jkswin commited on
Commit
b565131
·
1 Parent(s): e7c1e4f

[DOCS] Commit initial readme with examples

Browse files
Files changed (1) hide show
  1. README.md +34 -0
README.md ADDED
@@ -0,0 +1,34 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ language:
3
+ - en
4
+ library_name: sentence-transformers
5
+ pipeline_tag: sentence-similarity
6
+ widget:
7
+ - text: "Marshmallon should be [MASK]. It allows a 1 card FTK."
8
+ example_title: "Banlist Speculation"
9
+ - text: "Once per [MASK]: You can toss a coin and call it."
10
+ example_title: "Time Wizard"
11
+ - text: "You [MASK] so hard on turn 2."
12
+ example_title: "Misplay"
13
+
14
+ ---
15
+
16
+ # YGOMiniLM
17
+ ![time_wiz](https://ms.yugipedia.com//thumb/7/76/TimeWizard-MRD-EN-UR-UE-25thAnniversaryEdition.png/300px-TimeWizard-MRD-EN-UR-UE-25thAnniversaryEdition.png)
18
+ [ImgSource](https://yugipedia.com/wiki/Time_Wizard)
19
+
20
+ This is a paraphrase-MiniLM-L3-v2 model that has undergone further domain specific pretraining via Masked Language Modelling.
21
+
22
+ Its intended use is for fast vector search in the domain of YuGiOh discourse.
23
+
24
+ The training data was split into two parts:
25
+ 1) A private collection of data collected from YouTube Comments:
26
+
27
+ |CREATOR|N_COMMENTS|
28
+ |-----|-----|
29
+ |thecalieffect|20,592|
30
+ |MBTYuGiOh|5439|
31
+ |MSTTV |5340|
32
+ |mkohl40|5224|
33
+
34
+ 2) The Full Database of YuGiOh cards accessed via the [YGOProDeck API](https://ygoprodeck.com/api-guide/). The `name`, `type`, `race` and `desc` fields were concatenated and delimited by `\t` to create the training examples.