 sp
|
3521dbe471
|
disabled evalcallback
|
5 months ago |
|
sp
|
83076beb95
|
enabled evaluation
|
5 months ago |
|
sp
|
f891eea7b8
|
set agent to deadlock when no shield action
|
5 months ago |
|
sp
|
8cbb9c4073
|
debugging prism file passing
|
5 months ago |
|
sp
|
493f66a950
|
pass prism file to handler
|
5 months ago |
|
sp
|
70c1e63107
|
refactored has_key for python3
|
5 months ago |
|
sp
|
89675dfe80
|
added argument for predefined prism file
Needed when skipping M2P
|
5 months ago |
|
sp
|
eec87050eb
|
log if shield cannot provide mask
|
8 months ago |
|
sp
|
5587144815
|
removed the eval callback
|
8 months ago |
|
sp
|
5e824757d7
|
disabled debug reenabled training
|
8 months ago |
|
sp
|
4ae274047f
|
more debugging
|
8 months ago |
|
sp
|
31d833bc57
|
disable learning
|
8 months ago |
|
sp
|
f589b10692
|
more debugging
|
8 months ago |
|
sp
|
b5dd43ca23
|
debug printout for len of shield
|
8 months ago |
|
sp
|
8a201a1c3c
|
enable export of shield
|
9 months ago |
|
sp
|
a4d360f474
|
do not allowed movement when shield is empty
|
10 months ago |
|
sp
|
dc8e4f320d
|
reintroduced learning
|
10 months ago |
|
sp
|
5c6e69ccaf
|
disable learning
This will likely be reverted or squashed away, since I am doing this to
get the synthesis times on the server...
|
10 months ago |
|
sp
|
1a3ad544b7
|
added info picked_up
|
10 months ago |
|
sp
|
59f0011614
|
add info about opened_door
|
10 months ago |
|
sp
|
3f032dd8b3
|
re-init evalCallback and save final model
Disabled eval logging to csv explicitely in the library, as this is
causing issues with the final progess*.csv
|
10 months ago |
|
sp
|
27004e0916
|
check all actions for mask
|
10 months ago |
|
sp
|
775341c3af
|
fixed typo in info callback
|
10 months ago |
|
sp
|
62fddc1e27
|
fixed bug in info callback
|
10 months ago |
|
sp
|
b9ed2ac234
|
changed some callbacks
|
10 months ago |
|
sp
|
e21a578ad8
|
fixed some issues with start parsing regex
also keep track of time to postprocess shield
|
10 months ago |
|
sp
|
24dec631aa
|
record gifs when evaluating
This currently does not use tensorboards feature, we have to leave this
as a TBD. The gifs will be left in the experiments log_dir
|
10 months ago |
|
sp
|
d7e7a2411b
|
use shield in evaluation when full shielding
|
10 months ago |
|
sp
|
2a0b75bb15
|
added helper for ShieldingConfig
|
10 months ago |
|
sp
|
028c942625
|
evaluate sb3 training
WIP: Not a 100% sure whether the masking will be used in the evaluation
|
10 months ago |
|
sp
|
a69dab422c
|
log to stdout
|
10 months ago |
|
sp
|
f68a4052f9
|
add expname suffix
|
10 months ago |
|
sp
|
c387d99a6c
|
include unique id in experiment name
|
10 months ago |
|
sp
|
b696dac5f6
|
configure logger manually to change csv filename
|
10 months ago |
|
sp
|
d2aa224bc8
|
move isodate to expname
|
10 months ago |
|
sp
|
8cbbef4006
|
ensure that exp log directory exists
|
10 months ago |
|
sp
|
703b213248
|
log to csv and tensorboard only
|
10 months ago |
|
sp
|
95a06bd0f0
|
pass nocleanup flag correctly
|
10 months ago |
|
sp
|
d194537f56
|
shortened expname
|
10 months ago |
|
sp
|
09ef1aac10
|
fixed nocleanup argument
|
10 months ago |
|
sp
|
efbad0cc27
|
log to subdirectory
|
10 months ago |
|
sp
|
5ab83b7460
|
only create ShieldHandler when necessary
this also renames camelcase variable logDir
|
10 months ago |
|
sp
|
7c39ab3b87
|
pass cleanup flag to handler
|
10 months ago |
|
sp
|
11d5c3c811
|
store shield files in local tmp dir
|
10 months ago |
|
sp
|
29799bc52f
|
resynthesize on reset is default False
|
10 months ago |
|
sp
|
7627645e63
|
changed default expname
|
10 months ago |
|
sp
|
2bcb38f6af
|
refactored training without shield
|
10 months ago |
|
sp
|
36c04f1b81
|
set sb3 device to auto
This automatically detects whether a GPU can be used for training.
|
10 months ago |
|
sp
|
71854bae01
|
init evalCallback for training with sb3
|
10 months ago |
|
sp
|
59c795348e
|
changes in sb3 rl training
- included callbacks for initial image and info plotting
- switched to CnnPolicy
- changed GRID_TO_PRISM_BINARY to environment var M2P_BINARY
|
10 months ago |
