LisaMegaWatts commited on
Commit
550b9f1
·
verified ·
1 Parent(s): 0d2bb74

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +47 -0
README.md ADDED
@@ -0,0 +1,47 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ dataset_info:
3
+ features:
4
+ - name: text
5
+ dtype: string
6
+ splits:
7
+ - name: train
8
+ num_examples: 7614
9
+ - name: validation
10
+ num_examples: 846
11
+ language:
12
+ - en
13
+ license: mit
14
+ tags:
15
+ - character-level
16
+ - philosophy
17
+ - mathematics
18
+ - julia
19
+ - scriptio-continua
20
+ - reduced-vocab
21
+ size_categories:
22
+ - 1K<n<10K
23
+ ---
24
+
25
+ # JuliaGPT Training Data
26
+
27
+ Character-level training corpus for [JuliaGPT](https://huggingface.co/LisaMegaWatts/JuliaGPT), an experimental GPT in pure Julia exploring minimal vocabularies inspired by ancient Greek *scriptio continua*.
28
+
29
+ ## Sources (ordered, not shuffled)
30
+ 1. **Aristotle - Rhetoric** (5,460 chunks) — MIT Classics
31
+ 2. **Euclid - The Elements** (3,001 chunks) — Project Gutenberg
32
+
33
+ ## Vocabulary
34
+ 28 characters: a-z + space + period. Numerals converted to English words. All other punctuation removed.
35
+ Plus BOS token = 29 vocab total.
36
+
37
+ ## Processing
38
+ - Lowercase only
39
+ - Numerals converted to words (e.g. "42" becomes "forty two")
40
+ - All punctuation removed except period
41
+ - Chunks max 256 chars, split at sentence boundaries
42
+
43
+ ## Format
44
+ One chunk per line in .
45
+
46
+ ## Usage
47
+