pyerualjetwork 4.0.5__py3-none-any.whl

pyerualjetwork/data_operations_cuda.py

@@ -0,0 +1,463 @@
+from tqdm import tqdm
+import cupy as cp
+from colorama import Fore, Style
+import sys
+import math
+import numpy as np
+
+def encode_one_hot(y_train, y_test=None, summary=False):
+    """
+    Performs one-hot encoding on y_train and y_test data.
+
+    Args:
+        y_train (cupy.ndarray): Train label data.
+        y_test (cupy.ndarray): Test label data. (optional).
+        summary (bool): If True, prints the class-to-index mapping. Default: False
+   
+    Returns:
+        tuple: One-hot encoded y_train and (if given) y_test.
+    """
+
+    if len(y_train) < 256:
+        if y_train.dtype != cp.uint8:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint8, copy=False)
+    elif len(y_train) <= 32767:
+        if y_train.dtype != cp.uint16:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint16, copy=False)
+    else:
+        if y_train.dtype != cp.uint32:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint32, copy=False)
+
+    if y_test is not None:
+        if len(y_test) < 256:
+            if y_test.dtype != cp.uint8:
+                y_test = cp.array(y_test, copy=False).astype(cp.uint8, copy=False)
+        elif len(y_test) <= 32767:
+            if y_test.dtype != cp.uint16:
+                y_test = cp.array(y_test, copy=False).astype(cp.uint16, copy=False)
+        else:
+            if y_test.dtype != cp.uint32:
+                y_test = cp.array(y_test, copy=False).astype(cp.uint32, copy=False)
+
+    y_train = y_train.get()
+    y_test = y_test.get()
+
+    classes = np.unique(y_train)
+    class_count = len(classes)
+
+    class_to_index = {cls: idx for idx, cls in enumerate(classes)}
+
+    if summary:
+        print("Class-to-index mapping:")
+        for cls, idx in class_to_index.items():
+            print(f"  {idx}: {cls}")
+
+    y_train_encoded = np.zeros((y_train.shape[0], class_count))
+    for i, label in enumerate(y_train):
+        y_train_encoded[i, class_to_index[label]] = 1
+
+    if y_test is not None:
+        y_test_encoded = np.zeros((y_test.shape[0], class_count))
+        for i, label in enumerate(y_test):
+            y_test_encoded[i, class_to_index[label]] = 1
+        return cp.array(y_train_encoded), cp.array(y_test_encoded)
+
+    return cp.array(y_train_encoded)
+
+
+def decode_one_hot(encoded_data):
+    """
+    Decodes one-hot encoded data to original categorical labels.
+
+    Args:
+        encoded_data (cupy.ndarray): One-hot encoded data with shape (n_samples, n_classes).
+
+    Returns:
+        cupy.ndarray: Decoded categorical labels with shape (n_samples,).
+    """
+
+    decoded_labels = cp.argmax(encoded_data, axis=1)
+
+    return decoded_labels
+
+
+def split(X, y, test_size, random_state, dtype=cp.float32):
+    """
+    Splits the given X (features) and y (labels) data into training and testing subsets.
+
+    Args:
+        X (cupy.ndarray): Features data.
+        y (cupy.ndarray): Labels data.
+        test_size (float or int): Proportion or number of samples for the test subset.
+        random_state (int or None): Seed for random state.
+        dtype (cupy.dtype): Data type for the arrays. np.float32 by default. Example: cp.float64 or cp.float16. [fp32 for balanced devices, fp64 for strong devices, fp16 for weak devices: not reccomended!] (optional)
+
+    Returns:
+        tuple: x_train, x_test, y_train, y_test as ordered training and testing data subsets.
+    """
+    X = cp.array(X, copy=False).astype(dtype, copy=False)
+    if len(y) < 256:
+        if y.dtype != cp.uint8:
+            y = cp.array(y, copy=False).astype(cp.uint8, copy=False)
+    elif len(y) <= 32767:
+        if y.dtype != cp.uint16:
+            y = cp.array(y, copy=False).astype(cp.uint16, copy=False)
+    else:
+        if y.dtype != cp.uint32:
+            y = cp.array(y, copy=False).astype(cp.uint32, copy=False)
+
+    
+    num_samples = X.shape[0]
+
+    if isinstance(test_size, float):
+        test_size = int(test_size * num_samples)
+    elif isinstance(test_size, int):
+        if test_size > num_samples:
+            raise ValueError(
+                "test_size cannot be larger than the number of samples.")
