Added number of llama3 tokens to desc stats

.vscode/launch.json

@@ -0,0 +1,16 @@
+{
+    // Use IntelliSense to learn about possible attributes.
+    // Hover to view descriptions of existing attributes.
+    // For more information, visit: https://go.microsoft.com/fwlink/?linkid=830387
+    "version": "0.2.0",
+    "configurations": [
+        {
+            "name": "Python Debugger: Current File with Arguments",
+            "type": "debugpy",
+            "request": "launch",
+            "program": "${file}",
+            "console": "integratedTerminal",
+            "args": "--force"
+        }
+    ]
+}

README.md

@@ -114,6 +114,7 @@ language_bcp47:
 readme structure is inspired by:
 https://github.com/huggingface/datasets/blob/main/templates/README_guide.md -->
 
+
 # 🧨 Danish Dynaword
 
 |              |                                                                                                                                                             |
@@ -149,6 +150,7 @@ https://github.com/huggingface/datasets/blob/main/templates/README_guide.md -->
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 546.77K
+- **Number of tokens (Llama 3)**: 1.57B
 - **Average document length (characters)**: 8461.25
 
 <!-- END-DESC-STATS -->
@@ -262,7 +264,7 @@ Below follows a brief overview of the sources in the corpus along with their ind
 | [hest]              | [CC-0]               |
 | [jvj]               | [CC-BY-SA 4.0]       |
 | [naat]              | [CC-0]               |
-| [nordjyllandnews]              | [CC-0]               |
+| [nordjyllandnews]   | [CC-0]               |
 | [relig]             | [CC-0]               |
 | [retsinformationdk] | [Other (Danish Law)] |
 | [retspraksis]       | [CC-0]               |
@@ -314,3 +316,13 @@ We welcome contributions to the dataset such as new sources, better data filteri
 ### Citation Information
 
 This version expand upon existing dataset sources such as the [Danish gigaword](https://huggingface.co/danish-foundation-models). We recommend that you cite the source of the dataset when using these datasets.
+
+
+---
+
+<h3 style="display: flex; align-items: center;">
+  <a href="https://www.foundationmodels.dk">
+    <img src="./docs/icon.png" width="30" style="margin-right: 10px;" />
+  </a>
+  A&nbsp;<a href=https://www.foundationmodels.dk>Danish Foundation Models</a>&nbsp;dataset
+</h3>

data/adl/adl.md

@@ -23,6 +23,7 @@ Danish literature from 1700-2023 stemming for the Archive for Danish Literature
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 498
+- **Number of tokens (Llama 3)**: 58.49M
 - **Average document length (characters)**: 324932.24
 
 <!-- END-DESC-STATS -->

data/adl/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 498, "average_document_length": 324932.2429718876, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 498, "average_document_length": 324932.2429718876, "number_of_tokens": 58493311, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/botxt/botxt.md

@@ -24,6 +24,7 @@ Fictional texts of various kinds written in Bornholmsk, the dialect spoken on th
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 106
+- **Number of tokens (Llama 3)**: 847.97K
 - **Average document length (characters)**: 18972.42
 
 <!-- END-DESC-STATS -->

data/botxt/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 106, "average_document_length": 18972.415094339623, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 106, "average_document_length": 18972.415094339623, "number_of_tokens": 847973, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/dannet/dannet.md

@@ -25,6 +25,7 @@ A WordNet is a lexico-semantic network which show the meaning and the relation b
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 49.04K
+- **Number of tokens (Llama 3)**: 1.52M
 - **Average document length (characters)**: 90.80
 
 <!-- END-DESC-STATS -->

data/dannet/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 49040, "average_document_length": 90.80340538336053, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 49040, "average_document_length": 90.80340538336053, "number_of_tokens": 1523416, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/depbank/depbank.md

@@ -26,6 +26,7 @@ While the dataset was initially intended as a rich annotation, this corpora only
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 536
+- **Number of tokens (Llama 3)**: 185.45K
 - **Average document length (characters)**: 1018.90
 
