Aptlantis commited on
Commit
e234e79
·
verified ·
1 Parent(s): b77b42b

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +67 -59
README.md CHANGED
@@ -1,59 +1,67 @@
1
- # HolyC Dataset (TempleOS)
2
-
3
- > **Source**: TempleOS Source Code & Terry Davis Explanations
4
- > **Task**: Code Generation / Language Modeling
5
-
6
- ## Overview
7
- This dataset is a collection of **HolyC** code snippets paired with mock "Assistant" conversations, designed to train Large Language Models (LLMs) on the unique syntax and style of Terry Davis's HolyC programming language (used in TempleOS).
8
-
9
- The dataset was constructed by matching Terry Davis's video explanations (transcribed or contextually linked) with the original source code modules from the TempleOS codebase.
10
-
11
- ## Dataset Structure
12
- The dataset consists of two JSONL files located in `datasets/HolyC/` (or the root `HolyC` directory depending on your setup).
13
-
14
- ### Files
15
- - `train.jsonl`: Training set (~10MB)
16
- - `validation.jsonl`: Validation set (~2MB)
17
-
18
- ### Schema
19
- Each line is a JSON object representing a "training example" in a ChatML-like format.
20
-
21
- | Field | Type | Description |
22
- |-------|------|-------------|
23
- | `text` | string | The raw HolyC source code block. |
24
- | `formatted` | json-object | A structured conversation object containing `messages`. |
25
-
26
- #### `formatted` Object Structure
27
- The `formatted` field is designed for fine-tuning chat models. It typically follows this pattern:
28
-
29
- ```json
30
- {
31
- "messages": [
32
- {
33
- "role": "system",
34
- "content": "You are Terry Davis, the creator of TempleOS. Write HolyC code in your unique style."
35
- },
36
- {
37
- "role": "user",
38
- "content": "Please write some HolyC code."
39
- },
40
- {
41
- "role": "assistant",
42
- "content": "<ACTUAL_HOLYC_SOURCE_CODE>"
43
- }
44
- ]
45
- }
46
- ```
47
-
48
- ## Provenance
49
- - **Original Author**: Terry A. Davis
50
- - **Collection Method**: Matching ~103 technical explanation videos with their corresponding 120,933 lines of TempleOS source code.
51
- - **Goal**: To capture the specific dialect (HolyC) which includes features like:
52
- - `U0` (Void)
53
- - `I64` (64-bit Integer)
54
- - Unparenthesized function calls
55
- - JIT compilation directives (e.g., `#help_index`)
56
- - DolDoc syntax integration
57
-
58
- ## Future Work (Planned)
59
- - **Revision History**: Extracting keystroke-level revisions from the `TempoSpage.org` and `Sheikhs Place` HTML archives to model the *process* of writing HolyC, not just the final output.
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ language:
4
+ - en
5
+ tags:
6
+ - languages
7
+ pretty_name: tos-code-with-explaination
8
+ ---
9
+ # HolyC Dataset (TempleOS)
10
+
11
+ > **Source**: TempleOS Source Code & Terry Davis Explanations
12
+ > **Task**: Code Generation / Language Modeling
13
+
14
+ ## Overview
15
+ This dataset is a collection of **HolyC** code snippets paired with mock "Assistant" conversations, designed to train Large Language Models (LLMs) on the unique syntax and style of Terry Davis's HolyC programming language (used in TempleOS).
16
+
17
+ The dataset was constructed by matching Terry Davis's video explanations (transcribed or contextually linked) with the original source code modules from the TempleOS codebase.
18
+
19
+ ## Dataset Structure
20
+ The dataset consists of two JSONL files located in `datasets/HolyC/` (or the root `HolyC` directory depending on your setup).
21
+
22
+ ### Files
23
+ - `train.jsonl`: Training set (~10MB)
24
+ - `validation.jsonl`: Validation set (~2MB)
25
+
26
+ ### Schema
27
+ Each line is a JSON object representing a "training example" in a ChatML-like format.
28
+
29
+ | Field | Type | Description |
30
+ |-------|------|-------------|
31
+ | `text` | string | The raw HolyC source code block. |
32
+ | `formatted` | json-object | A structured conversation object containing `messages`. |
33
+
34
+ #### `formatted` Object Structure
35
+ The `formatted` field is designed for fine-tuning chat models. It typically follows this pattern:
36
+
37
+ ```json
38
+ {
39
+ "messages": [
40
+ {
41
+ "role": "system",
42
+ "content": "You are Terry Davis, the creator of TempleOS. Write HolyC code in your unique style."
43
+ },
44
+ {
45
+ "role": "user",
46
+ "content": "Please write some HolyC code."
47
+ },
48
+ {
49
+ "role": "assistant",
50
+ "content": "<ACTUAL_HOLYC_SOURCE_CODE>"
51
+ }
52
+ ]
53
+ }
54
+ ```
55
+
56
+ ## Provenance
57
+ - **Original Author**: Terry A. Davis
58
+ - **Collection Method**: Matching ~103 technical explanation videos with their corresponding 120,933 lines of TempleOS source code.
59
+ - **Goal**: To capture the specific dialect (HolyC) which includes features like:
60
+ - `U0` (Void)
61
+ - `I64` (64-bit Integer)
62
+ - Unparenthesized function calls
63
+ - JIT compilation directives (e.g., `#help_index`)
64
+ - DolDoc syntax integration
65
+
66
+ ## Future Work (Planned)
67
+ - **Revision History**: Extracting keystroke-level revisions from the `TempoSpage.org` and `Sheikhs Place` HTML archives to model the *process* of writing HolyC, not just the final output.