aprender-core 0.29.2

pub(crate) use super::*;

#[test]
fn test_conv1d_shape() {
    let conv = Conv1d::new(16, 32, 3);
    let x = Tensor::ones(&[4, 16, 100]);
    let y = conv.forward(&x);

    // Output length = (100 - 3) / 1 + 1 = 98
    assert_eq!(y.shape(), &[4, 32, 98]);
}

#[test]
fn test_conv1d_with_padding() {
    let conv = Conv1d::with_padding(16, 32, 3, 1);
    let x = Tensor::ones(&[4, 16, 100]);
    let y = conv.forward(&x);

    // Output length = (100 + 2*1 - 3) / 1 + 1 = 100
    assert_eq!(y.shape(), &[4, 32, 100]);
}

#[test]
fn test_conv1d_with_stride() {
    let conv = Conv1d::with_stride(16, 32, 3, 2);
    let x = Tensor::ones(&[4, 16, 100]);
    let y = conv.forward(&x);

    // Output length = (100 - 3) / 2 + 1 = 49
    assert_eq!(y.shape(), &[4, 32, 49]);
}

#[test]
fn test_conv1d_parameters() {
    let conv = Conv1d::new(16, 32, 3);
    let params = conv.parameters();

    assert_eq!(params.len(), 2); // weight + bias
    assert_eq!(params[0].shape(), &[32, 16, 3]); // weight
    assert_eq!(params[1].shape(), &[32]); // bias
}

#[test]
fn test_conv2d_shape() {
    let conv = Conv2d::new(3, 64, 3);
    let x = Tensor::ones(&[4, 3, 32, 32]);
    let y = conv.forward(&x);

    // Output: (32 - 3) / 1 + 1 = 30
    assert_eq!(y.shape(), &[4, 64, 30, 30]);
}

#[test]
fn test_conv2d_with_padding() {
    let conv = Conv2d::with_padding(3, 64, 3, 1);
    let x = Tensor::ones(&[4, 3, 32, 32]);
    let y = conv.forward(&x);

    // Output: (32 + 2 - 3) / 1 + 1 = 32 (same size)
    assert_eq!(y.shape(), &[4, 64, 32, 32]);
}

#[test]
fn test_conv2d_with_stride() {
    let conv = Conv2d::with_stride(3, 64, 3, 2);
    let x = Tensor::ones(&[4, 3, 32, 32]);
    let y = conv.forward(&x);

    // Output: (32 - 3) / 2 + 1 = 15
    assert_eq!(y.shape(), &[4, 64, 15, 15]);
}

#[test]
fn test_conv2d_parameters() {
    let conv = Conv2d::new(3, 64, 3);
    let params = conv.parameters();

    assert_eq!(params.len(), 2);
    assert_eq!(params[0].shape(), &[64, 3, 3, 3]); // weight
    assert_eq!(params[1].shape(), &[64]); // bias
}

#[test]
fn test_maxpool1d_shape() {
    let pool = MaxPool1d::new(2);
    let x = Tensor::ones(&[4, 16, 100]);
    let y = pool.forward(&x);

    assert_eq!(y.shape(), &[4, 16, 50]);
}

#[test]
fn test_maxpool2d_shape() {
    let pool = MaxPool2d::new(2);
    let x = Tensor::ones(&[4, 64, 32, 32]);
    let y = pool.forward(&x);

    assert_eq!(y.shape(), &[4, 64, 16, 16]);
}

#[test]
fn test_maxpool2d_values() {
    let pool = MaxPool2d::new(2);
    // Create input with known values
    let x = Tensor::new(
        &[
            1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
        ],
        &[1, 1, 4, 4],
    );
    let y = pool.forward(&x);

    assert_eq!(y.shape(), &[1, 1, 2, 2]);
    // Max of 2x2 windows:
    // [1,2,5,6] -> 6, [3,4,7,8] -> 8
    // [9,10,13,14] -> 14, [11,12,15,16] -> 16
    assert_eq!(y.data(), &[6.0, 8.0, 14.0, 16.0]);
}