 <!-- END-DESC-STATS -->

data/depbank/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 536, "average_document_length": 1018.8992537313433, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 536, "average_document_length": 1018.8992537313433, "number_of_tokens": 185454, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/ep/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 4213, "average_document_length": 74063.40469973891, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 4213, "average_document_length": 74063.40469973891, "number_of_tokens": 100888932, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/ep/ep.md

@@ -25,6 +25,7 @@ The europarl is a corpus of parallel text in 11 languages from the proceedings o
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 4.21K
+- **Number of tokens (Llama 3)**: 100.89M
 - **Average document length (characters)**: 74063.40
 
 <!-- END-DESC-STATS -->

data/ft/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 1315, "average_document_length": 266745.19163498096, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 1315, "average_document_length": 266745.19163498096, "number_of_tokens": 114087231, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/ft/ft.md

@@ -26,6 +26,7 @@ In the parliament hall, one speaker at a time addresses members of the parliamen
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 1.31K
+- **Number of tokens (Llama 3)**: 114.09M
 - **Average document length (characters)**: 266745.19
 
 <!-- END-DESC-STATS -->

data/gutenberg/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 66, "average_document_length": 290147.9393939394, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 66, "average_document_length": 290147.9393939394, "number_of_tokens": 6763317, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/gutenberg/gutenberg.md

@@ -24,6 +24,7 @@ Project Gutenberg is an online library of free eBooks. Project Gutenberg was the
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 66
+- **Number of tokens (Llama 3)**: 6.76M
 - **Average document length (characters)**: 290147.94
 
 <!-- END-DESC-STATS -->

data/hest/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 14391, "average_document_length": 82950.79104996179, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 14391, "average_document_length": 82950.79104996179, "number_of_tokens": 389325153, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/hest/hest.md

@@ -26,6 +26,7 @@ Its inclusion as training data for large language models have multiple times rea
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 14.39K
+- **Number of tokens (Llama 3)**: 389.33M
 - **Average document length (characters)**: 82950.79
 
 <!-- END-DESC-STATS -->

data/jvj/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 42, "average_document_length": 254893.66666666666, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 42, "average_document_length": 254893.66666666666, "number_of_tokens": 3549181, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/jvj/jvj.md

@@ -26,6 +26,7 @@ The works of the Danish author and poet, [Johannes V. Jensen](https://da.wikiped
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 42
+- **Number of tokens (Llama 3)**: 3.55M
 - **Average document length (characters)**: 254893.67
 
 <!-- END-DESC-STATS -->

data/naat/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 129, "average_document_length": 6832.387596899225, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 129, "average_document_length": 6832.387596899225, "number_of_tokens": 286677, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/naat/naat.md

@@ -22,6 +22,7 @@ A dataset of Danish speeches from 1930-2022.
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 129
+- **Number of tokens (Llama 3)**: 286.68K
 - **Average document length (characters)**: 6832.39
 
 <!-- END-DESC-STATS -->

data/nordjyllandnews/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 75219, "average_document_length": 1540.2673659580691, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 75219, "average_document_length": 1540.2673659580691, "number_of_tokens": 37905944, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/nordjyllandnews/nordjyllandnews.md

@@ -25,6 +25,7 @@ The data is derived from the Huggingface dataset [alexandrainst/nordjylland-news
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 75.22K
+- **Number of tokens (Llama 3)**: 37.91M
 - **Average document length (characters)**: 1540.27
 
 <!-- END-DESC-STATS -->

data/relig/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 66, "average_document_length": 53873.56060606061, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 66, "average_document_length": 53873.56060606061, "number_of_tokens": 1243970, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/relig/relig.md

@@ -22,6 +22,7 @@ Danish religious text from the 1700-2022.
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 66
+- **Number of tokens (Llama 3)**: 1.24M
 - **Average document length (characters)**: 53873.56
 
 <!-- END-DESC-STATS -->

data/retsinformationdk/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 64043, "average_document_length": 22248.525506300455, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 64043, "average_document_length": 22248.525506300455, "number_of_tokens": 516537034, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/retsinformationdk/retsinformationdk.md

