Format for poetry and add debugging

africat/aa_create_dataset.py

@@ -1,4 +1,10 @@
 #!/usr/bin/python
+'''
+  1. Load XML file
+  2. Create structure
+  3. Preprocess the data to remove punctuations, digits, spaces and making the text lower.
+    This helps reduce the vocab of the data (as now, "Cat ~" is "cat")
+'''
 
 import argparse
 import os
@@ -8,31 +14,16 @@ import re
 import string
 from string import digits
 import warnings
-
 import html
 from xml.etree import ElementTree as ET
-
+# data manupulation libs
-#data manupulation libs
 import csv
 import pandas as pd
 from pandarallel import pandarallel
 
-parser = argparse.ArgumentParser(
+def write_csv(data, output):
-  description='Turn XML data files into a dataset for use with pytorch', 
+  with open(output, 'w', encoding="utf-8") as f:
-  add_help=True,
+    data.to_csv(f, encoding="utf-8", quoting=csv.QUOTE_ALL)
-)
-parser.add_argument('--output', '-o', required=True, help='path of output CSV file')
-parser.add_argument('--input',  '-i', required=True, help='path of input directory containing XML files')
-args = parser.parse_args()
-
-if os.path.isdir(args.input) is False:
-  print(f"{args.input} is not a directory or does not exist");
-  sys.exit(1)
-
-#1. Load XML file
-#2. Create structure
-#3. Preprocess the data to remove punctuations, digits, spaces and making the text lower.
-#.  This helps reduce the vocab of the data (as now, "Cat ~" is "cat")
 
 def insert_line_numbers(txt):
   return "\n".join([f"{n+1:03d} {line}" for n, line in enumerate(txt.split("\n"))])
@@ -48,99 +39,153 @@ def partial_unescape(s):
       parts[i] = html.unescape(parts[i])
   return "".join(parts)
 
-articles = list()
+def parse_and_extract(input_dir, verbose):
-#allCats = list()
+  articles = list()
 
-total, plain, utf8, iso88591, failed = 0, 0, 0, 0, 0
+  total, plain, utf8, iso88591, failed = 0, 0, 0, 0, 0
-for root, dirs, files in os.walk(args.input):
+  for root, dirs, files in os.walk(input_dir):
     dirs.sort()
+    if verbose > 0:
       print(root)
     for file in sorted(files):
       #if re.search('2022\/10\/09', root) and re.search('0028.aans$', file):
       if re.search('.aans$', file):
         xml_file = os.path.join(root, file)
         total += 1
+
         try:
           with open(xml_file, 'r', encoding="ASCII") as f:
             content = f.read()
-        #print(f"ASCII read succeeded in {xml_file}")
+          if verbose > 1:
+            print(f"ASCII read succeeded in {xml_file}")
           plain += 1
         except Exception as e:
-        #print(f"ASCII read failed, trying UTF-8 in {xml_file} : {e}")
+          if verbose > 1:
+            print(f"ASCII read failed, trying UTF-8 in {xml_file} : {e}")
           try:
             with open(xml_file, 'r', encoding="UTF-8") as f:
               content = f.read()
-          #print(f"UTF-8 read succeeded in {xml_file}")
+            if verbose > 1:
+              print(f"UTF-8 read succeeded in {xml_file}")
             utf8 += 1
           except Exception as e:
-          #print(f"UTF-8 read failed, trying ISO-8859-1 in {xml_file} : {e}")
+            if verbose > 1:
+              print(f"UTF-8 read failed, trying ISO-8859-1 in {xml_file} : {e}")
             try:
               with open(xml_file, 'r', encoding="ISO-8859-1") as f:
                 content = f.read()
-            #print(f"ISO-8859-1 read succeeded in {xml_file}")
+              if verbose > 1:
+                print(f"ISO-8859-1 read succeeded in {xml_file}")
               iso88591 += 1
             except Exception as e:
               print(f"UTF-8 and ISO-8859-1 read failed in {xml_file} : {e}")
+              if verbose > 2:
                 print(content)
               failed += 1
         content = partial_unescape(content)
         content = local_clean(content)
-      #print(content)
+        if verbose > 3:
+          print(content)
 
