Fit new hf Dataset

src/hf_config.py

@@ -16,7 +16,8 @@ from huggingface_hub import snapshot_download
 # =============================================================================
 
 # HuggingFace Dataset repository ID
-HF_REPO_ID = os.environ.get("HF_REPO_ID", "TIME-benchmark/TIME-1.0")
+HF_REPO_ID = os.environ.get("HF_REPO_ID", "Real-TSF/TIME")
+HF_OUTPUT_REPO_ID = os.environ.get("HF_OUTPUT_REPO_ID", "Real-TSF/TIME-Output")
 
 # HuggingFace token (set via environment variable for security)
 # In HuggingFace Space, set this in Settings -> Repository secrets
@@ -51,28 +52,26 @@ def download_results_snapshot() -> Path:
     """
     if not USE_HF_HUB:
         # Return local path for development
-        # Priority: 1) LOCAL_RESULTS_PATH env var, 2) ../output/results, 3) /home/eee/qzz/TIME/output/results
+        # Priority: 1) LOCAL_RESULTS_PATH env var
         if LOCAL_RESULTS_PATH:
             local_path = Path(LOCAL_RESULTS_PATH)
         else:
             local_path = Path("../output/results")
-            if not local_path.exists():
-                local_path = Path("/home/eee/qzz/TIME/output/results")
         if not local_path.exists():
             print(f"⚠️ Warning: Local results path does not exist: {local_path}")
         return local_path
 
-    print(f"📥 Downloading results from HuggingFace Hub: {HF_REPO_ID}")
+    print(f"📥 Downloading results from HuggingFace Hub: {HF_OUTPUT_REPO_ID}")
 
     local_dir = snapshot_download(
-        repo_id=HF_REPO_ID,
+        repo_id=HF_OUTPUT_REPO_ID,
         repo_type="dataset",
         token=HF_TOKEN,
-        allow_patterns=["output/results/**"],
+        allow_patterns=["results/**"],
         cache_dir=HF_CACHE_DIR,
     )
 
-    results_path = Path(local_dir) / "output" / "results"
+    results_path = Path(local_dir) / "results"
     print(f"✅ Results cached at: {results_path}")
     return results_path
 
@@ -88,13 +87,11 @@ def download_datasets_snapshot() -> Path:
     """
     if not USE_HF_HUB:
         # Return local path for development
-        # Priority: 1) LOCAL_DATASETS_PATH env var, 2) ../data/hf_dataset, 3) /home/eee/qzz/TIME/data/hf_dataset
+        # Priority: 1) LOCAL_DATASETS_PATH env var, 2) ../data/hf_dataset
         if LOCAL_DATASETS_PATH:
             local_path = Path(LOCAL_DATASETS_PATH)
         else:
             local_path = Path("../data/hf_dataset")
-            if not local_path.exists():
-                local_path = Path("/home/eee/qzz/TIME/data/hf_dataset")
         if not local_path.exists():
             print(f"⚠️ Warning: Local datasets path does not exist: {local_path}")
         return local_path
@@ -105,11 +102,11 @@ def download_datasets_snapshot() -> Path:
         repo_id=HF_REPO_ID,
         repo_type="dataset",
         token=HF_TOKEN,
-        allow_patterns=["data/hf_dataset/**"],
+        allow_patterns=None,
         cache_dir=HF_CACHE_DIR,
     )
 
-    datasets_path = Path(local_dir) / "data" / "hf_dataset"
+    datasets_path = Path(local_dir)
     print(f"✅ Datasets cached at: {datasets_path}")
     return datasets_path
 
@@ -136,13 +133,9 @@ def download_config_snapshot() -> Path:
         pass
 
     # Fallback: Local development path
-    # Priority: 1) LOCAL_CONFIG_PATH env var, 2) ../config, 3) /home/eee/qzz/TIME/config
+    # Priority: 1) LOCAL_CONFIG_PATH env var, 2) ../config,
     if LOCAL_CONFIG_PATH:
         local_path = Path(LOCAL_CONFIG_PATH)
-    else:
-        local_path = Path("../config")
-        if not local_path.exists():
-            local_path = Path("/home/eee/qzz/TIME/config")
 
     if local_path.exists():
         print(f"📁 Using local config: {local_path}")
@@ -193,17 +186,17 @@ def get_features_root() -> Path:
         return local_path
 
