Fix Lesson 5 NLP PyTorch track: installable requirements and torchtext 0.18 imports

lessons/5-NLP/13-TextRep/TextRepresentationPyTorch.ipynb

@@ -21,6 +21,9 @@
    "source": [
     "import torch\n",
     "import torchtext\n",
+    "import torchtext.data.utils\n",
+    "import torchtext.datasets\n",
+    "import torchtext.vocab\n",
     "import os\n",
     "import collections\n",
     "os.makedirs('./data',exist_ok=True)\n",

lessons/5-NLP/14-Embeddings/EmbeddingsPyTorch.ipynb

@@ -43,6 +43,7 @@
    "source": [
     "import torch\n",
     "import torchtext\n",
+    "import torchtext.vocab\n",
     "import numpy as np\n",
     "from torchnlp import *\n",
     "train_dataset, test_dataset, classes, vocab = load_dataset()\n",

lessons/5-NLP/14-Embeddings/torchnlp.py

@@ -1,6 +1,9 @@
 import builtins
 import torch
 import torchtext
+import torchtext.data.utils
+import torchtext.datasets
+import torchtext.vocab
 import collections
 import os
 

lessons/5-NLP/15-LanguageModeling/CBoW-PyTorch.ipynb

@@ -18,6 +18,9 @@
       "source": [
         "import torch\n",
         "import torchtext\n",
+        "import torchtext.data.utils\n",
+        "import torchtext.datasets\n",
+        "import torchtext.vocab\n",
         "import os\n",
         "import collections\n",
         "import builtins\n",
@@ -66,7 +69,8 @@
         "        counter.update(torchtext.data.utils.ngrams_iterator(tokenizer(line),ngrams=ngrams))\n",
         "        if i == lines_cnt:\n",
         "            break\n",
-        "    vocab = torchtext.vocab.Vocab(collections.Counter(dict(counter.most_common(vocab_size))), min_freq=min_freq)\n",
+        "    vocab = torchtext.vocab.vocab(collections.OrderedDict(counter.most_common(vocab_size)), min_freq=min_freq, specials=['<unk>', '<pad>'])\n",
+        "    vocab.set_default_index(vocab['<unk>'])\n",
         "    return train_dataset, test_dataset, classes, vocab, tokenizer"
       ],
       "metadata": {
@@ -385,7 +389,7 @@
       },
       "outputs": [],
       "source": [
-        "vectors = torch.stack([embedder(torch.tensor(vocab[s])) for s in vocab.itos], 0)"
+        "vectors = torch.stack([embedder(torch.tensor(vocab[s])) for s in vocab.get_itos()], 0)"
       ]
     },
     {
@@ -460,7 +464,7 @@
         "def close_words(x, n = 5):\n",
         "  vec = embedder(torch.tensor(vocab[x]))\n",
         "  top5 = np.linalg.norm(vectors.detach().numpy() - vec.detach().numpy(), axis = 1).argsort()[:n]\n",
-        "  return [ vocab.itos[x] for x in top5 ]\n",
+        "  return [ vocab.get_itos()[x] for x in top5 ]\n",
         "\n",
         "close_words('microsoft')"
       ]

lessons/5-NLP/16-RNN/torchnlp.py

@@ -1,6 +1,9 @@
 import builtins
 import torch
 import torchtext
+import torchtext.data.utils
+import torchtext.datasets
+import torchtext.vocab
 import collections
 import os
 

lessons/5-NLP/17-GenerativeNetworks/GenerativePyTorch.ipynb

@@ -30,6 +30,7 @@
    "source": [
     "import torch\n",
     "import torchtext\n",
+    "import torchtext.vocab\n",
     "import numpy as np\n",
     "from torchnlp import *\n",
     "train_dataset,test_dataset,classes,vocab = load_dataset()"

lessons/5-NLP/17-GenerativeNetworks/torchnlp.py

@@ -1,6 +1,9 @@
 import builtins
 import torch
 import torchtext
+import torchtext.data.utils
+import torchtext.datasets
+import torchtext.vocab
 import collections
 import os
 

lessons/5-NLP/18-Transformers/torchnlp.py

@@ -1,6 +1,9 @@
 import builtins
 import torch
 import torchtext
+import torchtext.data.utils
+import torchtext.datasets
+import torchtext.vocab
 import collections
 import os
 

lessons/5-NLP/requirements-pytorch.txt

@@ -1,15 +1,25 @@
+# The PyTorch NLP notebooks use torchtext, which is deprecated and no longer
+# developed. torchtext 0.18.0 is its final release. It is built against
+# torch 2.3 and only ships wheels for Python 3.8-3.12, so use Python 3.11 or
+# 3.12 (Pillow 13 needs Python >= 3.11).
+# torch, torchaudio, torchdata, torchtext and torchvision must be upgraded together.
+# torchtext.datasets needs torchdata < 0.10 (DataPipes) and portalocker.
+# torch 2.3 is built against NumPy 1.x, hence numpy < 2.
+# transformers 5.x requires torch >= 2.4, so stay on the transformers 4.x line.
 gensim>=4.3.0,<5
 huggingface==0.0.1
 matplotlib
 nltk==3.10.4
-numpy==1.22.0
-opencv-python==4.5.1.48
+numpy>=1.22,<2
+opencv-python==4.11.0.86
 Pillow==13.0.0
+portalocker>=2.0.0
 scikit-learn
 scipy
-torch==2.13.0
-torchaudio==0.8.1
+torch==2.3.1
+torchaudio==2.3.1
+torchdata==0.8.0
 torchinfo==0.0.8
-torchtext==0.9.1
-torchvision==0.9.1
-transformers==5.10.1
+torchtext==0.18.0
+torchvision==0.18.1
+transformers==4.57.6