from io import open
import random
import glob
import os
import unicodedata
import string
from dotenv import load_dotenv
import sys
import torch
load_dotenv() # do not remove from here. do the imports all togetger
 

def getDeviceInfo():
    _platform = sys.platform # Check platform
    gpu = ''
    if _platform == "darwin" and torch.backends.mps.is_available():
        gpu = 'mps'
    elif torch.cuda.is_available():
        gpu = 'cuda'
    else:
        gpu = 'cpu'
    return gpu
 
device = torch.device(getDeviceInfo())
print(f"Using device: {device}")

dataset = os.getenv("NLP1_DATASET_PATH")
 
def findFiles(path):
    return glob.glob(path)

all_letters = string.ascii_letters + " .,;'"
n_letters = len(all_letters)
 
# turning a unicode string to ASCII
def unicodeToAscii(s):
    return "".join(
        c
        for c in unicodedata.normalize("NFD", s)
        if unicodedata.category(c) != "Mn" and c in all_letters
    )
 
 
unicodeToAscii("Ślusàrski")

all_cat = []
cat_lines = {}
 
def readlines(filepath : string):
    lines = open(filepath, "r", encoding='utf-8').read().strip().split()
    return [unicodeToAscii(line) for line in lines]
 
 
for filename in findFiles(dataset):
    category = os.path.splitext(os.path.basename(filename))[0]
    all_cat.append(category)
    lines  = readlines(filename)
    cat_lines[category] = lines
 
 
 
num_cat  = len(all_cat)
 

import torch
 
 
v_to_i = {v: i for i, v in enumerate([letter for letter in all_letters])}
 
 
def lineToTensor(line: string) -> torch.tensor:
    t = [[v_to_i[l] for l in w] for w in line]
    tens = torch.zeros(
        len(t), 1, n_letters
    )  # Make a zeros tensor with dimen [input length,1,vocab size]
    for i, val in enumerate(t):
        tens[i][0][val] = 1
 
    return tens
 
# lineToTensor("hello, hi"), lineToTensor("J"), lineToTensor("Jones")
 

v_to_i

## Did some testing to verify the learning code functions and my code returns the same results
 
# Tutorial version
def tutorial_lineToTensor(line):
    tensor = torch.zeros(len(line), 1, n_letters)
    for li, letter in enumerate(line):
        tensor[li][0][all_letters.find(letter)] = 1
    return tensor
 
 
# Test function
def compare_outputs(input_line):
    out = lineToTensor(input_line)
    tutorial_output = tutorial_lineToTensor(input_line)
 
    print(f"Input: {input_line}")
 
    shape_match = out.shape == tutorial_output.shape
    val_match = torch.all(out == tutorial_output)
    assert val_match
    print(f"Shapes match: {shape_match}:{out.shape} \nValues match: {val_match}")
 
 
    if not val_match:
        print("Differences:")
        print(torch.where(out != tutorial_output))
 
 
# Test with sample inputs
test_inputs = [
    "hello",
    "WORLD",
    "NLP",
    "PyTorch",
    "Machine Learning",
    "A.I.",
    "Natural Language Processing",
    "\'.,;",  # Note: These special characters might not be in all_letters
    "Hello, World",
    "Test case; with punctuation.",
    "Mixed Case and Nums",
    "EmptySpaces   Test",
    "a",  # Single character test
    "",  # Empty string test
    "'.;,\',;.",  # Only punctuation
    "aA.,;'zZ",  # Mix of everything
]
 
for input_line in test_inputs:
    compare_outputs(input_line)
    print()
 

import torch.nn as nn
import torch.nn.functional as F
 
class RNN(nn.Module):
    def __init__(self, input_size : int , hidden_size : int , output_size : int):
        super(RNN, self).__init__()
        self.hidden_size = hidden_size
        self.l1 =  nn.Linear(input_size, hidden_size)
        self.l2 = nn.Linear(hidden_size, hidden_size)
        self.l3 = nn.Linear(hidden_size, output_size)
 
