From 518b2764e71d43d145ea5bce447602bd51154da9 Mon Sep 17 00:00:00 2001
From: Ryszard Staruch <ryssta@amu.edu.pl>
Date: Wed, 4 Dec 2024 14:54:02 +0100
Subject: [PATCH] Update notebook 08

---
 08_vector_representations.ipynb | 71 +++++++++++++++++++++++++++++----
 1 file changed, 64 insertions(+), 7 deletions(-)

diff --git a/08_vector_representations.ipynb b/08_vector_representations.ipynb
index 71c770a..31b6b9c 100644
--- a/08_vector_representations.ipynb
+++ b/08_vector_representations.ipynb
@@ -19,6 +19,21 @@
     "https://github.com/unslothai/unsloth - biblioteka do efektywnego finetune'owania LLMów (są gotowe notebooki z kodem na platformie Colab)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "#### Co to jest wektor?\n",
+    "\n",
+    "Wektor - jednowymiarowa macierz\n",
+    "\n",
+    "[0, 1, 0, 0, 0] - one hot encoding - tylko wartości 0/1\n",
+    "\n",
+    "[0, 2, 0, 5, 1, 100] - frequency encoding - liczby całkowite >= 0\n",
+    "\n",
+    "[-1.5, 0.0002, 5000.01] - wektor"
+   ]
+  },
   {
    "cell_type": "markdown",
    "metadata": {},
@@ -62,9 +77,18 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 62,
+   "execution_count": 67,
    "metadata": {},
-   "outputs": [],
+   "outputs": [
+    {
+     "name": "stderr",
+     "output_type": "stream",
+     "text": [
+      "c:\\Users\\ryssta\\AppData\\Local\\anaconda3\\Lib\\site-packages\\transformers\\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884\n",
+      "  warnings.warn(\n"
+     ]
+    }
+   ],
    "source": [
     "from transformers import GPT2Tokenizer, GPT2Model\n",
     "import torch\n",
@@ -78,7 +102,23 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 63,
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": [
+    "[\n",
+    "    [0.1, 0.2, 0.3], # Ala\n",
+    "    [-0.5, 0.5, 0.9], # ma\n",
+    "    ...\n",
+    "    # 50254\n",
+    "    ...\n",
+    "    [0.1, -0.1, -0.2] # w GPT2 jest 768 wartości w pojedynczym wektorze, a nie 3\n",
+    "]"
+   ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": 78,
    "metadata": {},
    "outputs": [
     {
@@ -86,20 +126,31 @@
      "output_type": "stream",
      "text": [
       "Tekst 'cat' jest konwertowany do tokenu 9246\n",
+      "\n",
+      "Tokenizacja\n",
       "{'input_ids': [33215], 'attention_mask': [1]}\n",
-      "cat\n"
+      "\n",
+      "Detokenizacja\n",
+      "computer\n",
+      "\n",
+      "Liczba tokenów w słowniku\n",
+      "50257\n"
      ]
     }
    ],
    "source": [
     "print(\"Tekst 'cat' jest konwertowany do tokenu 9246\")\n",
+    "print(\"\\nTokenizacja\")\n",
     "print(tokenizer(\"computer\"))\n",
-    "print(tokenizer.decode([9246]))"
+    "print(\"\\nDetokenizacja\")\n",
+    "print(tokenizer.decode([33215]))\n",
+    "print(\"\\nLiczba tokenów w słowniku\")\n",
+    "print(len(tokenizer))"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 66,
+   "execution_count": 73,
    "metadata": {},
    "outputs": [
     {
@@ -107,7 +158,11 @@
      "output_type": "stream",
      "text": [
       "Embedding tokenu: 9246\n",
+      "\n",
+      "Rozmiar embeddingu (wektora)\n",
       "torch.Size([1, 768])\n",
+      "\n",
+      "Wartości embeddingu\n",
       "tensor([[-0.0164, -0.0934,  0.2425,  0.1398,  0.0388, -0.2592, -0.2724, -0.1625,\n",
       "          0.1683,  0.0829,  0.0136, -0.2788,  0.1493,  0.1408,  0.0557, -0.3691,\n",
       "          0.2200, -0.0428,  0.2206,  0.0865,  0.1237, -0.1499,  0.1446, -0.1150,\n",
@@ -211,13 +266,15 @@
    "source": [
     "print(\"Embedding tokenu: 9246\")\n",
     "cat_embedding = embedding_layer(torch.LongTensor([9246]))\n",
+    "print(\"\\nRozmiar embeddingu (wektora)\")\n",
     "print(cat_embedding.shape)\n",
+    "print(\"\\nWartości embeddingu\")\n",
     "print(cat_embedding)"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 65,
+   "execution_count": null,
    "metadata": {},
    "outputs": [
     {