+    else:
+        raise ValueError("test_size should be float or int.")
+
+    if random_state is not None:
+        cp.random.seed(random_state)
+
+    indices = cp.arange(num_samples)
+    cp.random.shuffle(indices)
+
+    test_indices = indices[:test_size]
+    train_indices = indices[test_size:]
+
+    x_train, x_test = X[train_indices], X[test_indices]
+    y_train, y_test = y[train_indices], y[test_indices]
+
+    return x_train, x_test, y_train, y_test
+
+
+def manuel_balancer(x_train, y_train, target_samples_per_class, dtype=cp.float32):
+    """
+    Generates synthetic examples to balance classes to the specified number of examples per class.
+
+    Arguments:
+    x_train -- Input dataset (examples) - cupy array format
+    y_train -- Class labels (one-hot encoded) - cupy array format
+    target_samples_per_class -- Desired number of samples per class
+    dtype (cupy.dtype): Data type for the arrays. np.float32 by default. Example: cp.float64 or cp.float16. [fp32 for balanced devices, fp64 for strong devices, fp16 for weak devices: not reccomended!] (optional)
+
+    Returns:
+    x_balanced -- Balanced input dataset (cupy array format)
+    y_balanced -- Balanced class labels (one-hot encoded, cupy array format)
+    """
+    from .ui import loading_bars
+
+    bar_format = loading_bars()[0]
+    
+    x_train = cp.array(x_train, copy=False).astype(dtype, copy=False)
+
+    if len(y_train[0]) < 256:
+        if y_train.dtype != cp.uint8:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint8, copy=False)
+    elif len(y_train[0]) <= 32767:
+        if y_train.dtype != cp.uint16:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint16, copy=False)
+    else:
+        if y_train.dtype != cp.uint32:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint32, copy=False)
+
+
+    classes = cp.arange(y_train.shape[1])
+    class_count = len(classes)
+    
+    x_balanced = []
+    y_balanced = []
+
+    for class_label in tqdm(range(class_count),leave=False, ascii="▱▰",
+            bar_format=bar_format,desc='Augmenting Data',ncols= 52):
+        class_indices = cp.where(cp.argmax(y_train, axis=1) == class_label)[0]
+        num_samples = len(class_indices)
+        
+        if num_samples > target_samples_per_class:
+      
+            selected_indices = cp.random.choice(class_indices, target_samples_per_class, replace=False)
+            x_balanced.append(x_train[selected_indices])
+            y_balanced.append(y_train[selected_indices])
+            
+        else:
+            
+            x_balanced.append(x_train[class_indices])
+            y_balanced.append(y_train[class_indices])
+
+            if num_samples < target_samples_per_class:
+                
+                samples_to_add = target_samples_per_class - num_samples
+                additional_samples = cp.zeros((samples_to_add, x_train.shape[1]))
+                additional_labels = cp.zeros((samples_to_add, y_train.shape[1]))
+                
+                for i in range(samples_to_add):
+
+                    random_indices = cp.random.choice(class_indices, 2, replace=False)
+                    sample1 = x_train[random_indices[0]]
+                    sample2 = x_train[random_indices[1]]
+
+                    
+                    synthetic_sample = sample1 + (sample2 - sample1) * cp.random.rand()
+
+                    additional_samples[i] = synthetic_sample
+                    additional_labels[i] = y_train[class_indices[0]]
+                    
+                    
+                x_balanced.append(additional_samples)
+                y_balanced.append(additional_labels)
+    
+    x_balanced = cp.vstack(x_balanced)
+    y_balanced = cp.vstack(y_balanced)
+
+    return x_balanced, y_balanced
+
+
+def auto_balancer(x_train, y_train, dtype=cp.float32):
+
+    """
+   Function to balance the training data across different classes.
+
+   Arguments:
+   x_train (list): Input data for training.
