leafspark commited on
Commit
681cc97
·
verified ·
1 Parent(s): 058659f

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +76 -0
README.md CHANGED
@@ -1,3 +1,79 @@
1
  ---
2
  license: apache-2.0
 
 
 
 
 
 
 
 
 
3
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  ---
2
  license: apache-2.0
3
+ language:
4
+ - arr
5
+ pretty_name: TrackMania Text
6
+ size_categories:
7
+ - 1M<n<10M
8
+ tags:
9
+ - driving
10
+ task_categories:
11
+ - text-classification
12
  ---
13
+
14
+ # tracktext
15
+
16
+ An experimental dataset that contains 128x64 greyscale images of TrackMania gameplay + keystrokes, designed for LLMs with 16k context or above.
17
+
18
+ Inspired by DOOM-Mistral-7b :)
19
+
20
+ ## Dataset Details
21
+
22
+ Format:
23
+ ```
24
+ {data}
25
+ [0.0, 0.0, 0.0, 0.0],
26
+ [0.0, 0.0, 0.0, 0.0],
27
+ ...
28
+ {action}
29
+ [0, 0, 0, 0]
30
+ ```
31
+
32
+ Greyscale Precision: 1 decimal
33
+ Capture rate: 6 frames per second
34
+
35
+ ### Dataset Description
36
+
37
+ - **Curated by:** leafspark
38
+ - **Language(s) (NLP):** Numbers
39
+ - **License:** apache-2.0
40
+
41
+ ## Uses
42
+
43
+ To train an LLM how to play TrackMania.
44
+
45
+ ### Direct Use
46
+
47
+ Specialized LLM
48
+
49
+ ### Out-of-Scope Use
50
+
51
+ Any regular LLM
52
+
53
+ ## Dataset Structure
54
+
55
+ Text files
56
+
57
+ ## Dataset Creation
58
+
59
+ Python script in repo
60
+
61
+ ### Curation Rationale
62
+
63
+ Self driving using an LLM? For fun
64
+
65
+ ### Source Data
66
+
67
+ TrackMania 2020 screengrabs
68
+
69
+ #### Personal and Sensitive Information
70
+
71
+ None
72
+
73
+ ## Bias, Risks, and Limitations
74
+
75
+ The resolution is not very high; there may be suboptimal results
76
+
77
+ ### Recommendations
78
+
79
+ Don't expect anything good