update

app.py

@@ -4,6 +4,7 @@ import gradio as gr
 import pandas as pd
 
 from apscheduler.schedulers.background import BackgroundScheduler
+
 from huggingface_hub import snapshot_download
 
 from src.display.about import (
@@ -14,8 +15,9 @@ from src.display.about import (
     LLM_BENCHMARKS_TEXT,
     LLM_BENCHMARKS_DETAILS,
     FAQ_TEXT,
-    TITLE,
+    TITLE
 )
+
 from src.display.css_html_js import custom_css
 
 from src.display.utils import (
@@ -38,33 +40,44 @@ from src.submission.submit import add_new_eval
 from src.utils import get_dataset_summary_table
 
 
-def restart_space():
-    API.restart_space(repo_id=REPO_ID, token=H4_TOKEN)
-
-
 def ui_snapshot_download(repo_id, local_dir, repo_type, tqdm_class, etag_timeout):
     try:
         print(local_dir)
         snapshot_download(repo_id=repo_id, local_dir=local_dir, repo_type=repo_type, tqdm_class=tqdm_class, etag_timeout=etag_timeout)
-    except Exception:
+    except Exception as e:
         restart_space()
 
-dataset_df = get_dataset_summary_table(file_path='blog/Hallucination-Leaderboard-Summary.csv')
 
-ui_snapshot_download(repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)
-ui_snapshot_download(repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)
+def restart_space():
+    API.restart_space(repo_id=REPO_ID, token=H4_TOKEN)
 
 
-raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
-leaderboard_df = original_df.copy()
+def init_space():
+    dataset_df = get_dataset_summary_table(file_path='blog/Hallucination-Leaderboard-Summary.csv')
+
+    import socket
+    if socket.gethostname() not in {'neuromancer'}:
+        ui_snapshot_download(repo_id=QUEUE_REPO, local_dir=EVAL_REQUESTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)
+        ui_snapshot_download(repo_id=RESULTS_REPO, local_dir=EVAL_RESULTS_PATH, repo_type="dataset", tqdm_class=None, etag_timeout=30)
 
-finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)
+    raw_data, original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, COLS, BENCHMARK_COLS)
+
+    finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = get_evaluation_queue_df(EVAL_REQUESTS_PATH, EVAL_COLS)
+    return dataset_df, original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df
+
+
+dataset_df, original_df, finished_eval_queue_df, running_eval_queue_df, pending_eval_queue_df = init_space()
+leaderboard_df = original_df.copy()
 
 
 # Searching and filtering
-def update_table(hidden_df: pd.DataFrame, columns: list, type_query: list, precision_query: list, size_query: list, query: str):
-    show_deleted = True
-    filtered_df = filter_models(hidden_df, type_query, size_query, precision_query, show_deleted)
+def update_table(hidden_df: pd.DataFrame,
+                 columns: list,
+                 type_query: list,
+                 precision_query: list,
+                 size_query: list,
+                 query: str):
+    filtered_df = filter_models(hidden_df, type_query, size_query, precision_query)
     filtered_df = filter_queries(query, filtered_df)
     df = select_columns(filtered_df, columns)
     return df
@@ -75,13 +88,15 @@ def search_table(df: pd.DataFrame, query: str) -> pd.DataFrame:
 
 
 def select_columns(df: pd.DataFrame, columns: list) -> pd.DataFrame:
-    always_here_cols = [
-        AutoEvalColumn.model_type_symbol.name,
-        AutoEvalColumn.model.name,
-    ]
+    # always_here_cols = [AutoEvalColumn.model_type_symbol.name, AutoEvalColumn.model.name]
+
+    always_here_cols = [c.name for c in fields(AutoEvalColumn) if c.never_hidden]
+    dummy_col = [AutoEvalColumn.dummy.name]
+
     # We use COLS to maintain sorting
     filtered_df = df[
-        always_here_cols + [c for c in COLS if c in df.columns and c in columns] + [AutoEvalColumn.dummy.name]
+        # always_here_cols + [c for c in COLS if c in df.columns and c in columns] + [AutoEvalColumn.dummy.name]
+        always_here_cols + [c for c in COLS if c in df.columns and c in columns] + dummy_col
     ]
     return filtered_df
 
@@ -98,19 +113,17 @@ def filter_queries(query: str, filtered_df: pd.DataFrame):
                     final_df.append(temp_filtered_df)
         if len(final_df) > 0:
             filtered_df = pd.concat(final_df)
-            filtered_df = filtered_df.drop_duplicates(
-                subset=[AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name]
-            )
-
+            subset = [AutoEvalColumn.model.name, AutoEvalColumn.precision.name, AutoEvalColumn.revision.name]
+            filtered_df = filtered_df.drop_duplicates(subset=subset)
     return filtered_df
 