+   
+   y_train (list): Labels corresponding to the input data.
+   
+   dtype (cupy.dtype): Data type for the arrays. np.float32 by default. Example: cp.float64 or cp.float16. [fp32 for balanced devices, fp64 for strong devices, fp16 for weak devices: not reccomended!] (optional)
+
+   Returns:
+   tuple: A tuple containing balanced input data and labels.
+    """
+    from .ui import loading_bars
+
+    x_train = cp.array(x_train, copy=False).astype(dtype, copy=False)
+
+    if len(y_train[0]) < 256:
+        if y_train.dtype != cp.uint8:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint8, copy=False)
+    elif len(y_train[0]) <= 32767:
+        if y_train.dtype != cp.uint16:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint16, copy=False)
+    else:
+        if y_train.dtype != cp.uint32:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint32, copy=False)
+
+    bar_format = loading_bars()[0]
+    
+    classes = cp.arange(y_train.shape[1])
+    class_count = len(classes)
+    
+    try:
+        ClassIndices = {i: cp.where(cp.array(y_train)[:, i] == 1)[
+            0] for i in range(class_count)}
+        classes = [len(ClassIndices[i]) for i in range(class_count)]
+
+        if len(set(classes)) == 1:
+            print(Fore.WHITE + "INFO: Data have already balanced. from: auto_balancer" + Style.RESET_ALL)
+            return x_train, y_train
+
+        MinCount = min(classes)
+
+        BalancedIndices = []
+        for i in tqdm(range(class_count),leave=False, ascii="▱▰",
+            bar_format= bar_format, desc='Balancing Data',ncols=70):
+            if len(ClassIndices[i]) > MinCount:
+                SelectedIndices = cp.random.choice(
+                    ClassIndices[i], MinCount, replace=False)
+            else:
+                SelectedIndices = ClassIndices[i]
+            BalancedIndices.extend(SelectedIndices)
+
+        BalancedInputs = [x_train[idx] for idx in BalancedIndices]
+        BalancedLabels = [y_train[idx] for idx in BalancedIndices]
+
+        permutation = cp.random.permutation(len(BalancedInputs))
+        BalancedInputs = cp.array(BalancedInputs)[permutation]
+        BalancedLabels = cp.array(BalancedLabels)[permutation]
+
+        print(Fore.GREEN + "Data Succesfully Balanced from: " + str(len(x_train)
+                                                                                 ) + " to: " + str(len(BalancedInputs)) + ". from: auto_balancer " + Style.RESET_ALL)
+    except:
+        print(Fore.RED + "ERROR: Inputs and labels must be same length check parameters")
+        sys.exit()
+
+    return BalancedInputs, BalancedLabels
+
+
+def synthetic_augmentation(x_train, y_train, dtype=cp.float32):
+    """
+    Generates synthetic examples to balance classes with fewer examples using CuPy.
+    Arguments:
+    
+    x_train -- Input dataset (examples) - cupy array format
+    
+    y_train -- Class labels (one-hot encoded) - cupy array format
+    
+    dtype (cupy.dtype): Data type for the arrays. np.float32 by default. Example: cp.float64 or cp.float16. [fp32 for balanced devices, fp64 for strong devices, fp16 for weak devices: not reccomended!] (optional)
+    
+    Returns:
+    x_train_balanced -- Balanced input dataset (cupy array format)
+    y_train_balanced -- Balanced class labels (one-hot encoded, cupy array format)
+    """
+    from .ui import loading_bars
+
+    bar_format = loading_bars()[0]
+    
+    x = x_train.astype(dtype, copy=False)
+
+    if len(y_train[0]) < 256:
+        if y_train.dtype != cp.uint8:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint8, copy=False)
+    elif len(y_train[0]) <= 32767:
+        if y_train.dtype != cp.uint16:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint16, copy=False)
+    else:
+        if y_train.dtype != cp.uint32:
