Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: unlicense
|
| 3 |
+
datasets:
|
| 4 |
+
- lucrbrtv/doom-e1-gameplay
|
| 5 |
+
library_name: transformers
|
| 6 |
+
---
|
| 7 |
+
This is the best checkpoint of the Inverse Dynamic Model I've trained on DOOM (1993) gameplay. It's good enought to labelize data for a world model; it can still have a lot of improvements
|
| 8 |
+
|
| 9 |
+
### Metrics
|
| 10 |
+
|
| 11 |
+
I've also calculated the balanced F-score (F1) on each labels.
|
| 12 |
+
|
| 13 |
+
eval/accuracy: 0.8335202932357788 (83,35%)
|
| 14 |
+
eval/f1_ATTACK: 0.5636902451515198
|
| 15 |
+
eval/f1_BCK: 0.5812191367149353
|
| 16 |
+
eval/f1_FWD: 0.8092130422592163
|
| 17 |
+
eval/f1_LEFT: 0.586035966873169
|
| 18 |
+
eval/f1_RIGHT: 0.6066449284553528
|
| 19 |
+
eval/f1_SPEED: 0 (I did not used this action in the training data)
|
| 20 |
+
eval/f1_TURN_L: 0.5413390398025513
|
| 21 |
+
eval/f1_TURN_R: 0.5131934285163879
|
| 22 |
+
eval/f1_USE: 0.4528301954269409
|
| 23 |
+
eval/loss: 0.37457531690597534
|
| 24 |
+
|
| 25 |
+
train/epoch: 60
|
| 26 |
+
train/loss:0.6065855598449708
|
| 27 |
+
|
| 28 |
+
### Licnece
|
| 29 |
+
|
| 30 |
+
The weights are in the public domain.
|