         key = re.sub('^.*\/(\d{4})\/(\d{2})\/(\d{2})\/(\d{4}).aans$', '\g<1>\g<2>\g<3>\g<4>', xml_file)
         try:
           doc = ET.fromstring(content)
+
           entry = dict()
           entry["key"] = key
+
           cats = list()
-        for cat in doc.findall('category'):
+          for cat in doc.findall('./category'):
-          #if cat not in allCats:
-          #  allCats.append(cat)
             cats.append(cat.text)
-        #entry["categories"] = cats
+          #entry["categories"] = cats          # if you want a list
-        entry["categories"] = ";".join(cats)
+          entry["categories"] = ";".join(cats) # if you want a string
+
           text = list()
+          lang = ""
           try:
-          #text = "\n".join([p.text for p in doc.find('./body')])
             for p in doc.find('./body'):
               if p.text is not None:
                 text.append(p.text)
-          if text is not None and len(cats) > 1:
+            lang = doc.find('./language').text
-            entry["content"] = "\n".join(text)
-            articles.append(entry)
           except Exception as e:
             print(f"{xml_file} : {e}")
+
+          if text is not None and len(cats) > 1:
+            entry["content"] = "\n".join(text)
+            entry["language"] = lang
+            articles.append(entry)
+
         except ET.ParseError as e:
+          if verbose > 1:
             print(insert_line_numbers(content))
           print("Parse error in " + xml_file + " : ", e)
           raise(SystemExit)
 
-print("total:    {: 7d}\nplain:    {: 7d}\nutf8:     {: 7d}\niso88591: {: 7d}\nfailed:   {: 7d}\n".format(total, plain, utf8, iso88591, failed))
+  if verbose > 0:
+    print("total:    {: 7d}\nplain:    {: 7d}\nutf8:     {: 7d}\niso88591: {: 7d}\nfailed:   {: 7d}\n".format(total, plain, utf8, iso88591, failed))
 
-#sys.exit(0)
+  #sys.exit(0)
+  return articles
 
-data = pd.DataFrame(articles)
-data.set_index("key", inplace=True)
 
-#print(data.categories)
+def scrub_data(articles, verbose):
+  data = pd.DataFrame(articles)
+  data.set_index("key", inplace=True)
 
-# Initialization
+  #if verbose > 2:
-pandarallel.initialize()
+  #  print(data.categories)
 
-# Lowercase everything
+  # Initialization
-data['content'] = data.content.parallel_apply(lambda x: x.lower())
+  pandarallel.initialize()
 
-# Remove special characters
+  # Lowercase everything
-exclude = set(string.punctuation) #set of all special chars
+  data['content'] = data.content.parallel_apply(lambda x: x.lower())
-data['content'] = data.content.parallel_apply(lambda x: ''.join(ch for ch in x if ch not in exclude))
 
-# Remove digits
+  # Remove special characters
-remove_digits = str.maketrans('','',digits)
+  exclude = set(string.punctuation) #set of all special chars
-data['content'] = data.content.parallel_apply(lambda x: x.translate(remove_digits))
+  data['content'] = data.content.parallel_apply(lambda x: ''.join(ch for ch in x if ch not in exclude))
 
-# Remove extra spaces
+  # Remove digits
-data['content']=data.content.parallel_apply(lambda x: x.strip())
+  remove_digits = str.maketrans('','',digits)
-data['content']=data.content.parallel_apply(lambda x: re.sub(" +", " ", x))
+  data['content'] = data.content.parallel_apply(lambda x: x.translate(remove_digits))
 
-with open(args.output, 'w', encoding="utf-8") as f:
+  # Remove extra spaces
-  data.to_csv(f, encoding="utf-8", quoting=csv.QUOTE_ALL)
+  data['content']=data.content.parallel_apply(lambda x: x.strip())
+  data['content']=data.content.parallel_apply(lambda x: re.sub(" +", " ", x))
+
+  # TODO: lemmas? See spaCy
+
+  return data
+
+def main():
+  parser = argparse.ArgumentParser(
+    description='Turn XML data files into a dataset for use with pytorch',
+    add_help=True,
+  )
+  parser.add_argument('--output', '-o',
+    required=True,
+    help='path of output CSV file')
+  parser.add_argument('--input',  '-i',
+    required=True,
+    help='path of input directory containing XML files')
+  parser.add_argument('--verbose',  '-v',
+    type=int, nargs='?',
+    const=1,     # Default value if -v is supplied
+    default=0,   # Default value if -v is not supplied
+    help='print debugging')
+  args = parser.parse_args()
+
+  if os.path.isdir(args.input) is False:
+    print(f"{args.input} is not a directory or does not exist");
+    sys.exit(1)
+
+  articles = parse_and_extract(args.input, args.verbose)
+
+  data = scrub_data(articles, args.verbose)
+
+  write_csv(data, args.output)
+
+  return
+
+if __name__ == "__main__":
+  main()

africat/categorise.py

@@ -6,80 +6,84 @@ import sys
 import pprint
 import re
 import string
+import time
 import warnings
-
+# data manupulation
-#data manupulation libs
 import csv
 import random
 import pandas as pd
 import numpy as np
 #from pandarallel import pandarallel
 from tqdm import tqdm
-
+# torch
-#torch libs
 import torch
 import torchdata.datapipes as dp
 import torchtext.transforms as T
 from torchtext.vocab import build_vocab_from_iterator
 from torch.utils.data import Dataset, DataLoader
+from torch import nn
 
