Create load_to_hub.py

src/load_to_hub.py

@@ -0,0 +1,56 @@
+import json
+import math
+
+import pandas as pd
+
+from src.assets.symbols import UP_ARROW, DOWN_ARROW
+
+from src.tasks import TASKS, TASK_CODES, TASK_TO_METRIC
+
+
+def load_from_hub(fs, repo_path):
+    files = fs.glob(f"{repo_path}/**/*.json")
+
+    set_organization_models = {}
+    tasks = {}
+
+    for file in files:
+        organization, model, task = file.split("/")[-3:]
+
+        organization_model = f"{organization}/{model}"
+        task_code = task.replace(".json", "")
+
+        if task_code not in TASK_CODES:
+            continue
+
+        set_organization_models[organization_model] = 1
+        tasks[task_code] = 1
+
+    table = pd.DataFrame(
+        index=list(set_organization_models.keys()),
+        columns=["Organization", "Model"] + list(tasks.keys()),
+        data=None,
+    )
+
+    for file in files:
+        organization, model, task = file.split("/")[-3:]
+
+        organization_model = f"{organization}/{model}"
+        task_code = task.replace(".json", "")
+
+        if task_code not in TASK_CODES:
+            continue
+
+        data = json.loads(fs.open(file, "r").read())
+
+        result = round(data["results"][task_code][TASK_TO_METRIC[task_code]], 4)
+
+        table.loc[organization_model, task_code] = result
+        table.loc[organization_model, "Organization"] = organization
+        table.loc[organization_model, "Model"] = model
+
+    table.rename(columns={
+        task.code: f"{task.name} {UP_ARROW if task.higher_is_better else DOWN_ARROW}"
+        for task in TASKS}, inplace=True)
+
+    return table