Deploy from GitHub

streamlit_simulation/app.py

@@ -1,3 +1,4 @@
+import os
 import time
 import warnings
 
@@ -8,6 +9,7 @@ import pandas as pd
 import streamlit as st
 import torch
 from config_streamlit import DATA_PATH, PLOT_COLOR, TRAIN_RATIO
+from huggingface_hub import hf_hub_download
 
 from lightgbm_model.scripts.config_lightgbm import FEATURES
 from lightgbm_model.scripts.model_loader_wrapper import load_lightgbm_model
@@ -84,6 +86,21 @@ init_session_state()
 
 
 # ============================== Loaders Cache ==============================
+HF_REPO = "dlaj/energy-forecasting-files"
+HF_FILENAME = "data/processed/energy_consumption_aggregated_cleaned.csv"
+
+# if local data, use them, if not, download from huggingface
+if os.path.exists(DATA_PATH):
+    CSV_PATH = DATA_PATH
+else:
+    CSV_PATH = hf_hub_download(
+        repo_id=HF_REPO,
+        filename=HF_FILENAME,
+        repo_type="dataset",
+        cache_dir="hf_cache",  # Optional
+    )
+
+
 @st.cache_data
 def load_cached_lightgbm_model():
     return load_lightgbm_model()
@@ -449,7 +466,7 @@ if model_choice == "Transformer Model (moments)":
             len(InformerDataset(data_split="train", forecast_horizon=FORECAST_HORIZON))
             + SEQ_LEN
         )
-        base_timestamp = pd.read_csv(DATA_PATH, parse_dates=["date"])["date"].iloc[
+        base_timestamp = pd.read_csv(CSV_PATH, parse_dates=["date"])["date"].iloc[
             test_start_idx
         ]  # get original timestamp for later, cause not in dataset anymore
 

streamlit_simulation/utils_streamlit.py

@@ -27,3 +27,17 @@ def load_data():
 
     print(f"Lade lokale Datei: {DATA_PATH}")
     return pd.read_csv(DATA_PATH, parse_dates=["date"])
+
+
+def resolve_csv_path() -> str:
+    if os.path.exists(DATA_PATH):
+        print(f"Lokale Datei verwendet: {DATA_PATH}")
+        return DATA_PATH
+    else:
+        print(f"Lokale Datei nicht gefunden, lade von HF: {HF_FILENAME}")
+        return hf_hub_download(
+            repo_id=HF_REPO,
+            filename=HF_FILENAME,
+            repo_type="dataset",
+            cache_dir="hf_cache",
+        )

transformer_model/scripts/utils/informer_dataset_class.py

@@ -12,6 +12,21 @@ from transformer_model.scripts.config_transformer import DATA_PATH, SEQ_LEN
 
 logging.basicConfig(level=logging.INFO)
 
+HF_REPO = "dlaj/energy-forecasting-files"
+HF_FILENAME = "data/processed/energy_consumption_aggregated_cleaned.csv"
+
+if os.path.exists(DATA_PATH):
+    print(f"Lokale Datei gefunden: {DATA_PATH}")
+    CSV_PATH = DATA_PATH
+else:
+    print("Lokale Datei NICHT gefunden! Lade von Hugging Face...")
+    CSV_PATH = hf_hub_download(
+        repo_id=HF_REPO,
+        filename=HF_FILENAME,
+        repo_type="dataset",
+        cache_dir="hf_cache",  # Optional
+    )
+
 
 class InformerDataset:
     def __init__(
@@ -39,28 +54,12 @@ class InformerDataset:
 
         self.seq_len = SEQ_LEN
         self.forecast_horizon = forecast_horizon
+        self.full_file_path_and_name = CSV_PATH
         self.data_split = data_split
         self.data_stride_len = data_stride_len
         self.task_name = task_name
         self.random_seed = random_seed
 
-        # use local dataset if available, else download it from huggingface
-        HF_REPO = "dlaj/energy-forecasting-files"
-        HF_FILENAME = "data/processed/energy_consumption_aggregated_cleaned.csv"
-
-        if not os.path.exists(DATA_PATH):
-            print(f"Lokale Datei nicht gefunden: {DATA_PATH}")
-            print("Lade von Hugging Face Hub...")
-
-            self.full_file_path_and_name = hf_hub_download(
-                repo_id=HF_REPO,
-                filename=HF_FILENAME,
-                repo_type="dataset",
-                cache_dir="hf_cache",  # optional
-            )
-        else:
-            self.full_file_path_and_name = DATA_PATH
-
         self._read_data()
 
     def _get_borders(self):