-parser = argparse.ArgumentParser(
+story_num = 40 # XXX None for all
-  description='Classify text data according to categories',
-  add_help=True,
-)
-parser.add_argument('action', help='train or classify')
-parser.add_argument('--input', '-i', required=True, help='path of CSV file containing dataset')
-parser.add_argument('--output', '-o', help='path to trained model')
-args = parser.parse_args()
 
-if args.action != 'train' and args.action != 'classify':
+# Hyperparameters
-  print("ERROR: train or classify data")
+EPOCHS = 10       # epoch
-  sys.exit(1)
+LR = 5            # learning rate
+BATCH_SIZE = 64   # batch size for training
 
-if args.action == 'classify' and s.path.isfile(model_storage) is None:
+def read_csv(input_csv, rows=None, verbose=0):
-  print("No model found for classification; running training instead")
+  if verbose > 0:
-  args.action = 'train'
+    with open(input_csv, 'r', encoding="utf-8") as f:
-
-if os.path.isfile(args.input) is False:
-  print(f"{args.input} is not a valid file")
-  sys.exit(1)
-
-#with open(args.input, 'r', encoding="utf-8") as f:
-#  data = pd.read_csv(f, encoding="utf-8", quoting=csv.QUOTE_ALL)
-
-with open(args.input, 'r', encoding="utf-8") as f:
       data = pd.concat(
         [chunk for chunk in tqdm(
           pd.read_csv(f,
             encoding="utf-8",
             quoting=csv.QUOTE_ALL,
-        nrows=200,              ## XXX
+            nrows=rows,
-        chunksize=100),
+            chunksize=50,
+          ),
           desc='Loading data'
         )])
+  else:
+    with open(input_csv, 'r', encoding="utf-8") as f:
+      data = pd.read_csv(f,
+        encoding="utf-8",
+        quoting=csv.QUOTE_ALL,
+        nrows=rows,
+      )
 
-data.dropna(axis='index', inplace=True)
+  data.dropna(axis='index', inplace=True)
+  #print(data)
+  #sys.exit(0)
+  return data
 
-#print(data)
-#sys.exit(0)
 
 '''
   Create Training and Validation sets
 '''
-# Create a list of ints till len of data
+def split_dataset(data, verbose=0):
-data_idx = list(range(len(data)))
+  # Create a list of ints till len of data
-np.random.shuffle(data_idx)
+  data_idx = list(range(len(data)))
+  np.random.shuffle(data_idx)
 
-# Get indexes for validation and train
+  # Get indexes for validation and train
-split_percent = 0.95
+  split_percent = 0.05
-num_train = int(len(data) * split_percent)
+  num_valid = int(len(data) * split_percent)
-valid_idx, train_idx = data_idx[num_train:], data_idx[:num_train]
+  #num_tests = int(len(data) * split_percent)
-print("Length of train_data: {}".format(len(train_idx)))
+  #train_idx = data_idx[num_valid:-num_tests]
-print("Length of valid_data: {}".format(len(valid_idx)))
+  train_idx = data_idx[num_valid:]
+  valid_idx = data_idx[:num_valid]
+  #tests_idx = data_idx[-num_tests:]
+  if verbose > 0:
+    print("Length of train_data: {}".format(len(train_idx)))
+    print("Length of valid_data: {}".format(len(valid_idx)))
+    #print("Length of tests_data: {}".format(len(tests_idx)))
 
-# Create the training and validation sets, as dataframes
+  # Create the training and validation sets, as dataframes
-train_data = data.iloc[train_idx].reset_index().drop('index', axis=1)
+  train_data = data.iloc[train_idx].reset_index().drop('index', axis=1)
-valid_data = data.iloc[valid_idx].reset_index().drop('index', axis=1)
+  valid_data = data.iloc[valid_idx].reset_index().drop('index', axis=1)
+  #tests_data = data.iloc[tests_idx].reset_index().drop('index', axis=1)
+  #return(train_data, valid_data, tests_data)
+  return(train_data, valid_data)
 
