mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-07-16 10:32:24 -05:00
6f86460ec0
- module_01 → 01_tensor - module_02 → 02_activations - module_03 → 03_layers - module_04 → 04_losses - module_05 → 05_autograd - module_06 → 06_optimizers - module_07 → 07_training - module_08 → 08_dataloader - module_09 → 09_spatial - module_10 → 10_tokenization - module_11 → 11_embeddings - module_12 → 12_attention - module_13 → 13_transformers - module_14 → 14_kvcaching - module_15 → 15_profiling This prevents misalignment between source and test directories. Tests now mirror the exact structure of modules/source/.
116 lines
3.5 KiB
Python
116 lines
3.5 KiB
Python
"""
|
|
Tests for Module 04: Dense/Networks
|
|
"""
|
|
|
|
import pytest
|
|
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 TestDenseExports:
|
|
"""Test that Dense layer is properly exported."""
|
|
|
|
def test_dense_import(self):
|
|
"""Test Dense can be imported from correct location."""
|
|
from tinytorch.core.layers import Dense
|
|
assert Dense is not None
|
|
|
|
def test_dense_creation(self):
|
|
"""Test Dense layer can be created."""
|
|
from tinytorch.core.layers import Dense
|
|
layer = Dense(10, 5)
|
|
assert layer.weights.shape == (10, 5)
|
|
|
|
|
|
class TestDenseForward:
|
|
"""Test Dense layer forward pass."""
|
|
|
|
def test_forward_shape(self):
|
|
"""Test output shape is correct."""
|
|
from tinytorch.core.layers import Dense
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
layer = Dense(10, 5)
|
|
x = Tensor(np.random.randn(32, 10))
|
|
output = layer(x)
|
|
|
|
assert output.shape == (32, 5)
|
|
|
|
def test_forward_with_bias(self):
|
|
"""Test forward pass with bias."""
|
|
from tinytorch.core.layers import Dense
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
layer = Dense(10, 5, use_bias=True)
|
|
x = Tensor(np.zeros((1, 10)))
|
|
output = layer(x)
|
|
|
|
# With zero input, output should equal bias
|
|
assert np.allclose(output.data, layer.bias.data)
|
|
|
|
def test_forward_without_bias(self):
|
|
"""Test forward pass without bias."""
|
|
from tinytorch.core.layers import Dense
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
layer = Dense(10, 5, use_bias=False)
|
|
x = Tensor(np.zeros((1, 10)))
|
|
output = layer(x)
|
|
|
|
# With zero input and no bias, output should be zero
|
|
assert np.allclose(output.data, 0)
|
|
|
|
|
|
class TestDenseIntegration:
|
|
"""Test Dense layer integration with other modules."""
|
|
|
|
def test_dense_with_tensor(self):
|
|
"""Test Dense works with Tensor (Module 02)."""
|
|
from tinytorch.core.layers import Dense
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
layer = Dense(10, 5)
|
|
|
|
# Weights and bias should be Tensors
|
|
assert isinstance(layer.weights, Tensor)
|
|
if layer.use_bias:
|
|
assert isinstance(layer.bias, Tensor)
|
|
|
|
def test_dense_with_activations(self):
|
|
"""Test Dense works with activations (Module 03)."""
|
|
from tinytorch.core.layers import Dense
|
|
from tinytorch.core.activations import ReLU, Sigmoid
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
layer = Dense(10, 5)
|
|
relu = ReLU()
|
|
sigmoid = Sigmoid()
|
|
|
|
x = Tensor(np.random.randn(16, 10))
|
|
h = layer(x)
|
|
h_relu = relu(h)
|
|
h_sigmoid = sigmoid(h)
|
|
|
|
assert h_relu.shape == h.shape
|
|
assert h_sigmoid.shape == h.shape
|
|
assert np.all(h_sigmoid.data >= 0) and np.all(h_sigmoid.data <= 1)
|
|
|
|
def test_dense_chain(self):
|
|
"""Test chaining multiple Dense layers."""
|
|
from tinytorch.core.layers import Dense
|
|
from tinytorch.core.tensor import Tensor
|
|
|
|
layer1 = Dense(784, 128)
|
|
layer2 = Dense(128, 64)
|
|
layer3 = Dense(64, 10)
|
|
|
|
x = Tensor(np.random.randn(32, 784))
|
|
h1 = layer1(x)
|
|
h2 = layer2(h1)
|
|
output = layer3(h2)
|
|
|
|
assert output.shape == (32, 10) |