Spaces:
Sleeping
Sleeping
File size: 5,927 Bytes
75bb597 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 | {
"id": "data_tokenization",
"concept": "Data & Tokenization",
"fiction": "The knight enters the Great Archive and meets the Runesmiths, learning how to mine raw knowledge and chisel it into mathematical runes.",
"questions": [
{
"id": "dt_q1",
"shown": {
"question": "What is the name of the official Hugging Face library used to efficiently download, load, and process vast amounts of data?",
"concept_brief": "Ask the knight for the name of the specific grimoire dedicated entirely to handling the great tomes of knowledge."
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"datasets",
"dataset"
]
}
},
"reveal": "The `datasets` library."
},
{
"id": "dt_q2",
"shown": {
"question": "Which main function from the `datasets` library is used to fetch a dataset from the Hub or a local file?",
"concept_brief": "What is the precise incantation used to summon a tome from the archive to the reading desk?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"load_dataset"
]
}
},
"reveal": "The `load_dataset` function."
},
{
"id": "dt_q3",
"shown": {
"question": "When a dataset is too massive to fit into your computer's memory (RAM), what feature allows you to load and process the data on-the-fly as you iterate through it?",
"concept_brief": "If the tome is too heavy to lift, what magic allows a knight to read the pages as they flow past like a river?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"streaming",
"stream"
]
}
},
"reveal": "Streaming."
},
{
"id": "dt_q4",
"shown": {
"question": "What method is used to efficiently apply a transformation function to every single example in a dataset?",
"concept_brief": "By what command does a scribe cast a translation spell over every single page of a tome at once?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"map",
"map()"
]
}
},
"reveal": "The `map` method."
},
{
"id": "dt_q5",
"shown": {
"question": "Explain in your own words why machine learning models cannot understand raw text directly and require tokenization.",
"concept_brief": "Ask the knight why spirits of logic cannot hear mortal words. Why must speech be turned into numbers?"
},
"check": {
"strategy": "llm_judge",
"spec": {
"rubric": [
"Mentions that models are mathematical/computational engines",
"Mentions the need to convert text into numbers or IDs"
]
}
},
"reveal": "Models are mathematical entities and require text to be converted into numerical token IDs to process them."
},
{
"id": "dt_q6",
"shown": {
"question": "Which class in the `transformers` library is most commonly used to automatically load the correct tokenizer for a given model?",
"concept_brief": "What specific incantation from the grimoire summons the exact Runesmith matched to a spirit?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"autotokenizer",
"AutoTokenizer"
]
}
},
"reveal": "The `AutoTokenizer` class."
},
{
"id": "dt_q7",
"shown": {
"question": "Write one line of Python code that encodes the text \"hello\" into token IDs using a `tokenizer` object.",
"concept_brief": "The Runesmith's own tool can transform plain speech -- invoke it upon the text."
},
"check": {
"strategy": "ast",
"spec": {
"must_call_any": [
"encode",
"tokenizer"
]
}
},
"reveal": "tokenizer(\"hello\") (or tokenizer.encode(\"hello\"))"
},
{
"id": "dt_q8",
"shown": {
"question": "What is the term for the complete, fixed set of unique tokens a specific tokenizer knows how to process?",
"concept_brief": "What is the name of the great ledger that lists every single rune a specific smith has ever learned?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"vocabulary",
"vocab"
]
}
},
"reveal": "The vocabulary (or vocab)."
},
{
"id": "dt_q9",
"shown": {
"question": "When processing multiple texts of different lengths in a batch, what technique is used to add empty tokens so they all match the length of the longest text?",
"concept_brief": "When casting multiple spells at once, they must all be the same length. How do we fill the empty space in the shorter incantations?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"padding",
"pad"
]
}
},
"reveal": "Padding."
},
{
"id": "dt_q10",
"shown": {
"question": "When padding is applied, what secondary array of 1s and 0s is generated to tell the model which tokens are real text and which are just padding?",
"concept_brief": "What is the name of the filter that tells the spirit to ignore the empty filler runes and focus only on the true magic?"
},
"check": {
"strategy": "keyword",
"spec": {
"correct_any": [
"attention mask",
"attention_mask"
]
}
},
"reveal": "The attention mask."
}
]
} |