@@ -31,6 +31,7 @@ It serves as a central repository for Danish legislation, administrative regulat
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 64.04K
+- **Number of tokens (Llama 3)**: 516.54M
 - **Average document length (characters)**: 22248.53
 
 <!-- END-DESC-STATS -->

data/retspraksis/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 4413, "average_document_length": 46323.66915930206, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 4413, "average_document_length": 46323.66915930206, "number_of_tokens": 57080441, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/retspraksis/retspraksis.md

@@ -25,6 +25,7 @@ It encompasses the body of legal decisions made by Danish courts, which play a s
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 4.41K
+- **Number of tokens (Llama 3)**: 57.08M
 - **Average document length (characters)**: 46323.67
 
 <!-- END-DESC-STATS -->

data/skat/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 14716, "average_document_length": 24026.800964936123, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 14716, "average_document_length": 24026.800964936123, "number_of_tokens": 122120389, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/skat/skat.md

@@ -22,6 +22,7 @@ Skat is the Danish tax authority. This dataset contains content from its website
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 14.72K
+- **Number of tokens (Llama 3)**: 122.12M
 - **Average document length (characters)**: 24026.80
 
 <!-- END-DESC-STATS -->

data/spont/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 411, "average_document_length": 9736.79805352798, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 411, "average_document_length": 9736.79805352798, "number_of_tokens": 1555942, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/spont/spont.md

@@ -24,6 +24,7 @@ The conversational corpus included originates from interdisciplinary research co
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 411
+- **Number of tokens (Llama 3)**: 1.56M
 - **Average document length (characters)**: 9736.80
 
 <!-- END-DESC-STATS -->

data/synne/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 178, "average_document_length": 785.3258426966293, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 178, "average_document_length": 785.3258426966293, "number_of_tokens": 52508, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/synne/synne.md

@@ -22,6 +22,7 @@ Dataset collected from [synnejysk forening's website](https://www.synnejysk.dk),
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 178
+- **Number of tokens (Llama 3)**: 52.51K
 - **Average document length (characters)**: 785.33
 
 <!-- END-DESC-STATS -->

data/tv2r/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 49137, "average_document_length": 1303.3820746077295, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 49137, "average_document_length": 1303.3820746077295, "number_of_tokens": 21671943, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/tv2r/tv2r.md

@@ -24,6 +24,7 @@ It contains articles of regional interest, written following editorial standards
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 49.14K
+- **Number of tokens (Llama 3)**: 21.67M
 - **Average document length (characters)**: 1303.38
 
 <!-- END-DESC-STATS -->

data/wiki/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 264502, "average_document_length": 1386.6366794958071, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 264502, "average_document_length": 1386.6366794958071, "number_of_tokens": 122003880, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/wiki/wiki.md

@@ -24,6 +24,7 @@ You can read more about wikipeadia on their [about](https://en.wikipedia.org/wik
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 264.50K
+- **Number of tokens (Llama 3)**: 122.00M
 - **Average document length (characters)**: 1386.64
 
 <!-- END-DESC-STATS -->

data/wikibooks/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 1319, "average_document_length": 13291.027293404093, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 1319, "average_document_length": 13291.027293404093, "number_of_tokens": 6239127, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/wikibooks/wikibooks.md

@@ -23,6 +23,7 @@ The Danish Subsection of [**Wikibooks**](https://www.wikibooks.org).
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 1.32K
+- **Number of tokens (Llama 3)**: 6.24M
 - **Average document length (characters)**: 13291.03
 
 <!-- END-DESC-STATS -->

data/wikisource/descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 2429, "average_document_length": 6372.487855084397, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 2429, "average_document_length": 6372.487855084397, "number_of_tokens": 5344937, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

data/wikisource/wikisource.md

@@ -22,6 +22,7 @@ The Danish subsection of [Wikisource](https://en.wikisource.org/wiki/Main_Page).
 