 
-def filter_models(df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool) -> pd.DataFrame:
+def filter_models(df: pd.DataFrame,
+                  type_query: list,
+                  size_query: list,
+                  precision_query: list) -> pd.DataFrame:
     # Show all models
-    if show_deleted:
-        filtered_df = df
-    else:  # Show only still on the hub models
-        filtered_df = df[df[AutoEvalColumn.still_on_hub.name] is True]
+    filtered_df = df
 
     type_emoji = [t[0] for t in type_query]
     filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
@@ -124,17 +137,27 @@ def filter_models(df: pd.DataFrame, type_query: list, size_query: list, precisio
     return filtered_df
 
 
+# triggered only once at startup => read query parameter if it exists
+def load_query(request: gr.Request):
+    query = request.query_params.get("query") or ""
+    return query
+
+
 demo = gr.Blocks(css=custom_css)
 with demo:
     gr.HTML(TITLE)
     gr.Markdown(INTRODUCTION_TEXT, elem_classes="markdown-text")
 
     with gr.Tabs(elem_classes="tab-buttons") as tabs:
-        with gr.TabItem("Hallucination Benchmark", elem_id="llm-benchmark-tab-table", id=0):
+        with gr.TabItem("Hallucinations Benchmark",
+                        elem_id="llm-benchmark-tab-table",
+                        id=0):
             with gr.Row():
                 with gr.Column():
                     with gr.Row():
-                        search_bar = gr.Textbox(placeholder=" 🔍 Model search (separate multiple queries with `;`)", show_label=False, elem_id="search-bar",)
+                        search_bar = gr.Textbox(placeholder=" 🔍 Model search (separate multiple queries with `;`)",
+                                                show_label=False,
+                                                elem_id="search-bar")
                     with gr.Row():
                         shown_columns = gr.CheckboxGroup(
                             choices=[
@@ -149,8 +172,7 @@ with demo:
                             ],
                             label="Select columns to show",
                             elem_id="column-select",
-                            interactive=True,
-                        )
+                            interactive=True)
 
                 with gr.Column(min_width=320):
                     filter_columns_type = gr.CheckboxGroup(
@@ -158,44 +180,39 @@ with demo:
                         choices=[t.to_str() for t in ModelType],
                         value=[t.to_str() for t in ModelType],
                         interactive=True,
-                        elem_id="filter-columns-type",
-                    )
+                        elem_id="filter-columns-type")
+
                     filter_columns_precision = gr.CheckboxGroup(
                         label="Precision",
                         choices=[i.value.name for i in Precision],
                         value=[i.value.name for i in Precision],
                         interactive=True,
-                        elem_id="filter-columns-precision",
-                    )
+                        elem_id="filter-columns-precision")
+
                     filter_columns_size = gr.CheckboxGroup(
                         label="Model sizes (in billions of parameters)",
                         choices=list(NUMERIC_INTERVALS.keys()),
                         value=list(NUMERIC_INTERVALS.keys()),
                         interactive=True,
-                        elem_id="filter-columns-size",
-                    )
+                        elem_id="filter-columns-size")
 
             leaderboard_table = gr.components.Dataframe(
                 value=leaderboard_df[
-                    [c.name for c in fields(AutoEvalColumn) if c.never_hidden]
-                    + shown_columns.value
-                    + [AutoEvalColumn.dummy.name]
+                    [c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value + [AutoEvalColumn.dummy.name]
                 ] if leaderboard_df.empty is False else leaderboard_df,
                 headers=[c.name for c in fields(AutoEvalColumn) if c.never_hidden] + shown_columns.value,
                 datatype=TYPES,
                 elem_id="leaderboard-table",
                 interactive=False,
-                visible=True,
-                column_widths=["2%", "20%"]
-            )
+                visible=True)
 
