mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-05-01 13:42:33 -05:00
9361cbf987
- Create professional examples directory showcasing TinyTorch as real ML framework - Add examples: XOR, MNIST, CIFAR-10, text generation, autograd demo, optimizer comparison - Fix import paths in exported modules (training.py, dense.py) - Update training module with autograd integration for loss functions - Add progressive integration tests for all 16 modules - Document framework capabilities and usage patterns This commit establishes the examples gallery that demonstrates TinyTorch works like PyTorch/TensorFlow, validating the complete framework.
143 lines
4.8 KiB
Python
143 lines
4.8 KiB
Python
"""
|
|
Module 02: Tensor - Integration Tests
|
|
Tests that Tensor works as foundation for all other modules
|
|
"""
|
|
|
|
import numpy as np
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
# Add project root to path
|
|
sys.path.insert(0, str(Path(__file__).parent.parent.parent))
|
|
|
|
|
|
class TestTensorFoundation:
|
|
"""Test Tensor as foundation for the framework."""
|
|
|
|
def test_tensor_import(self):
|
|
"""Test Tensor can be imported from package."""
|
|
from tinytorch.core.tensor import Tensor
|
|
assert Tensor is not None
|
|
|
|
def test_tensor_creation(self):
|
|
"""Test various ways to create tensors."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
# From list
|
|
t1 = Tensor([1, 2, 3])
|
|
assert t1.shape == (3,)
|
|
|
|
# From numpy array
|
|
t2 = Tensor(np.array([[1, 2], [3, 4]]))
|
|
assert t2.shape == (2, 2)
|
|
|
|
# From other tensor data
|
|
t3 = Tensor(t2.data)
|
|
assert t3.shape == t2.shape
|
|
|
|
def test_tensor_properties(self):
|
|
"""Test tensor properties work correctly."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
data = np.random.randn(3, 4, 5)
|
|
t = Tensor(data)
|
|
|
|
assert t.shape == (3, 4, 5)
|
|
assert t.dtype == data.dtype
|
|
assert np.array_equal(t.data, data)
|
|
|
|
def test_tensor_for_neural_networks(self):
|
|
"""Test tensor supports operations needed by neural networks."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
# Weights and inputs
|
|
weights = Tensor(np.random.randn(10, 20))
|
|
inputs = Tensor(np.random.randn(32, 10))
|
|
|
|
# Should support matrix multiplication (key for Dense layers)
|
|
# This tests if the tensor data can be used with numpy operations
|
|
output_data = inputs.data @ weights.data
|
|
output = Tensor(output_data)
|
|
|
|
assert output.shape == (32, 20)
|
|
|
|
|
|
class TestTensorMemoryManagement:
|
|
"""Test tensor memory usage and copying behavior."""
|
|
|
|
def test_tensor_memory_sharing(self):
|
|
"""Test tensor memory behavior."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
original_data = np.array([1, 2, 3])
|
|
t = Tensor(original_data)
|
|
|
|
# Tensor should maintain reference to data
|
|
assert np.shares_memory(t.data, original_data) or np.array_equal(t.data, original_data)
|
|
|
|
def test_tensor_copy_semantics(self):
|
|
"""Test tensor copying doesn't break."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
t1 = Tensor([1, 2, 3])
|
|
t2 = Tensor(t1.data.copy())
|
|
|
|
# Should be different tensors with same values
|
|
assert np.array_equal(t1.data, t2.data)
|
|
assert not np.shares_memory(t1.data, t2.data)
|
|
|
|
|
|
class TestTensorIntegrationReadiness:
|
|
"""Test tensor is ready for integration with other modules."""
|
|
|
|
def test_ready_for_activations(self):
|
|
"""Test tensor works with activation-like operations."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
t = Tensor(np.array([-1, 0, 1, 2]))
|
|
|
|
# Should support element-wise operations (for ReLU, Sigmoid, etc.)
|
|
relu_result = Tensor(np.maximum(0, t.data))
|
|
assert relu_result.shape == t.shape
|
|
assert np.array_equal(relu_result.data, [0, 0, 1, 2])
|
|
|
|
def test_ready_for_layers(self):
|
|
"""Test tensor works with layer-like operations."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
# Batch of inputs
|
|
x = Tensor(np.random.randn(32, 784)) # MNIST-like
|
|
weights = Tensor(np.random.randn(784, 128))
|
|
bias = Tensor(np.random.randn(128))
|
|
|
|
# Dense layer operation: x @ W + b
|
|
output_data = x.data @ weights.data + bias.data
|
|
output = Tensor(output_data)
|
|
|
|
assert output.shape == (32, 128)
|
|
|
|
def test_ready_for_spatial_operations(self):
|
|
"""Test tensor works with spatial/CNN operations."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
# Image tensor (batch, height, width, channels)
|
|
image = Tensor(np.random.randn(8, 32, 32, 3))
|
|
|
|
# Should support reshaping for spatial operations
|
|
flattened = Tensor(image.data.reshape(8, -1))
|
|
assert flattened.shape == (8, 32*32*3)
|
|
|
|
# Should support slicing for convolution-like operations
|
|
patch = Tensor(image.data[:, :3, :3, :]) # 3x3 patch
|
|
assert patch.shape == (8, 3, 3, 3)
|
|
|
|
def test_ready_for_autograd(self):
|
|
"""Test tensor is ready for gradient computation."""
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
# Should be able to create tensors that could track gradients
|
|
t = Tensor(np.array([1.0, 2.0, 3.0]))
|
|
|
|
# Should support operations that will need gradients
|
|
squared = Tensor(t.data ** 2)
|
|
assert squared.shape == t.shape |