#[test]
fn test_avgpool2d_shape() {
    let pool = AvgPool2d::new(2);
    let x = Tensor::ones(&[4, 64, 32, 32]);
    let y = pool.forward(&x);

    assert_eq!(y.shape(), &[4, 64, 16, 16]);
}

#[test]
fn test_avgpool2d_values() {
    let pool = AvgPool2d::new(2);
    let x = Tensor::new(
        &[
            1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0, 12.0, 13.0, 14.0, 15.0, 16.0,
        ],
        &[1, 1, 4, 4],
    );
    let y = pool.forward(&x);

    // Avg of 2x2 windows:
    // [1,2,5,6] -> 3.5, [3,4,7,8] -> 5.5
    // [9,10,13,14] -> 11.5, [11,12,15,16] -> 13.5
    assert_eq!(y.data(), &[3.5, 5.5, 11.5, 13.5]);
}

#[test]
fn test_global_avg_pool2d() {
    let pool = GlobalAvgPool2d::new();
    let x = Tensor::ones(&[2, 64, 7, 7]);
    let y = pool.forward(&x);

    assert_eq!(y.shape(), &[2, 64]);
    // All ones, so average is 1.0
    assert!(y.data().iter().all(|&v| (v - 1.0).abs() < 1e-5));
}

#[test]
fn test_flatten() {
    let flatten = Flatten::new();
    let x = Tensor::ones(&[4, 64, 7, 7]);
    let y = flatten.forward(&x);

    assert_eq!(y.shape(), &[4, 64 * 7 * 7]);
}

#[test]
fn test_flatten_from_dim() {
    let flatten = Flatten::from_dim(2);
    let x = Tensor::ones(&[4, 64, 7, 7]);
    let y = flatten.forward(&x);

    assert_eq!(y.shape(), &[4, 64, 49]);
}

// ==========================================================================
// Additional Coverage Tests
// ==========================================================================

// Conv1d tests
#[test]
fn test_conv1d_no_bias() {
    let conv = Conv1d::with_options(16, 32, 3, 1, 0, false);
    let params = conv.parameters();

    assert_eq!(params.len(), 1); // weight only
    assert_eq!(params[0].shape(), &[32, 16, 3]);
}

#[test]
fn test_conv1d_getters() {
    let conv = Conv1d::with_options(16, 32, 5, 2, 3, true);
    assert_eq!(conv.kernel_size(), 5);
    assert_eq!(conv.stride(), 2);
    assert_eq!(conv.padding(), 3);
}

#[test]
fn test_conv1d_parameters_mut() {
    let mut conv = Conv1d::new(8, 16, 3);
    let params = conv.parameters_mut();
    assert_eq!(params.len(), 2);
    assert_eq!(params[0].shape(), &[16, 8, 3]);
}

#[test]
fn test_conv1d_parameters_mut_no_bias() {
    let mut conv = Conv1d::with_options(8, 16, 3, 1, 0, false);
    let params = conv.parameters_mut();
    assert_eq!(params.len(), 1);
}

#[test]
fn test_conv1d_debug() {
    let conv = Conv1d::new(16, 32, 3);
    let debug_str = format!("{:?}", conv);
    assert!(debug_str.contains("Conv1d"));
    assert!(debug_str.contains("in_channels"));
    assert!(debug_str.contains("out_channels"));
}

#[test]
fn test_conv1d_forward_no_bias() {
    let conv = Conv1d::with_options(4, 8, 3, 1, 0, false);
    let x = Tensor::ones(&[2, 4, 10]);
    let y = conv.forward(&x);