 
 '''
@@ -88,21 +92,24 @@ valid_data = data.iloc[valid_idx].reset_index().drop('index', axis=1)
 '''
 class TextCategoriesDataset(Dataset):
   ''' Dataset of Text and Categories '''
-  def __init__(self, df, text_column, cats_column, transform=None):
+  def __init__(self, df, text_column, cats_column, lang_column, transform=None, verbose=0):
     '''
     Arguments:
       df (panda.Dataframe): csv content, loaded as dataframe
       text_column (str): the name of the column containing the text
       cats_column (str): the name of the column containing
         semicolon-separated categories
+      text_column (str): the name of the column containing the language
       transform (callable, optional): Optional transform to be
         applied on a sample.
     '''
     self.df = df
     self.transform = transform
+    self.verbose = verbose
 
-    self.texts = self.df[text_column]
+    self.text = self.df[text_column]
     self.cats = self.df[cats_column]
+    self.lang = self.df[lang_column]
 
     # index-to-token dict
     # <pad> : padding, used for padding the shorter sentences in a batch
@@ -146,8 +153,9 @@ class TextCategoriesDataset(Dataset):
       idx = idx.tolist()
 
     # Get the raw data
-    text = self.texts[idx]
+    text = self.text[idx]
     cats = self.cats[idx]
+    lang = self.lang[idx]
 
     if self.transform:
       text, cats = self.transform(text, cats)
@@ -186,25 +194,6 @@ class TextCategoriesDataset(Dataset):
       T.AddToken(2, begin=False)
     )
 
-'''
-dataset = TextCategoriesDataset(df=data,
-  text_column="content",
-  cats_column="categories",
-)
-'''
-train_dataset = TextCategoriesDataset(df=train_data,
-  text_column="content",
-  cats_column="categories",
-)
-valid_dataset = TextCategoriesDataset(df=valid_data,
-  text_column="content",
-  cats_column="categories",
-)
-#print(dataset[2])
-#for text, cat in enumerate(valid_dataset):
-#  print(text, cat)
-#sys.exit(0)
-
 
 '''
   Now that we have a dataset, let's create dataloader,
@@ -232,44 +221,183 @@ class CollateBatch:
     )
 
 
-# Hyperparameters
+class TextClassificationModel(nn.Module):
-EPOCHS = 10       # epoch
+  def __init__(self, input_size, output_size, verbose):
-LR = 5            # learning rate
+    super().__init__()
-BATCH_SIZE = 64   # batch size for training
 
-# Get cpu, gpu or mps device for training.
+  def forward(self, x):
-# Move tensor to the NVIDIA GPU if available
+    return x
-device = (
+
+
+def train(dataloader):
+  model.train()
+  total_acc, total_count = 0, 0
+  log_interval = 500
+  start_time = time.time()
+
+  for idx, (label, text) in enumerate(dataloader):
+    optimizer.zero_grad()
+    predicted_label = model(text)
+    loss = criterion(predicted_label, label)
+    loss.backward()
+    torch.nn.utils.clip_grad_norm_(model.parameters(), 0.1)
+    optimizer.step()
+    total_acc += (predicted_label.argmax(1) == label).sum().item()
+    total_count += label.size(0)
+    if idx % log_interval == 0 and idx > 0:
+      elapsed = time.time() - start_time
+      print(
+        "| epoch {:3d} | {:5d}/{:5d} batches "
+        "| accuracy {:8.3f}".format(
+          epoch, idx, len(dataloader), total_acc / total_count
+        )
+      )
+      total_acc, total_count = 0, 0
+      start_time = time.time()
+
+
+def evaluate(dataloader):
+  model.eval()
+  total_acc, total_count = 0, 0
+
+  with torch.no_grad():
+    for idx, (label, text) in enumerate(dataloader):
+      predicted_label = model(text)
+      loss = criterion(predicted_label, label)
+      total_acc += (predicted_label.argmax(1) == label).sum().item()
+      total_count += label.size(0)
+  return total_acc / total_count
+
+
+def main():
+  parser = argparse.ArgumentParser(
+    description='Classify text data according to categories',
+    add_help=True,
+  )
+  parser.add_argument('action',
+    help='train or classify')
+  parser.add_argument('--input', '-i',
+    required=True,
+    help='path of CSV file containing dataset')
+  parser.add_argument('--model', '-m',
+    #required=True, # XXX
+    help='path to training model')
+  parser.add_argument('--verbose',  '-v',
+    type=int, nargs='?',
+    const=1,     # Default value if -v is supplied
+    default=0,   # Default value if -v is not supplied
+    help='print debugging')
+  args = parser.parse_args()
+
+  if args.action != 'train' and args.action != 'classify':
+    print("ERROR: train or classify data")
+    sys.exit(1)
+
+  if args.action == 'classify' and s.path.isfile(model_storage) is None:
+    print("No model found for classification; running training instead")
+    args.action = 'train'
+
+  if os.path.isfile(args.input) is False:
+    print(f"{args.input} is not a valid file")
+    sys.exit(1)
+
+  data = read_csv(input_csv=args.input, rows=story_num, verbose=args.verbose)
+  train_data, valid_data, = split_dataset(data, verbose=args.verbose)
+
+  '''
+  dataset = TextCategoriesDataset(df=data,
+    text_column="content",
+    cats_column="categories",
+    lang_column="language",
+    verbose=args.verbose,
+  )
+  '''
+  train_dataset = TextCategoriesDataset(df=train_data,
+    text_column="content",
+    cats_column="categories",
+    lang_column="language",
+    verbose=args.verbose,
+  )
+  valid_dataset = TextCategoriesDataset(df=valid_data,
+    text_column="content",
+    cats_column="categories",
+    lang_column="language",
+    verbose=args.verbose,
+  )
+  #print(dataset[2])
+  #for text, cat in enumerate(valid_dataset):
+  #  print(text, cat)
+  #sys.exit(0)
+
+  # Get cpu, gpu or mps device for training.
+  # Move tensor to the NVIDIA GPU if available
+  device = (
     "cuda" if torch.cuda.is_available()
     else "xps" if hasattr(torch, "xpu") and torch.xpu.is_available()
     else "mps" if torch.backends.mps.is_available()
     else "cpu"
-)
+  )
-print(f"Using {device} device")
+  print(f"Using {device} device")
 