+            y_train = cp.array(y_train, copy=False).astype(cp.uint32, copy=False)
+
+    y = y_train
+
+    classes = cp.arange(y_train.shape[1])
+    class_count = len(classes)
+    class_distribution = {i: 0 for i in range(class_count)}
+    
+    y_cpu = cp.asnumpy(y) if isinstance(y, cp.ndarray) else y
+    for label in y_cpu:
+        class_distribution[cp.argmax(label).item()] += 1
+        
+    max_class_count = max(class_distribution.values())
+    x_balanced = list(x)
+    y_balanced = list(y)
+    
+    for class_label in tqdm(range(class_count), leave=False, ascii="▱▰",
+            bar_format=bar_format, desc='Augmenting Data', ncols=52):
+        class_indices = [i for i, label in enumerate(y) if cp.argmax(label) == class_label]
+        num_samples = len(class_indices)
+        
+        if num_samples < max_class_count:
+            while num_samples < max_class_count:
+                random_indices = cp.random.choice(
+                    cp.array(class_indices), 2, replace=False)
+                sample1 = x[random_indices[0]]
+                sample2 = x[random_indices[1]]
+                
+                synthetic_sample = sample1 + \
+                    (sample2 - sample1) * cp.random.rand()
+                
+                x_balanced.append(synthetic_sample)
+                y_balanced.append(y[class_indices[0]])
+                num_samples += 1
+    
+    x_balanced = cp.array(x_balanced)
+    y_balanced = cp.array(y_balanced)
+    
+    return x_balanced, y_balanced
+
+def standard_scaler(x_train=None, x_test=None, scaler_params=None, dtype=cp.float32):
+    """
+    Standardizes training and test datasets. x_test may be None.
+
+    Args:
+        x_train: cupy.ndarray
+        
+        x_test: cupy.ndarray (optional)
+        
+        scaler_params (optional for using model)
+        
+        dtype (cupy.dtype): Data type for the arrays. cp.float32 by default. Example: cp.float64 or cp.float16. [fp32 for balanced devices, fp64 for strong devices, fp16 for weak devices: not reccomended!] (optional)
+
+    Returns:
+        list:
+        Scaler parameters: mean and std
+        tuple
+        Standardized training and test datasets
+    """
+    if x_train is not None and scaler_params is None and x_test is not None:
+        x_train = x_train.astype(dtype, copy=False)
+        x_test = x_test.astype(dtype, copy=False)
+
+        mean = cp.mean(x_train, axis=0)
+        std = cp.std(x_train, axis=0)
+        
+        train_data_scaled = (x_train - mean) / std
+        test_data_scaled = (x_test - mean) / std
+
+        train_data_scaled = cp.nan_to_num(train_data_scaled, nan=0)
+        test_data_scaled = cp.nan_to_num(test_data_scaled, nan=0)
+
+        scaler_params = [mean, std]
+
+        return scaler_params, train_data_scaled, test_data_scaled
+    
+    if scaler_params is None and x_train is None and x_test is not None:
+        return x_test.astype(dtype, copy=False)  # sample data not scaled
+            
+    if scaler_params is not None:
+        x_test = x_test.astype(dtype, copy=False)
+        scaled_data = (x_test - scaler_params[0]) / scaler_params[1]
+        scaled_data = cp.nan_to_num(scaled_data, nan=0)
+
+        return scaled_data  # sample data scaled
+    
+    
+def normalization(
+    Input,  # num: Input data to be normalized.
+    dtype=cp.float32
+):
+    """
+    Normalizes the input data using maximum absolute scaling.
+
+    Args:
+        Input (num): Input data to be normalized.
+
+        dtype (cupy.dtype): Data type for the arrays. cp.float32 by default. Example: cp.float64 or cp.float16. [fp32 for balanced devices, fp64 for strong devices, fp16 for weak devices: not reccomended!] (optional)
+
+    Returns:
+        (num) Scaled input data after normalization.