    // Output: (10 - 3) / 1 + 1 = 8
    assert_eq!(y.shape(), &[2, 8, 8]);
}

#[test]
fn test_conv1d_padding_zero_values() {
    // Test that padding correctly zero-pads
    let conv = Conv1d::with_padding(1, 1, 3, 1);
    let x = Tensor::new(&[1.0, 2.0, 3.0], &[1, 1, 3]);
    let y = conv.forward(&x);

    // With padding=1, output should have same length
    assert_eq!(y.shape(), &[1, 1, 3]);
}

// Conv2d tests
#[test]
fn test_conv2d_no_bias() {
    let conv = Conv2d::with_options(3, 64, (3, 3), (1, 1), (0, 0), false);
    let params = conv.parameters();

    assert_eq!(params.len(), 1); // weight only
    assert_eq!(params[0].shape(), &[64, 3, 3, 3]);
}

#[test]
fn test_conv2d_getters() {
    let conv = Conv2d::with_options(3, 64, (5, 7), (2, 3), (1, 2), true);
    assert_eq!(conv.kernel_size(), (5, 7));
    assert_eq!(conv.stride(), (2, 3));
    assert_eq!(conv.padding(), (1, 2));
}

#[test]
fn test_conv2d_parameters_mut() {
    let mut conv = Conv2d::new(3, 32, 3);
    let params = conv.parameters_mut();
    assert_eq!(params.len(), 2);
    assert_eq!(params[0].shape(), &[32, 3, 3, 3]);
}

#[test]
fn test_conv2d_parameters_mut_no_bias() {
    let mut conv = Conv2d::with_options(3, 32, (3, 3), (1, 1), (0, 0), false);
    let params = conv.parameters_mut();
    assert_eq!(params.len(), 1);
}

#[test]
fn test_conv2d_debug() {
    let conv = Conv2d::new(3, 64, 3);
    let debug_str = format!("{:?}", conv);
    assert!(debug_str.contains("Conv2d"));
    assert!(debug_str.contains("in_channels"));
    assert!(debug_str.contains("kernel_size"));
}

#[test]
fn test_conv2d_forward_no_bias() {
    let conv = Conv2d::with_options(2, 4, (3, 3), (1, 1), (0, 0), false);
    let x = Tensor::ones(&[2, 2, 8, 8]);
    let y = conv.forward(&x);

    // Output: (8 - 3) / 1 + 1 = 6
    assert_eq!(y.shape(), &[2, 4, 6, 6]);
}

#[test]
fn test_conv2d_padding_zero_values() {
    // Test padding in 2D conv
    let conv = Conv2d::with_padding(1, 1, 3, 1);
    let x = Tensor::ones(&[1, 1, 4, 4]);
    let y = conv.forward(&x);

    // With padding=1 and kernel=3, output should be same size
    assert_eq!(y.shape(), &[1, 1, 4, 4]);
}

#[test]
fn test_conv2d_non_square_kernel() {
    let conv = Conv2d::with_options(2, 4, (3, 5), (1, 1), (0, 0), true);
    let x = Tensor::ones(&[1, 2, 10, 12]);
    let y = conv.forward(&x);

    // Output H: (10 - 3) / 1 + 1 = 8
    // Output W: (12 - 5) / 1 + 1 = 8
    assert_eq!(y.shape(), &[1, 4, 8, 8]);
}

// MaxPool1d tests
#[test]
fn test_maxpool1d_with_stride() {
    let pool = MaxPool1d::with_stride(3, 2);
    let x = Tensor::ones(&[2, 4, 10]);
    let y = pool.forward(&x);

    // Output: (10 - 3) / 2 + 1 = 4
    assert_eq!(y.shape(), &[2, 4, 4]);
}

#[test]
fn test_maxpool1d_values() {
    let pool = MaxPool1d::new(2);
    let x = Tensor::new(&[1.0, 4.0, 2.0, 3.0], &[1, 1, 4]);
    let y = pool.forward(&x);

