from pathlib import Path
|
|
@@ -53,27 +54,52 @@ def _dataset_tables(grouped: dict, datasets: list[str]) -> tuple[list, list]:
|
|
| 53 |
return values_tables, rank_tables
|
| 54 |
|
| 55 |
|
| 56 |
-
def
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 60 |
root = Path(root_dir)
|
| 61 |
if not root.exists():
|
| 62 |
-
return
|
| 63 |
|
| 64 |
for model_dir in sorted(root.iterdir()):
|
| 65 |
if not model_dir.is_dir():
|
| 66 |
continue
|
|
|
|
| 67 |
for snapshot_dataset, snapshot in load_forecast_snapshots(root_dir, model_dir.name).items():
|
| 68 |
-
|
| 69 |
-
if
|
| 70 |
continue
|
| 71 |
-
|
| 72 |
-
|
| 73 |
-
|
| 74 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 75 |
|
| 76 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 77 |
|
| 78 |
|
| 79 |
def refresh_leaderboard() -> tuple:
|
|
@@ -98,23 +124,21 @@ def refresh_leaderboard() -> tuple:
|
|
| 98 |
load_live_aggregate_table(RESULTS_PATH, "rank"), # [13]
|
| 99 |
]
|
| 100 |
|
| 101 |
-
# [
|
| 102 |
for domain, domain_datasets in DOMAIN_GROUPS.items():
|
| 103 |
outputs.append(build_domain_status_md(domain, domain_datasets, RESULTS_PATH))
|
| 104 |
|
| 105 |
value_tables, rank_tables = _dataset_tables(grouped, datasets)
|
| 106 |
-
forecast_plots = _dataset_forecast_plots(RESULTS_PATH, datasets)
|
| 107 |
# [14+D .. 14+D+N-1] value tables
|
| 108 |
outputs.extend(value_tables)
|
| 109 |
# [14+D+N .. 14+D+2N-1] rank tables
|
| 110 |
outputs.extend(rank_tables)
|
| 111 |
-
# [14+D+2N .. 14+D+3N-1] forecast plots
|
| 112 |
-
outputs.extend(forecast_plots)
|
| 113 |
return tuple(outputs)
|
| 114 |
|
| 115 |
|
| 116 |
grouped_initial = get_grouped_dfs(RESULTS_PATH)
|
| 117 |
DATASETS: list[str] = grouped_initial["datasets"]
|
|
|
|
| 118 |
DOMAIN_GROUPS = group_datasets_by_domain(DATASETS, RESULTS_PATH)
|
| 119 |
N_DOMAINS = len(DOMAIN_GROUPS)
|
| 120 |
INITIAL_OUTPUTS = refresh_leaderboard()
|
|
@@ -127,7 +151,6 @@ with demo:
|
|
| 127 |
|
| 128 |
dataset_value_dfs: list[gr.Dataframe] = []
|
| 129 |
dataset_rank_dfs: list[gr.Dataframe] = []
|
| 130 |
-
dataset_forecast_plots: list[gr.Plot] = []
|
| 131 |
domain_status_mds: list[gr.Markdown] = []
|
| 132 |
dataset_idx = 0
|
| 133 |
|
|
@@ -283,11 +306,23 @@ with demo:
|
|
| 283 |
show_label=False,
|
| 284 |
)
|
| 285 |
)
|
| 286 |
-
|
| 287 |
-
|
| 288 |
-
|
| 289 |
-
|
| 290 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 291 |
)
|
| 292 |
|
| 293 |
dataset_idx += 1
|
|
@@ -321,7 +356,6 @@ with demo:
|
|
| 321 |
*domain_status_mds,
|
| 322 |
*dataset_value_dfs,
|
| 323 |
*dataset_rank_dfs,
|
| 324 |
-
*dataset_forecast_plots,
|
| 325 |
]
|
| 326 |
|
| 327 |
demo.load(refresh_leaderboard, inputs=[], outputs=refresh_outputs)
|
|
|
|
| 1 |
+
import json
|
| 2 |
import os
|
| 3 |
import sys
|
| 4 |
from pathlib import Path
|
|
|
|
| 54 |
return values_tables, rank_tables
|
| 55 |
|
| 56 |
|
| 57 |
+
def _model_label(root: Path, model_slug: str) -> str:
|
| 58 |
+
config_path = root / model_slug / "config.json"
|
| 59 |
+
if not config_path.exists():
|
| 60 |
+
return model_slug
|
| 61 |
+
try:
|
| 62 |
+
config = json.loads(config_path.read_text())
|
| 63 |
+
return str(config.get("model") or model_slug)
|
| 64 |
+
except Exception:
|
| 65 |
+
return model_slug
|
| 66 |
+
|
| 67 |
+
|
| 68 |
+
def _forecast_choices(root_dir: str, datasets: list[str]) -> dict[str, dict[str, dict]]:
|
| 69 |
+
dataset_keys = {dataset.split("/")[0] for dataset in datasets}
|
| 70 |
+
choices: dict[str, dict[str, dict]] = {dataset_key: {} for dataset_key in dataset_keys}
|
| 71 |
root = Path(root_dir)