+    """
+
+    MaxAbs = cp.max(cp.abs(Input.astype(dtype, copy=False)))
+    return (Input / MaxAbs)
+
+
+def find_closest_factors(a):
+
+    root = int(math.sqrt(a))
+    
+    for i in range(root, 0, -1):
+        if a % i == 0:
+            j = a // i
+            return i, j
+        
+
+def batcher(x_test, y_test, batch_size=1):
+    y_labels = cp.argmax(y_test, axis=1)  # Sınıf etiketlerini belirle
+    
+    unique_labels = cp.unique(y_labels)  # Tüm sınıfları bul
+    total_samples = sum(
+        int(cp.sum(y_labels == class_label) * batch_size) for class_label in unique_labels
+    )
+    
+    sampled_x = cp.empty((total_samples, x_test.shape[1]), dtype=x_test.dtype)
+    sampled_y = cp.empty((total_samples, y_test.shape[1]), dtype=y_test.dtype)
+    
+    offset = 0
+    for class_label in unique_labels:
+        # Sınıfa ait indeksleri bulun
+        class_indices = cp.where(y_labels == class_label)[0]
+        
+        # Örnek sayısını belirle
+        num_samples = int(len(class_indices) * batch_size)
+        
+        # Rastgele örnek seç
+        sampled_indices = cp.random.choice(class_indices, num_samples, replace=False)
+        
+        # Veriyi sampled dizilerine yaz
+        sampled_x[offset:offset + num_samples] = x_test[sampled_indices]
+        sampled_y[offset:offset + num_samples] = y_test[sampled_indices]
+        
+        # Kaydırmayı güncelle
+        offset += num_samples
+    
+    return sampled_x, sampled_y

pyerualjetwork/help.py

@@ -0,0 +1,16 @@
+from .activation_functions import all_activations
+
+def activation_potentiation():
+
+    activations_list = all_activations()
+
+    print('All available activations: ',  activations_list, "\n\nYOU CAN COMBINE EVERY ACTIVATION. EXAMPLE: ['linear', 'tanh'] or ['waveakt', 'linear', 'sine'].")
+
+    return activations_list
+
+def docs_and_examples():
+
+    print('PLAN document: https://github.com/HCB06/PyerualJetwork/tree/main/Welcome_to_PLAN\n')
+    print('PLAN examples: https://github.com/HCB06/PyerualJetwork/tree/main/Welcome_to_PyerualJetwork/ExampleCodes\n')
+    print('PLANEAT examples: https://github.com/HCB06/PyerualJetwork/tree/main/Welcome_to_PyerualJetwork/ExampleCodes/PLANEAT\n')
+    print('PyerualJetwork document and examples: https://github.com/HCB06/PyerualJetwork/tree/main/Welcome_to_PyerualJetwork')

pyerualjetwork/loss_functions.py

@@ -0,0 +1,21 @@
+
+import numpy as np
+
+def categorical_crossentropy(y_true_batch, y_pred_batch):
+    epsilon = 1e-7
+    y_pred_batch = np.clip(y_pred_batch, epsilon, 1. - epsilon)
+    
+    losses = -np.sum(y_true_batch * np.log(y_pred_batch), axis=1)
+
+    mean_loss = np.mean(losses)
+    return mean_loss
+
+
+def binary_crossentropy(y_true_batch, y_pred_batch):
+    epsilon = 1e-7
+    y_pred_batch = np.clip(y_pred_batch, epsilon, 1. - epsilon)
+    
+    losses = -np.mean(y_true_batch * np.log(y_pred_batch) + (1 - y_true_batch) * np.log(1 - y_pred_batch), axis=1)
+
+    mean_loss = np.mean(losses)
+    return mean_loss

pyerualjetwork/loss_functions_cuda.py

@@ -0,0 +1,21 @@
+
+import cupy as cp
+
+def categorical_crossentropy(y_true_batch, y_pred_batch):
+    epsilon = 1e-7
+    y_pred_batch = cp.clip(y_pred_batch, epsilon, 1. - epsilon)
+    
+    losses = -cp.sum(y_true_batch * cp.log(y_pred_batch), axis=1)
+
+    mean_loss = cp.mean(losses)
+    return mean_loss
+
+
+def binary_crossentropy(y_true_batch, y_pred_batch):
+    epsilon = 1e-7
+    y_pred_batch = cp.clip(y_pred_batch, epsilon, 1. - epsilon)
+    
+    losses = -cp.mean(y_true_batch * cp.log(y_pred_batch) + (1 - y_true_batch) * cp.log(1 - y_pred_batch), axis=1)
+
+    mean_loss = cp.mean(losses)
+    return mean_loss