/
MAJOR CHANGE LOG-CURRENT WORK.txt
46 lines (29 loc) · 2.24 KB
/
MAJOR CHANGE LOG-CURRENT WORK.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
[CURENT WORK] Version 0.1
1) Three projects have been completed: Find Target (V0.07), Run From Zombie (V0.09), and AI Plays Tag (V0.1). Each will be found in their respective folders.
2) [General Work Update 4/13/24]:
2a) After a lot of fine-tuning I have learned a lot about the possibilities of the framework I created and I am confident in its capabilities. I will now be moving onto a more ambitios "Attack Zombie" project that will serve as a major mile stone to larger projects. The main reason is because up until now the agent has only had to make choises based on movement (move x, move y), and now I will be giving it more actions.
[MAJOR CHANGE LOG]
[11/17/23] : Memory upload/download implemented
[11/20/23] : Model upload/download implemented
[11/25/23] : Action Replay implemented
[11/25/23]
1) "Find Target" version 0.02 complete.
2) version 0.03 and work on "Run From Zombie" Started.
[11/28/23]
1) reduced agents actions back down to 2 actions [moveX, moveY].
2) Created a "Train From Memory" button. This directly trains the agent "episodes" number of times. The agent does not act.
3) created a new method of training. The "Record User Actions" will store user input and states into the memory buffer. Then, you can use the "Train From Memory"
button to directly train the agent from the memory (agent does not act).
[11/30/23]
1) "Run From Zombie" version 0.03 complete.
2) Re-organized GitHub project directories so that each mini-project can have it's own folder.
[12/21/23]
1) I added a "Test Agent" button that will test the actor_main model (without training or saving to memory).
2) Because of the "Test Agent" button I can now include trained actor_main model data files in the build folders to be used as examples.
[2/7/24]
1) Uploaded final "Find Target" V0.07 with a 99-100% accurate model and increasing version to V0.08.
[4/3/24]
1) Uploaded final "Run From Zombie" V0.09 with a 99% success rate.
2) Starting work on "AI Plays Tag"
[4/13/2024]
1) Uploaded final "AI Plays Tag" which uses actor models from Find Target V0.07 and Run From Zombie V0.09. It can handle multiple AI controlled players that each make calls to the uploaded models depending on what role they are in (chaser or runner).