     # For HF Hub, features are in the same repo as results
-    print(f"📥 Downloading features from HuggingFace Hub: {HF_REPO_ID}")
+    print(f"📥 Downloading features from HuggingFace Hub: {HF_OUTPUT_REPO_ID}")
 
     local_dir = snapshot_download(
-        repo_id=HF_REPO_ID,
+        repo_id=HF_OUTPUT_REPO_ID,
         repo_type="dataset",
         token=HF_TOKEN,
-        allow_patterns=["output/features/**"],
+        allow_patterns=["features/**"],
         cache_dir=HF_CACHE_DIR,
     )
 
-    features_path = Path(local_dir) / "output" / "features"
+    features_path = Path(local_dir) / "features"
     print(f"✅ Features cached at: {features_path}")
     return features_path
 

src/tab.py

@@ -221,6 +221,26 @@ def update_horizon_choices(display_name):
     return gr.Radio(choices=choices, value=current_value)
 
 
+def update_horizon_checkbox_choices(display_name):
+    """
+    根据数据集更新horizon CheckboxGroup组件的choices和value
+    用于 Per Dataset tab
+
+    Args:
+        display_name: Dataset display name from UI dropdown
+
+    Returns:
+        gr.CheckboxGroup: 更新后的CheckboxGroup组件
+    """
+    available_horizons = get_available_horizons(display_name)
+
+    # 创建choices列表，只包含可用的horizons
+    choices = [h for h in ALL_HORIZONS if h in available_horizons]
+
+    # 默认全部选中
+    return gr.CheckboxGroup(choices=choices, value=choices)
+
+
 def update_series_variate_and_window(display_name, horizon):
     """
     根据 dataset display_name 和 horizon 更新 series, variate, window 的下拉选项
@@ -697,7 +717,7 @@ def plot_window_series(display_name, series, variate, window_id, horizon, select
         yaxis=dict(showgrid=True, gridcolor='lightgray', gridwidth=1)
     )
 
-    
+
     # Create info message for prediction window
     if timestamps is not None and test_window_start_idx is not None and test_window_end_idx is not None:
         pred_start_ts = timestamps[test_window_start_idx]
@@ -889,23 +909,26 @@ def init_per_dataset_tab(demo):
         elem_classes="markdown-text"
     )
 
+    # Initialize horizon choices based on first dataset
+    initial_dataset = DATASET_CHOICES[0]
+    initial_horizons = get_available_horizons(initial_dataset)
+
     with gr.Row():
         with gr.Column(scale=1):
             horizons = gr.CheckboxGroup(
-                choices=ALL_HORIZONS,
-                value=ALL_HORIZONS,
+                choices=initial_horizons,
+                value=initial_horizons,
                 label="Horizons"
             )
 
             dataset_dropdown = gr.Dropdown(
                 choices=DATASET_CHOICES,
-                value=DATASET_CHOICES[0],
+                value=initial_dataset,
                 label="Dataset",
                 interactive=True
             )
 
             # Initialize series and variate dropdowns
-            initial_dataset = DATASET_CHOICES[0]
             series_dropdown, variate_dropdown = update_series_and_variate(
                 initial_dataset
             )
@@ -913,15 +936,23 @@ def init_per_dataset_tab(demo):
     msg = gr.Textbox(label="Message", interactive=False)
     table = gr.DataFrame(elem_classes="custom-table", interactive=False)
 
-    # Update series and variate dropdowns when dataset changes
+    # Update horizons, series, and variate dropdowns when dataset changes
     dataset_dropdown.change(
+        fn=update_horizon_checkbox_choices,
+        inputs=[dataset_dropdown],
+        outputs=[horizons],
+    ).then(
         fn=update_series_and_variate,
         inputs=[dataset_dropdown],
         outputs=[series_dropdown, variate_dropdown],
+    ).then(
+        fn=get_dataset_multilevel_leaderboard,
+        inputs=[dataset_dropdown, series_dropdown, variate_dropdown, horizons],
+        outputs=[msg, table]
     )
 
-    # Update leaderboard when any selection changes
-    for comp in [dataset_dropdown, series_dropdown, variate_dropdown, horizons]:
+    # Update leaderboard when series, variate, or horizons change
+    for comp in [series_dropdown, variate_dropdown, horizons]:
         comp.change(
             fn=get_dataset_multilevel_leaderboard,
             inputs=[dataset_dropdown, series_dropdown, variate_dropdown, horizons],