    // Max of [1,4]=4, [2,3]=3
    assert_eq!(y.data(), &[4.0, 3.0]);
}

#[test]
fn test_maxpool1d_debug() {
    let pool = MaxPool1d::new(2);
    let debug_str = format!("{:?}", pool);
    assert!(debug_str.contains("MaxPool1d"));
}

// MaxPool2d tests
#[test]
fn test_maxpool2d_with_stride() {
    let pool = MaxPool2d::with_stride(3, 2);
    let x = Tensor::ones(&[2, 4, 10, 10]);
    let y = pool.forward(&x);

    // Output: (10 - 3) / 2 + 1 = 4
    assert_eq!(y.shape(), &[2, 4, 4, 4]);
}

#[test]
fn test_maxpool2d_with_options() {
    let pool = MaxPool2d::with_options((3, 5), (2, 3));
    let x = Tensor::ones(&[1, 2, 10, 14]);
    let y = pool.forward(&x);

    // Output H: (10 - 3) / 2 + 1 = 4
    // Output W: (14 - 5) / 3 + 1 = 4
    assert_eq!(y.shape(), &[1, 2, 4, 4]);
}

#[test]
fn test_maxpool2d_debug() {
    let pool = MaxPool2d::new(2);
    let debug_str = format!("{:?}", pool);
    assert!(debug_str.contains("MaxPool2d"));
}

// AvgPool2d tests
#[test]
fn test_avgpool2d_with_stride() {
    let pool = AvgPool2d::with_stride(3, 2);
    let x = Tensor::ones(&[2, 4, 10, 10]);
    let y = pool.forward(&x);

    // Output: (10 - 3) / 2 + 1 = 4
    assert_eq!(y.shape(), &[2, 4, 4, 4]);
}

#[test]
fn test_avgpool2d_debug() {
    let pool = AvgPool2d::new(2);
    let debug_str = format!("{:?}", pool);
    assert!(debug_str.contains("AvgPool2d"));
}

// GlobalAvgPool2d tests
#[test]
fn test_global_avgpool2d_default() {
    let pool = GlobalAvgPool2d::default();
    let x = Tensor::ones(&[2, 8, 4, 4]);
    let y = pool.forward(&x);

    assert_eq!(y.shape(), &[2, 8]);
}

#[test]
fn test_global_avgpool2d_debug() {
    let pool = GlobalAvgPool2d::new();
    let debug_str = format!("{:?}", pool);
    assert!(debug_str.contains("GlobalAvgPool2d"));
}

#[test]
fn test_global_avgpool2d_varied_values() {
    let pool = GlobalAvgPool2d::new();
    let x = Tensor::new(&[1.0, 2.0, 3.0, 4.0], &[1, 1, 2, 2]);
    let y = pool.forward(&x);

    // Average of [1, 2, 3, 4] = 2.5
    assert!((y.data()[0] - 2.5).abs() < 1e-5);
}

// Flatten tests
#[test]
fn test_flatten_default() {
    // Default sets start_dim to 0, so flattens everything
    let flatten = Flatten::default();
    let x = Tensor::ones(&[4, 8, 8, 8]);
    let y = flatten.forward(&x);

    // start_dim=0 flattens all: [4*8*8*8] = [2048]
    assert_eq!(y.shape(), &[2048]);
}

#[test]
fn test_flatten_debug() {
    let flatten = Flatten::new();
    let debug_str = format!("{:?}", flatten);
    assert!(debug_str.contains("Flatten"));
}

#[test]
fn test_flatten_no_op_for_2d() {
    // If input is already 2D, flatten should return it unchanged
    let flatten = Flatten::new();
    let x = Tensor::ones(&[4, 64]);
    let y = flatten.forward(&x);

    assert_eq!(y.shape(), &[4, 64]);
}

#[path = "tests_conv1d_contract.rs"]
mod tests_conv1d_contract;
#[path = "tests_flatten.rs"]
mod tests_flatten;