Upload folder using huggingface_hub

README.md

@@ -1,62 +1,10 @@
 ---
-license: mit
-tags:
-- futures-prediction
-- multi-dimensional
-- mixture-of-experts
-- state-space-model
+title: Interactive Futures Model
+emoji: 🚀
+colorFrom: blue
+colorTo: green
+sdk: gradio
+sdk_version: "6.2.0"
+app_file: app.py
+pinned: false
 ---
-
-# Futures Prediction Model (MoE + SSM + FiLM)
-
-This repository contains the code and trained weights for a novel architecture designed for multi-dimensional futures prediction. The model was trained on the `futures_dataset_v2.json` dataset.
-
-## Model Description
-
-The model architecture is a combination of:
-
-*   **Mixture of Experts (MoE):** To handle the multi-dimensional nature of futures scenarios.
-*   **State Space Model (SSM):** To capture the temporal evolution of futures.
-*   **FiLM Conditioning:** To modulate the model's behavior based on the different future axes.
-
-The model is trained to predict a 12-dimensional vector of weights, each corresponding to a different future "axis".
-
-## How to Use
-
-To use this model, you will need to have PyTorch installed. You can then use the `load_model.py` script to load the model and tokenizer.
-
-```python
-from load_model import load_model_and_tokenizer
-
-model, tokenizer = load_model_and_tokenizer()
-
-text = "In a future dominated by hyper-automation, societal structures adapt to new forms of labor and community."
-token_ids = tokenizer.encode(text)
-tokens_tensor = torch.LongTensor(token_ids).unsqueeze(0)
-
-with torch.no_grad():
-    axis_logits, _, _ = model(tokens_tensor)
-    axis_predictions = torch.sigmoid(axis_logits)
-
-print(axis_predictions)
-```
-
-## Training Data
-
-The model was trained on the `futures_dataset_v2.json` dataset, which contains 3,000 rich, multi-dimensional futures scenarios.
-
-## Training Procedure
-
-The model was trained for 100 epochs with a batch size of 16 and a learning rate of 1e-4. The training script `train_futures_model.py` is available in the original repository.
-
-## Citing
-
-If you use this model or code, please cite:
-
-```
-@article{futures-representation-learning,
-  title={Learning Multi-Dimensional Futures Representations with Mixture-of-Experts and State Space Models},
-  author={Your Name},
-  year={2024}
-}
-```

app.py

@@ -0,0 +1,87 @@
+import gradio as gr
+import torch
+from app.load_model import load_model_and_tokenizer
+
+# --- 1. Load Model and Tokenizer ---
+# This is done once when the Gradio app starts.
+try:
+    print("Loading model and tokenizer...")
+    model, tokenizer = load_model_and_tokenizer()
+    print("✅ Model and tokenizer loaded successfully.")
+except Exception as e:
+    print(f"❌ Failed to load model: {e}")
+    model, tokenizer = None, None
+
+# --- 2. Define the Prediction Function ---
+# This function is called every time a user interacts with the demo.
+def predict_futures(text):
+    """
+    Takes raw text input, tokenizes it, gets model predictions,
+    and formats the output for the Gradio interface.
+    """
+    if not model or not tokenizer:
+        return "Model not loaded. Please check the logs.", {}
+
+    try:
+        # a. Preprocess: Tokenize the input text
+        token_ids = tokenizer.encode(text)
+        tokens_tensor = torch.LongTensor(token_ids).unsqueeze(0) # Add batch dimension
+
+        # b. Predict: Get model's raw output (logits)
+        with torch.no_grad():
+            axis_logits, _, _ = model(tokens_tensor)
+            # c. Post-process: Apply sigmoid to get probabilities (0-1)
+            axis_predictions = torch.sigmoid(axis_logits)
+
+        # d. Format Output: Create a dictionary for the label component
+        axis_names = [
+            "Hyper-Automation", "Human-Tech Symbiosis", "Abundance", "Individualism",
+            "Community Focus", "Global Interconnectedness", "Crisis & Collapse", "Restoration & Healing",
+            "Adaptation & Resilience", "Digital Dominance", "Physical Embodiment", "Collaboration"
+        ]
+        
+        # Create a dictionary of {label: confidence}
+        confidences = {name: float(weight) for name, weight in zip(axis_names, axis_predictions[0])}
+        
+        # You can return a simple message and the formatted labels
+        return "Prediction complete.", confidences
+
+    except Exception as e:
+        print(f"Error during prediction: {e}")
+        return f"An error occurred: {e}", {}
+
+# --- 3. Create and Launch the Gradio Interface ---
+print("Creating Gradio interface...")
+
+# Define the input and output components
+input_text = gr.Textbox(
+    lines=5,
+    label="Input Scenario",
+    placeholder="Describe a future scenario here..."
+)
+
+output_text = gr.Textbox(label="Status")
+output_labels = gr.Label(label="Predicted Axis Weights", num_top_classes=12)
+
+# Build the interface
+demo = gr.Interface(
+    fn=predict_futures,
+    inputs=input_text,
+    outputs=[output_text, output_labels],
+    title="Futures Prediction Model",
+    description=(
+        "Explore multi-dimensional futures. "
+        "Write a text describing a potential future scenario and see how the model scores it "
+        "across 12 different axes, from 'Hyper-Automation' to 'Crisis & Collapse'."
+    ),
+    examples=[
+        ["In a future dominated by hyper-automation, societal structures adapt to new forms of labor and community."],
+        ["Coastal cities adopt divergent strategies as sea levels rise. Singapore invests in autonomous seawall monitoring, while Jakarta facilitates managed retreat."],
+        ["A global pandemic leads to a surge in community-focused initiatives and a renewed appreciation for local supply chains."]
+    ]
+)
+
+if __name__ == "__main__":
+    print("Launching Gradio demo...")
+    # The launch() command creates a shareable link to the demo.
+    demo.launch()