mirror of
https://github.com/MLSysBook/TinyTorch.git
synced 2026-06-03 08:51:06 -05:00
69d65d38c3
- Add new 'tito modules' command for comprehensive module status checking - Scans all modules in modules/ directory automatically - Shows file structure (dev file, tests, README) - Runs tests with --test flag - Provides detailed breakdown with --details flag - Remove duplicate/stub commands: - Remove 'tito status' (unimplemented stub) - Remove 'tito submit' (unimplemented stub) - Update 'tito test' command: - Focus on individual module testing with detailed output - Redirect 'tito test --all' to 'tito modules --test' with recommendation - Better error handling with available modules list - Add comprehensive documentation: - docs/development/testing-separation.md - explains module vs package checking - docs/development/command-cleanup-summary.md - documents CLI cleanup Key benefit: Clear separation between module development status (tito modules) and TinyTorch package functionality (tito info) with no confusing overlaps.
58 lines
2.1 KiB
Python
58 lines
2.1 KiB
Python
# AUTOGENERATED! DO NOT EDIT! File to edit: ../../modules/cnn/cnn_dev.ipynb.
|
|
|
|
# %% auto 0
|
|
__all__ = ['conv2d_naive', 'Conv2D', 'flatten']
|
|
|
|
# %% ../../modules/cnn/cnn_dev.ipynb 4
|
|
def conv2d_naive(input: np.ndarray, kernel: np.ndarray) -> np.ndarray:
|
|
"""
|
|
Naive 2D convolution (single channel, no stride, no padding).
|
|
Args:
|
|
input: 2D input array (H, W)
|
|
kernel: 2D filter (kH, kW)
|
|
Returns:
|
|
2D output array (H-kH+1, W-kW+1)
|
|
TODO: Implement the sliding window convolution using for-loops.
|
|
"""
|
|
raise NotImplementedError("Student implementation required")
|
|
|
|
# %% ../../modules/cnn/cnn_dev.ipynb 5
|
|
def conv2d_naive(input: np.ndarray, kernel: np.ndarray) -> np.ndarray:
|
|
H, W = input.shape
|
|
kH, kW = kernel.shape
|
|
out_H, out_W = H - kH + 1, W - kW + 1
|
|
output = np.zeros((out_H, out_W), dtype=input.dtype)
|
|
for i in range(out_H):
|
|
for j in range(out_W):
|
|
output[i, j] = np.sum(input[i:i+kH, j:j+kW] * kernel)
|
|
return output
|
|
|
|
# %% ../../modules/cnn/cnn_dev.ipynb 9
|
|
class Conv2D:
|
|
"""
|
|
2D Convolutional Layer (single channel, single filter, no stride/pad).
|
|
Args:
|
|
kernel_size: (kH, kW)
|
|
TODO: Initialize a random kernel and implement the forward pass using conv2d_naive.
|
|
"""
|
|
def __init__(self, kernel_size: Tuple[int, int]):
|
|
raise NotImplementedError("Student implementation required")
|
|
def forward(self, x: Tensor) -> Tensor:
|
|
raise NotImplementedError("Student implementation required")
|
|
def __call__(self, x: Tensor) -> Tensor:
|
|
return self.forward(x)
|
|
|
|
# %% ../../modules/cnn/cnn_dev.ipynb 10
|
|
class Conv2D:
|
|
def __init__(self, kernel_size: Tuple[int, int]):
|
|
self.kernel = np.random.randn(*kernel_size).astype(np.float32)
|
|
def forward(self, x: Tensor) -> Tensor:
|
|
return Tensor(conv2d_naive(x.data, self.kernel))
|
|
def __call__(self, x: Tensor) -> Tensor:
|
|
return self.forward(x)
|
|
|
|
# %% ../../modules/cnn/cnn_dev.ipynb 12
|
|
def flatten(x: Tensor) -> Tensor:
|
|
"""Flatten a 2D tensor to 1D (for connecting to Dense)."""
|
|
return Tensor(x.data.flatten()[None, :])
|