Create app.py

app.py

@@ -0,0 +1,93 @@
+# A simple script that loops over all public models and get their libary_name
+import gradio as gr
+import pandas as pd
+import numpy as np
+
+from collections import Counter
+
+from huggingface_hub import HfApi
+from datasets import load_dataset
+
+api = HfApi()
+list_models = api.list_models()
+
+def fetch_dataset_and_init():
+    dataset = load_dataset("librarian-bots/model_cards_with_metadata", split="train")
+    library_names = dataset["library_name"]
+
+    string_counts = Counter(library_names)
+    string_counts_series = pd.Series(string_counts)
+
+    # Sort the series in descending order
+    df = string_counts_series.sort_values(ascending=False).to_frame()
+    df.columns = ["count"]
+    df = df.reset_index()
+    df_log = df.copy()
+    df_log['count'] = np.log(df_log['count'])
+
+    return df, df_log
+
+df, df_log = fetch_dataset_and_init()
+
+def get_current_nb_models():
+    # We need this hack since `list_models` returns a generator..
+    total_models = sum(1 for _ in list_models)
+    diff_models = total_models - len(df)
+    return str(diff_models)
+
+plot_height = 512
+plot_width = 1512
+top_k = len(df)
+
+def bar_plot_fn(display, top_k):
+    if display == "simple":
+        return gr.BarPlot(
+            df[:top_k],
+            x="index",
+            y="count",
+            tooltip=["index", "count"],
+            height=plot_height,
+            width=plot_width
+        )
+    elif display == "log":
+        return gr.BarPlot(
+            df_log[:top_k],
+            x="index",
+            y="count",
+            tooltip=["index", "count"],
+            height=plot_height,
+            width=plot_width
+        )
+
+
+with gr.Blocks() as bar_plot:
+    with gr.Column():
+        with gr.Column():
+            display = gr.Dropdown(
+                choices=[
+                    "simple",
+                    "log",
+                ],
+                value="simple",
+                label="Type of Bar Plot",
+            )
+            top_k = gr.Slider(
+                label="Select top-K most used library_name",
+                value=len(df),
+                minimum=1,
+                maximum=len(df),
+                step=1,
+            )
+        with gr.Column():
+            plot = gr.BarPlot()
+
+        with gr.Row():
+            fetch_button = gr.Button(value="Fetch number of models without model cards (takes up to 1min to fetch everything)")
+            text_box = gr.Textbox(value="", label="Number of models without model cards")
+
+    top_k.change(bar_plot_fn, inputs=[display, top_k], outputs=plot)
+    display.change(bar_plot_fn, inputs=[display, top_k], outputs=plot)
+    fetch_button.click(get_current_nb_models, outputs=[text_box])
+    bar_plot.load(fn=bar_plot_fn, inputs=[display, top_k], outputs=plot)
+
+bar_plot.launch()