        # self.softmax = nn.LogSoftmax(dim=1)
 
 
    def forward(self, input, hidden):
        hidden  = F.tanh(self.l1(input) + self.l2(hidden)) # adding lin(hidden size) + lin(hidden size)
        output = self.l3(hidden)
        # output = self.softmax(output)
        return output, hidden
 
    def initHidden(self):
        return torch.zeros(1, self.hidden_size)

n_hidden = 128 # Some arbitary number for now
model = RNN(n_letters, n_hidden, num_cat).to(device)
 

print(model)

from torchinfo import summary
 
# Define input and hidden tensors for the summary
input_tensor = torch.zeros(1, n_letters)
hidden_tensor = model.initHidden()
 
# Print model summary using torchinfo
summary(RNN(n_letters, n_hidden, num_cat), input_data=[input_tensor, hidden_tensor], col_names=["input_size", "output_size", "num_params", "mult_adds"])

============================================================================================================================================
Layer (type:depth-idx)                   Input Shape               Output Shape              Param #                   Mult-Adds
============================================================================================================================================
RNN                                      [1, 57]                   [1, 18]                   --                        --
├─Linear: 1-1                            [1, 57]                   [1, 128]                  7,424                     7,424
├─Linear: 1-2                            [1, 128]                  [1, 128]                  16,512                    16,512
├─Linear: 1-3                            [1, 128]                  [1, 18]                   2,322                     2,322
├─LogSoftmax: 1-4                        [1, 18]                   [1, 18]                   --                        --
============================================================================================================================================
Total params: 26,258
Trainable params: 26,258
Non-trainable params: 0
Total mult-adds (Units.MEGABYTES): 0.03
============================================================================================================================================
Input size (MB): 0.00
Forward/backward pass size (MB): 0.00
Params size (MB): 0.11
Estimated Total Size (MB): 0.11
============================================================================================================================================

input = lineToTensor("Albert")
hidden = torch.zeros(1, n_hidden) # make a 1, 128 array
 
output, next_hidden = model(input, hidden)
 
print(output.shape, next_hidden.shape)

def categoryFromOutput(output):
    top_n, top_i = output.topk(1)
    category_i = top_i[0].item()
    return all_cat[category_i], category_i
 
 
print(categoryFromOutput(output))

randomChoice = lambda l: l[random.randint(0, len(l) - 1)]
 
def randomTrainingExample():
    category = randomChoice(all_cat) # get a random category
    line = randomChoice(cat_lines[category])
    category_tens = torch.tensor([all_cat.index(category)], dtype= torch.long)
    line_tensor  = lineToTensor(line)
    return category , line , category_tens , line_tensor
 
for i in range(10):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    print('category =', category, '/ line =', line)
 

criterion = nn.CrossEntropyLoss().to(device)

lr= 5e-3
 
def train(category_tensor, line_tensor):
    hidden  = model.initHidden()
    model.zero_grad()
 
 
    for i in range(line_tensor.size()[0]):
        output, hidden  = model(line_tensor[i] , hidden)
 
    loss = criterion(output, category_tensor)
    loss.backward()
 
 
    for p in model.parameters():
        p.data.add_(p.grad.data, alpha=-lr)
 
    return output, loss.item()
 
 
# train()

import time
import math
 
n_iters = 100000
print_every = 5000
plot_every = 1000
 
# Keep track of losses for plotting
current_loss = 0
all_losses = []
 
def timeSince(since):
    now = time.time()
    s = now - since
    m = math.floor(s / 60)
    s -= m * 60
    return '%dm %ds' % (m, s)
 
start = time.time()
 
for iter in range(1, n_iters + 1):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    output, loss = train(category_tensor, line_tensor)
    current_loss += loss
 
    # Print ``iter`` number, loss, name and guess
    if iter % print_every == 0:
        guess, guess_i = categoryFromOutput(output)
        correct = '✓' if guess == category else '✗ (%s)' % category
        print('%d %d%% (%s) %.4f %s / %s %s' % (iter, iter / n_iters * 100, timeSince(start), loss, line, guess, correct))
 
