debug_logic.py

import sys
import types
from unittest.mock import MagicMock

# 1. Mock PyTorch Class BEFORE importing data_processor
class MockTorch:
    def __init__(self):
        self.float32 = "float32"
        self.long = "long"
        
        # Mock Tensor class for type hints
        class MockTensor:
            pass
        self.Tensor = MockTensor
    
    def FloatTensor(self, x): return x # Return numpy array directly
    def LongTensor(self, x): return x
    def tensor(self, x): return x
    def randn(self, *args): 
        import numpy as np
        return np.random.randn(*args)
    def randint(self, low, high, size): 
        import numpy as np
        return np.random.randint(low, high, size)
    def stack(self, tensors, axis=0):
        import numpy as np
        return np.stack(tensors, axis=axis)

# 2. Inject into sys.modules
mock_torch = MockTorch()
sys.modules['torch'] = mock_torch

# 3. Now safe to import
import numpy as np
import pandas as pd
import os

# Add parent dir
sys.path.append(os.getcwd())

from data_processor import AlphaDataProcessor

def test_logic():
    print("✅ Successfully imported data_processor with Mock Torch")
    
    # 1. Simulate get_deeplob_tensors return
    N, T, Levels = 10, 100, 20
    # DataProcessor returns:
    # X = stack([p_window, v_window], axis=0) -> (2, T, 2*Levels) for each sample
    # Then np.array(X) -> (N, 2, T, 2*Levels)
    
    # Mock simulating what data_processor actually builds
    # (N samples, 2 channels, T timesteps, 2*Levels features)
    mock_X = np.random.randn(N, 2, T, 2*Levels)
    
    print(f"Mock DataProcessor Output Shape: {mock_X.shape}")
    
    # 2. Verify DeepLOB Input Requirement
    # DeepLOB Conv2d(2, 16, ...) expects (N, C, H, W) = (N, 2, T, Features)
    # Our features = 2*Levels = 40
    
    # If the shape is 4D: (N, 2, 100, 40) -> IT IS CORRECT
    # If we unsqueeze(1) -> (N, 1, 2, 100, 40) -> 5D -> INCORRECT
    
    if mock_X.ndim == 4 and mock_X.shape[1] == 2:
        print("✅ Data Shape matches Conv2d Expectation (N, 2, T, F)")
        print("   -> (N, Channels=2, Height=100, Width=40)")
        print("   -> NO unsqueeze(1) needed!")
    else:
        print(f"❌ Data Shape Mismatch: {mock_X.shape}")

    # 3. Verify TRM logic
    # TRM needs (N, T, F)
    # F = 6 (Vol, Imb, CVD, Spr, Mom, OFI)
    N_trm, T_trm, F_trm = 10, 60, 6
    mock_X_trm = np.random.randn(N_trm, T_trm, F_trm)
    
    print(f"\nMock TRM Output Shape: {mock_X_trm.shape}")
    if mock_X_trm.ndim == 3 and mock_X_trm.shape[2] == 6:
        print("✅ TRM Shape matches Transformer Expectation (N, T, F)")
    else:
        print(f"❌ TRM Shape Mismatch: {mock_X_trm.shape}")

if __name__ == "__main__":
    test_logic()