             # Dummy leaderboard for handling the case when the user uses backspace key
             hidden_leaderboard_table_for_search = gr.components.Dataframe(
                 value=original_df[COLS] if original_df.empty is False else original_df,
                 headers=COLS,
                 datatype=TYPES,
-                visible=False
-            )
+                visible=False)
+
             search_bar.submit(
                 update_table,
                 [
@@ -206,8 +223,11 @@ with demo:
                     filter_columns_size,
                     search_bar,
                 ],
-                leaderboard_table,
-            )
+                leaderboard_table)
+
+            # Check query parameter once at startup and update search bar
+            demo.load(load_query, inputs=[], outputs=[search_bar])
+
             for selector in [shown_columns, filter_columns_type, filter_columns_precision, filter_columns_size]:
                 selector.change(
                     update_table,
@@ -220,8 +240,7 @@ with demo:
                         search_bar,
                     ],
                     leaderboard_table,
-                    queue=True,
-                )
+                    queue=True)
 
         with gr.TabItem("📝 About", elem_id="llm-benchmark-tab-table", id=2):
             gr.Markdown(LLM_BENCHMARKS_TEXT, elem_classes="markdown-text")
@@ -238,48 +257,38 @@ with demo:
             gr.Markdown(LLM_BENCHMARKS_DETAILS, elem_classes="markdown-text")
             gr.Markdown(FAQ_TEXT, elem_classes="markdown-text")
 
-        with gr.TabItem("🚀 Submit here! ", elem_id="llm-benchmark-tab-table", id=3):
+        with gr.TabItem("Submit a model ", elem_id="llm-benchmark-tab-table", id=3):
             with gr.Column():
                 with gr.Row():
                     gr.Markdown(EVALUATION_QUEUE_TEXT, elem_classes="markdown-text")
 
                 with gr.Column():
-                    with gr.Accordion(
-                        f"✅ Finished Evaluations ({len(finished_eval_queue_df)})",
-                        open=False,
-                    ):
+                    with gr.Accordion(f"✅ Finished Evaluations ({len(finished_eval_queue_df)})", open=False):
                         with gr.Row():
                             finished_eval_table = gr.components.Dataframe(
                                 value=finished_eval_queue_df,
                                 headers=EVAL_COLS,
                                 datatype=EVAL_TYPES,
-                                row_count=5
-                            )
-                    with gr.Accordion(
-                        f"🔄 Running Evaluation Queue ({len(running_eval_queue_df)})",
-                        open=False,
-                    ):
+                                row_count=5)
+
+                    with gr.Accordion(f"🔄 Running Evaluation Queue ({len(running_eval_queue_df)})", open=False):
                         with gr.Row():
                             running_eval_table = gr.components.Dataframe(
                                 value=running_eval_queue_df,
                                 headers=EVAL_COLS,
                                 datatype=EVAL_TYPES,
-                                row_count=5
-                            )
+                                row_count=5)
 
-                    with gr.Accordion(
-                        f"⏳ Pending Evaluation Queue ({len(pending_eval_queue_df)})",
-                        open=False,
-                    ):
+                    with gr.Accordion(f"⏳ Scheduled Evaluation Queue ({len(pending_eval_queue_df)})", open=False):
                         with gr.Row():
                             pending_eval_table = gr.components.Dataframe(
                                 value=pending_eval_queue_df,
                                 headers=EVAL_COLS,
                                 datatype=EVAL_TYPES,
-                                row_count=5
-                            )
+                                row_count=5)
+
             with gr.Row():
-                gr.Markdown("# ✉️✨ Submit your model here!", elem_classes="markdown-text")
+                gr.Markdown("# Submit your model here", elem_classes="markdown-text")
 
             with gr.Row():
                 with gr.Column():
@@ -291,8 +300,7 @@ with demo:
                         label="Model type",
                         multiselect=False,
                         value=None,
-                        interactive=True,
-                    )
+                        interactive=True)
 
                 with gr.Column():
                     precision = gr.Dropdown(
@@ -300,15 +308,15 @@ with demo:
                         label="Precision",
                         multiselect=False,
                         value="float32",
-                        interactive=True,
-                    )
+                        interactive=True)
+
                     weight_type = gr.Dropdown(
                         choices=[i.value.name for i in WeightType],
                         label="Weights type",
                         multiselect=False,
                         value="Original",
-                        interactive=True,
-                    )
+                        interactive=True)
+
                     base_model_name_textbox = gr.Textbox(label="Base model (for delta or adapter weights)")
 
             submit_button = gr.Button("Submit Eval")
@@ -324,18 +332,16 @@ with demo:
                     weight_type,
                     model_type,
                 ],
-                submission_result,
-            )
+                submission_result)
 
     with gr.Row():
-        with gr.Accordion("📙 Citation", open=False):
+        with gr.Accordion("Citing this leaderboard", open=False):
             citation_button = gr.Textbox(
                 value=CITATION_BUTTON_TEXT,
                 label=CITATION_BUTTON_LABEL,
                 lines=20,
                 elem_id="citation-button",
-                show_copy_button=True,
-            )
+                show_copy_button=True)
 
 scheduler = BackgroundScheduler()