    # Add current loss avg to list of losses
    if iter % plot_every == 0:
        all_losses.append(current_loss / plot_every)
        current_loss = 0

5000 5% (0m 2s) 1.0455 Dagher / Arabic ✓
10000 10% (0m 5s) 0.0624 Pantelakos / Greek ✓
15000 15% (0m 7s) 1.3450 Breitbarth / German ✓
20000 20% (0m 10s) 1.1973 Luo / Vietnamese ✗ (Chinese)
25000 25% (0m 12s) 5.3349 Comino / Portuguese ✗ (Greek)
30000 30% (0m 15s) 0.7361 Marek / Polish ✓
35000 35% (0m 17s) 1.2471 Chong / Chinese ✗ (Korean)
40000 40% (0m 20s) 1.7424 Siena / Spanish ✗ (Italian)
45000 45% (0m 22s) 1.5558 Brant / Scottish ✗ (German)
50000 50% (0m 25s) 0.5561 Vega / Spanish ✓
55000 55% (0m 27s) 0.2479 Guadarrama / Spanish ✓
60000 60% (0m 30s) 4.7581 Donohoe / Greek ✗ (English)
65000 65% (0m 32s) 0.0341 Esimontovsky / Russian ✓
70000 70% (0m 35s) 0.2816 Teague / Irish ✓
75000 75% (0m 37s) 0.0618 Russell / Scottish ✓
80000 80% (0m 40s) 0.7942 Chu / Vietnamese ✗ (Korean)
85000 85% (0m 42s) 0.6592 Trujillo / Spanish ✓
90000 90% (0m 45s) 0.0065 Ruadhain / Irish ✓
95000 95% (0m 48s) 0.0178 Aitken / Scottish ✓
100000 100% (0m 50s) 1.9986 Brady / Irish ✗ (English)

import matplotlib.pyplot as plt
import matplotlib.ticker as ticker
 
plt.figure()
plt.plot(all_losses)

# Keep track of correct guesses in a confusion matrix
confusion = torch.zeros(num_cat, num_cat)
n_confusion = 10000
 
# Just return an output given a line
def evaluate(line_tensor):
    hidden = model.initHidden()
 
    for i in range(line_tensor.size()[0]):
        output, hidden = model(line_tensor[i], hidden)
 
    return output
 
# Go through a bunch of examples and record which are correctly guessed
for i in range(n_confusion):
    category, line, category_tensor, line_tensor = randomTrainingExample()
    output = evaluate(line_tensor)
    guess, guess_i = categoryFromOutput(output)
    category_i = all_cat.index(category)
    confusion[category_i][guess_i] += 1
 
# Normalize by dividing every row by its sum
for i in range(num_cat):
    confusion[i] = confusion[i] / confusion[i].sum()
 
# Set up plot
fig = plt.figure()
ax = fig.add_subplot(111)
cax = ax.matshow(confusion.numpy())
fig.colorbar(cax)
 
# Set up axes
ax.set_xticklabels([''] + all_cat, rotation=90)
ax.set_yticklabels([''] + all_cat)
 
# Force label at every tick
ax.xaxis.set_major_locator(ticker.MultipleLocator(1))
ax.yaxis.set_major_locator(ticker.MultipleLocator(1))
 
# sphinx_gallery_thumbnail_number = 2
plt.show()

def predict(input_line, n_predictions=3):
    print('\n> %s' % input_line)
    with torch.no_grad():
        output = evaluate(lineToTensor(input_line))
 
        # Get top N categories
        topv, topi = output.topk(n_predictions, 1, True)
        predictions = []
 
        for i in range(n_predictions):
            value = topv[0][i].item()
            category_index = topi[0][i].item()
            print('(%.2f) %s' % (value, all_cat[category_index]))
            predictions.append([value, all_cat[category_index]])
 
predict('Dovesky')
predict('Jackson')
predict('Satoshi')