|
| 72 |
if not root.exists():
|
| 73 |
+
return choices
|
| 74 |
|
| 75 |
for model_dir in sorted(root.iterdir()):
|
| 76 |
if not model_dir.is_dir():
|
| 77 |
continue
|
| 78 |
+
model_label = _model_label(root, model_dir.name)
|
| 79 |
for snapshot_dataset, snapshot in load_forecast_snapshots(root_dir, model_dir.name).items():
|
| 80 |
+
dataset_key = str(snapshot_dataset).split("/")[0]
|
| 81 |
+
if dataset_key not in choices:
|
| 82 |
continue
|
| 83 |
+
choices[dataset_key][model_label] = snapshot
|
| 84 |
+
|
| 85 |
+
return choices
|
| 86 |
+
|
| 87 |
+
|
| 88 |
+
def _available_forecast_models(dataset_key: str) -> list[str]:
|
| 89 |
+
return sorted(FORECAST_CHOICES.get(dataset_key, {}).keys())
|
| 90 |
+
|
| 91 |
|
| 92 |
+
def _forecast_plot_for(dataset_key: str, model_label: str):
|
| 93 |
+
if not model_label:
|
| 94 |
+
return None
|
| 95 |
+
return make_forecast_plot(FORECAST_CHOICES.get(dataset_key, {}).get(model_label))
|
| 96 |
+
|
| 97 |
+
|
| 98 |
+
def _make_forecast_plotter(dataset_key: str):
|
| 99 |
+
def update_forecast_plot(model_label: str):
|
| 100 |
+
return _forecast_plot_for(dataset_key, model_label)
|
| 101 |
+
|
| 102 |
+
return update_forecast_plot
|
| 103 |
|
| 104 |
|
| 105 |
def refresh_leaderboard() -> tuple:
|
|
|
|
| 124 |
load_live_aggregate_table(RESULTS_PATH, "rank"), # [13]
|
| 125 |
]
|
| 126 |
|
| 127 |
+
# [14 .. 14+D-1] per-domain status lines
|
| 128 |
for domain, domain_datasets in DOMAIN_GROUPS.items():
|
| 129 |
outputs.append(build_domain_status_md(domain, domain_datasets, RESULTS_PATH))
|
| 130 |
|
| 131 |
value_tables, rank_tables = _dataset_tables(grouped, datasets)
|
|
|
|
| 132 |
# [14+D .. 14+D+N-1] value tables
|
| 133 |
outputs.extend(value_tables)
|
| 134 |
# [14+D+N .. 14+D+2N-1] rank tables
|
| 135 |
outputs.extend(rank_tables)
|
|
|
|
|
|
|
| 136 |
return tuple(outputs)
|
| 137 |
|
| 138 |
|
| 139 |
grouped_initial = get_grouped_dfs(RESULTS_PATH)
|
| 140 |
DATASETS: list[str] = grouped_initial["datasets"]
|
| 141 |
+
FORECAST_CHOICES = _forecast_choices(RESULTS_PATH, DATASETS)
|
| 142 |
DOMAIN_GROUPS = group_datasets_by_domain(DATASETS, RESULTS_PATH)
|
| 143 |
N_DOMAINS = len(DOMAIN_GROUPS)
|
| 144 |
INITIAL_OUTPUTS = refresh_leaderboard()
|
|
|
|
| 151 |
|
| 152 |
dataset_value_dfs: list[gr.Dataframe] = []
|
| 153 |
dataset_rank_dfs: list[gr.Dataframe] = []
|
|
|
|
| 154 |
domain_status_mds: list[gr.Markdown] = []
|
| 155 |
dataset_idx = 0
|
| 156 |
|
|
|
|
| 306 |
show_label=False,
|
| 307 |
)
|
| 308 |
)
|
| 309 |
+
dataset_key = dataset.split("/")[0]
|
| 310 |
+
forecast_model_choices = _available_forecast_models(dataset_key)
|
| 311 |
+
default_forecast_model = forecast_model_choices[0] if forecast_model_choices else None
|
| 312 |
+
forecast_model_dropdown = gr.Dropdown(
|
| 313 |
+
choices=forecast_model_choices,
|
| 314 |
+
value=default_forecast_model,
|
| 315 |
+
label="Model",
|
| 316 |
+
interactive=bool(forecast_model_choices),
|
| 317 |
+
)
|
| 318 |
+
forecast_plot = gr.Plot(
|
| 319 |
+
value=_forecast_plot_for(dataset_key, default_forecast_model),
|
| 320 |
+
label="Forecast snapshot",
|
| 321 |
+
)
|
| 322 |
+
forecast_model_dropdown.change(
|
| 323 |
+
fn=_make_forecast_plotter(dataset_key),
|
| 324 |
+
inputs=[forecast_model_dropdown],
|
| 325 |
+
outputs=[forecast_plot],
|
| 326 |
)
|
| 327 |
|
| 328 |
dataset_idx += 1
|
|
|
|
| 356 |
*domain_status_mds,
|
| 357 |
*dataset_value_dfs,
|
| 358 |
*dataset_rank_dfs,
|
|
|
|
| 359 |
]
|
| 360 |
|
| 361 |
demo.load(refresh_leaderboard, inputs=[], outputs=refresh_outputs)
|