Create src.details module

app.py

@@ -1,185 +1,14 @@
-import json
-
 import gradio as gr
-import pandas as pd
 from huggingface_hub import HfFileSystem
 
-from src.constants import DETAILS_DATASET_ID, DETAILS_FILENAME, SUBTASKS, TASKS
+from src.constants import SUBTASKS, TASKS
+from src.details import update_subtasks_component, update_load_details_component, load_details_dataframes, \
+    display_details, update_sample_idx_component, clear_details
 from src.results import fetch_result_paths, filter_latest_result_path_per_model, update_load_results_component, \
     load_results_dataframes, display_results, update_tasks_component, clear_results
 
-fs = HfFileSystem()
-
-
-def fetch_result_paths():
-    paths = fs.glob(f"{RESULTS_DATASET_ID}/**/**/*.json")
-    return paths
-
-
-def filter_latest_result_path_per_model(paths):
-    from collections import defaultdict
-
-    d = defaultdict(list)
-    for path in paths:
-        model_id, _ = path[len(RESULTS_DATASET_ID) + 1:].rsplit("/", 1)
-        d[model_id].append(path)
-    return {model_id: max(paths) for model_id, paths in d.items()}
-
-
-def get_result_path_from_model(model_id, result_path_per_model):
-    return result_path_per_model[model_id]
-
-
-def update_load_results_component():
-    return gr.Button("Load Results", interactive=True)
-
-
-def load_data(result_path) -> pd.DataFrame:
-    with fs.open(result_path, "r") as f:
-        data = json.load(f)
-    return data
-
-
-def load_results_dataframe(model_id):
-    if not model_id:
-        return
-    result_path = get_result_path_from_model(model_id, latest_result_path_per_model)
-    data = load_data(result_path)
-    model_name = data.get("model_name", "Model")
-    df = pd.json_normalize([{key: value for key, value in data.items()}])
-    # df.columns = df.columns.str.split(".")  # .split return a list instead of a tuple
-    return df.set_index(pd.Index([model_name])).reset_index()
-
-
-def load_results_dataframes(*model_ids):
-    return [load_results_dataframe(model_id) for model_id in model_ids]
-
-
-def display_results(task, *dfs):
-    dfs = [df.set_index("index") for df in dfs if "index" in df.columns]
-    if not dfs:
-        return None, None
-    df = pd.concat(dfs)
-    df = df.T.rename_axis(columns=None)
-    return display_tab("results", df, task), display_tab("configs", df, task)
-
-
-def display_tab(tab, df, task):
-    df = df.style.format(na_rep="")
-    df.hide(
-        [
-            row
-            for row in df.index
-            if (
-                not row.startswith(f"{tab}.")
-                or row.startswith(f"{tab}.leaderboard.")
-                or row.endswith(".alias")
-                or (not row.startswith(f"{tab}.{task}") if task != "All" else False)
-            )
-        ],
-        axis="index",
-    )
-    start = len(f"{tab}.leaderboard_") if task == "All" else len(f"{tab}.{task} ")
-    df.format_index(lambda idx: idx[start:].removesuffix(",none"), axis="index")
-    return df.to_html()
-
 
