maxgfr commited on
Commit
9ddcd59
·
verified ·
1 Parent(s): 23fa6c3

skycmd 6cbcce8c: skycmd numeric curriculum

Browse files
Files changed (3) hide show
  1. MANIFEST.json +9 -0
  2. README.md +54 -0
  3. data/numeric.parquet +3 -0
MANIFEST.json ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "commit": "6cbcce8cbbbf27e8e952e61b6cb1ed03ee1fca58",
3
+ "source_files": {
4
+ "numeric.jsonl": {
5
+ "sha256": "4d2f38a9bb9da098b6c09a197e717d0260ed9f6a26ab771457eeb5cbb827c5e7",
6
+ "rows": 100000
7
+ }
8
+ }
9
+ }
README.md ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ language:
4
+ - en
5
+ - fr
6
+ pretty_name: skycmd numeric curriculum
7
+ task_categories:
8
+ - text-classification
9
+ tags:
10
+ - skycmd
11
+ - typed-decisions
12
+ - system-one
13
+ configs:
14
+ - config_name: numeric
15
+ data_files: data/numeric.parquet
16
+ ---
17
+
18
+ # skycmd numeric curriculum
19
+
20
+ Generated questions about numbers (comparisons, ranges, units) of the skycmd numeric curriculum (SPEC-S3 §3). No third-party data.
21
+
22
+ Part of [skycmd](https://github.com/maxgfr/skycmd), tiny decision models trained from scratch in Rust that run in the
23
+ browser. Written by `sky-train s2-numeric --n 100000 (scripts/prepare-data.sh)` at commit [`6cbcce8c`](https://github.com/maxgfr/skycmd/commit/6cbcce8cbbbf27e8e952e61b6cb1ed03ee1fca58).
24
+ Licences of every set: [DATA_LICENSES.md](https://github.com/maxgfr/skycmd/blob/main/DATA_LICENSES.md).
25
+
26
+ Each row is one typed decision question (docs/SPEC.md):
27
+
28
+ | field | type | meaning |
29
+ |---|---|---|
30
+ | `src` | string | source |
31
+ | `t` | string | `choice`, `score` or `noul` |
32
+ | `ins` | string | the question |
33
+ | `opts` | list of strings | the options, rendered as text |
34
+ | `state` | string | the state the question is about |
35
+ | `target` | list of floats | a probability per option (soft or one-hot) |
36
+ | `split` | string | `train`, `val` or `test` |
37
+
38
+ ## Licence
39
+
40
+ Apache-2.0: generated by skycmd's own code, no third-party data.
41
+
42
+ ## Files
43
+
44
+ | file | rows |
45
+ |---|---:|
46
+ | `data/numeric.parquet` | 100,000 |
47
+
48
+ ```python
49
+ from datasets import load_dataset
50
+ ds = load_dataset("maxgfr/skycmd-numeric", "numeric")
51
+ ```
52
+
53
+ skycmd wrote no row with a language model: the targets come from the upstream labels (some upstream
54
+ sets carry their own teacher soft labels, see their cards), the games' oracles or our generators.
data/numeric.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c3fcc21251d88f367ebe5d59e7350a3b620a024d3fe06008015b6cc5fa947ae
3
+ size 2494454