-
+  '''
-'''
+  dataloader = DataLoader(dataset,
-dataloader = DataLoader(dataset,
     batch_size=4,
     shuffle=True,
     num_workers=0,
     collate_fn=CollateBatch(pad_idx=dataset.stoi['<pad>']),
-)
+  )
-'''
+  '''
-train_dataloader = DataLoader(train_dataset,
+  train_dataloader = DataLoader(train_dataset,
     batch_size=BATCH_SIZE,
     shuffle=True,
     num_workers=0,
     collate_fn=CollateBatch(pad_idx=train_dataset.stoi['<pad>']),
-)
+  )
-valid_dataloader = DataLoader(valid_dataset,
+  valid_dataloader = DataLoader(valid_dataset,
     batch_size=BATCH_SIZE,
     shuffle=True,
     num_workers=0,
     collate_fn=CollateBatch(pad_idx=valid_dataset.stoi['<pad>']),
-)
+  )
-#for i_batch, sample_batched in enumerate(dataloader):
+  #for i_batch, sample_batched in enumerate(dataloader):
-#  print(i_batch, sample_batched[0], sample_batched[1])
+  #  print(i_batch, sample_batched[0], sample_batched[1])
-#sys.exit(0)
+  #sys.exit(0)
+
+  num_class = len(set([cats for key, cats, text, lang in train_data.values]))
+  input_size = len(train_dataset.text_vocab)
+  output_size = len(train_dataset.cats_vocab)
+  emsize = 64
+  model = TextClassificationModel(input_size, output_size, args.verbose).to(device)
 
 
+  criterion = torch.nn.CrossEntropyLoss()
+  optimizer = torch.optim.SGD(model.parameters(), lr=LR)
+  scheduler = torch.optim.lr_scheduler.StepLR(optimizer, 1.0, gamma=0.1)
+  total_accu = None
+
+  for epoch in range(1, EPOCHS + 1):
+    epoch_start_time = time.time()
+    train(train_dataloader)
+    accu_val = evaluate(valid_dataloader)
+    if total_accu is not None and total_accu > accu_val:
+      scheduler.step()
+    else:
+      total_accu = accu_val
+    print("-" * 59)
+    print(
+      "| end of epoch {:3d} | time: {:5.2f}s | "
+      "valid accuracy {:8.3f} ".format(
+        epoch, time.time() - epoch_start_time, accu_val
+      )
+    )
+    print("-" * 59)
+
+  print("Checking the results of test dataset.")
+  accu_test = evaluate(test_dataloader)
+  print("test accuracy {:8.3f}".format(accu_test))
+
+  return
+
+if __name__ == "__main__":
+  main()
+