-def update_tasks_component():
-    return gr.Radio(
-            ["All"] + list(TASKS.values()),
-            label="Tasks",
-            info="Evaluation tasks to be displayed",
-            value="All",
-            interactive=True,
-        )
-
-
-def clear_results():
-    # model_id_1, model_id_2, dataframe_1, dataframe_2, task
-    return (
-        None, None, None, None,
-        gr.Radio(
-            ["All"] + list(TASKS.values()),
-            label="Tasks",
-            info="Evaluation tasks to be displayed",
-            value="All",
-            interactive=False,
-        ),
-    )
-
-
-def update_subtasks_component(task):
-    return gr.Radio(
-        SUBTASKS.get(task),
-        info="Evaluation subtasks to be displayed",
-        value=None,
-    )
-
-
-def update_load_details_component(model_id_1, model_id_2, subtask):
-    if (model_id_1 or model_id_2) and subtask:
-        return gr.Button("Load Details", interactive=True)
-    else:
-        return gr.Button("Load Details", interactive=False)
-
-
-def load_details_dataframe(model_id, subtask):
-    if not model_id or not subtask:
-        return
-    model_name_sanitized = model_id.replace("/", "__")
-    paths = fs.glob(
-        f"{DETAILS_DATASET_ID}/**/{DETAILS_FILENAME}".format(
-            model_name_sanitized=model_name_sanitized, subtask=subtask
-        )
-    )
-    if not paths:
-        return
-    path = max(paths)
-    with fs.open(path, "r") as f:
-        data = [json.loads(line) for line in f]
-    df = pd.json_normalize(data)
-    # df = df.rename_axis("Parameters", axis="columns")
-    df["model_name"] = model_id  # Keep model_name
-    return df
-    # return df.set_index(pd.Index([model_id])).reset_index()
-
-
-def load_details_dataframes(subtask, *model_ids):
-    return [load_details_dataframe(model_id, subtask) for model_id in model_ids]
-
-
-def display_details(sample_idx, *dfs):
-    rows = [df.iloc[sample_idx] for df in dfs if "model_name" in df.columns and sample_idx < len(df)]
-    if not rows:
-        return
-    # Pop model_name and add it to the column name
-    df = pd.concat([row.rename(row.pop("model_name")) for row in rows], axis="columns")
-    return (
-        df.style
-        .format(na_rep="")
-        # .hide(axis="index")
-        .to_html()
-    )
-
-
-def update_sample_idx_component(*dfs):
-    maximum = max([len(df) - 1 for df in dfs])
-    return gr.Number(
-        label="Sample Index",
-        info="Index of the sample to be displayed",
-        value=0,
-        minimum=0,
-        maximum=maximum,
-        visible=True,
-    )
-
-
-def clear_details():
-    # model_id_1, model_id_2, details_dataframe_1, details_dataframe_2, details_task, subtask, sample_idx
-    return (
-        None, None, None, None, None, None,
-        gr.Number(label="Sample Index", info="Index of the sample to be displayed", value=0, minimum=0,visible=False),
-    )
+fs = HfFileSystem()
 
 
 # if __name__ == "__main__":

src/details.py

@@ -0,0 +1,81 @@
+import json
+
+import gradio as gr
+import pandas as pd
+
+from app import fs
+from src.constants import SUBTASKS, DETAILS_DATASET_ID, DETAILS_FILENAME
+
+
+def update_subtasks_component(task):
+    return gr.Radio(
+        SUBTASKS.get(task),
+        info="Evaluation subtasks to be displayed",
+        value=None,
+    )
+
+
+def update_load_details_component(model_id_1, model_id_2, subtask):
+    if (model_id_1 or model_id_2) and subtask:
+        return gr.Button("Load Details", interactive=True)
+    else:
+        return gr.Button("Load Details", interactive=False)
+
+
+def load_details_dataframe(model_id, subtask):
+    if not model_id or not subtask:
+        return
+    model_name_sanitized = model_id.replace("/", "__")
+    paths = fs.glob(
+        f"{DETAILS_DATASET_ID}/**/{DETAILS_FILENAME}".format(
+            model_name_sanitized=model_name_sanitized, subtask=subtask
+        )
+    )
+    if not paths:
+        return
+    path = max(paths)
+    with fs.open(path, "r") as f:
+        data = [json.loads(line) for line in f]
+    df = pd.json_normalize(data)
+    # df = df.rename_axis("Parameters", axis="columns")
+    df["model_name"] = model_id  # Keep model_name
+    return df
+    # return df.set_index(pd.Index([model_id])).reset_index()
+
+
+def load_details_dataframes(subtask, *model_ids):
+    return [load_details_dataframe(model_id, subtask) for model_id in model_ids]
+
+
+def display_details(sample_idx, *dfs):
+    rows = [df.iloc[sample_idx] for df in dfs if "model_name" in df.columns and sample_idx < len(df)]
+    if not rows:
+        return
+    # Pop model_name and add it to the column name
+    df = pd.concat([row.rename(row.pop("model_name")) for row in rows], axis="columns")
+    return (
+        df.style
+        .format(na_rep="")
+        # .hide(axis="index")
+        .to_html()
+    )
+
+
+def update_sample_idx_component(*dfs):
+    maximum = max([len(df) - 1 for df in dfs])
+    return gr.Number(
+        label="Sample Index",
+        info="Index of the sample to be displayed",
+        value=0,
+        minimum=0,
+        maximum=maximum,
+        visible=True,
+    )
+
+
+def clear_details():
+    # model_id_1, model_id_2, details_dataframe_1, details_dataframe_2, details_task, subtask, sample_idx
+    return (
+        None, None, None, None, None, None,
+        gr.Number(label="Sample Index", info="Index of the sample to be displayed", value=0, minimum=0,visible=False),
+    )