 - **Language**: dan, dansk, Danish
 - **Number of samples**: 2.43K
+- **Number of tokens (Llama 3)**: 5.34M
 - **Average document length (characters)**: 6372.49
 
 <!-- END-DESC-STATS -->

descriptive_stats.json

@@ -1 +1 @@
-{"number_of_samples": 546769, "average_document_length": 8461.252984349881, "language": "dan, dansk, Danish", "revision": "f2ad4e16f68745b42ccd1b8e614b6e6908d15ffc"}
+{"number_of_samples": 546769, "average_document_length": 8461.252984349881, "number_of_tokens": 1567706760, "language": "dan, dansk, Danish", "revision": "ab78b9132d5697343896be76ff8f99a6b544b74b"}

pyproject.toml

@@ -3,7 +3,7 @@ name = "danish-dynaword"
 version = "1.0.4"
 description = "project code for the danish dynaword project"
 readme = "README.md"
-requires-python = ">=3.13"
+requires-python = ">=3.12"
 dependencies = [
     "datasets>=3.0.0",
     "gitpython>=3.1.43",
@@ -15,4 +15,5 @@ dependencies = [
     "ruff>=0.8.3",
     "seaborn>=0.13.2",
     "tomlkit>=0.13.2",
+    "transformers>=4.47.1",
 ]

scripts/update_descriptive_statistics.py

@@ -10,6 +10,7 @@ Example use:
 import argparse
 import json
 import logging
+import multiprocessing
 from dataclasses import dataclass
 from pathlib import Path
 from textwrap import dedent
@@ -17,11 +18,15 @@ from typing import Self, cast
 
 from datasets import Dataset, load_dataset
 from git_utilities import check_is_ancestor, get_current_revision, get_latest_revision
+from transformers import AutoTokenizer
 
 logger = logging.getLogger(__name__)
 
 repo_path = Path(__file__).parent.parent
 
+tokenizer_name = "AI-Sweden-Models/Llama-3-8B-instruct"
+tokenizer = AutoTokenizer.from_pretrained(tokenizer_name, use_fast=True)
+
 
 def human_readable_large_int(value: int) -> str:
     thresholds = [
@@ -36,29 +41,56 @@ def human_readable_large_int(value: int) -> str:
     return str(value)
 
 
+def calculate_average_document_length(
+    dataset: Dataset, text_column: str = "text"
+) -> float:
+    texts = sum(len(t) for t in dataset[text_column])
+    return texts / len(dataset)
+
+
+def _count_tokens(batch):
+    return {
+        "token_count": [
+            len(tokens)
+            for tokens in tokenizer(batch["text"], padding=False)["input_ids"]
+        ]
+    }
+
+
+def calculate_number_of_tokens(
+    dataset: Dataset,
+    text_column: str = "text",
+) -> int:
+    token_counts = dataset.map(
+        _count_tokens,
+        batched=True,
+        batch_size=1000,
+        num_proc=multiprocessing.cpu_count(),
+    )
+    return sum(token_counts["token_count"])
+
+
 @dataclass()
 class DescriptiveStatsOverview:
     number_of_samples: int
     average_document_length: float
+    number_of_tokens: int
     language: str = "dan, dansk, Danish"
 
     @classmethod
     def from_dataset(cls, dataset: Dataset) -> Self:
         return cls(
             number_of_samples=len(dataset),
-            average_document_length=cls.calculate_average_document_length(dataset),
+            average_document_length=calculate_average_document_length(dataset),
+            number_of_tokens=calculate_number_of_tokens(dataset),
         )
 
-    @staticmethod
-    def calculate_average_document_length(dataset: Dataset) -> float:
-        texts = sum(len(t) for t in dataset["text"])
-        return texts / len(dataset)
-
     def to_markdown(self) -> str:
         format = dedent(f"""
 
         - **Language**: {self.language}
         - **Number of samples**: {human_readable_large_int(self.number_of_samples)}
+        - **Number of tokens (Llama 3)**: {human_readable_large_int(self.number_of_tokens)}
         - **Average document length (characters)**: {self.average_document_length:.2f}
 
         """)