jalf/TIR_PROJ
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Cleanup and new approach

Browse Source
This commit is contained in:
Johnny Fernandes
2026-04-26 01:50:01 +01:00
parent b031473758
commit 61f8a7db15
139 changed files with 510 additions and 16170 deletions
Download Patch File Download Diff File Expand all files Collapse all files
training/runs/expA_fresh2.log
-35
View File
@@ -1,35 +0,0 @@
Config: {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
Run dir: runs/expA_fresh2
Curriculum: 2 → 2 sheep, 2,000,000 steps/stage
[Stage n_sheep=2] training 2,000,000 steps
... [trial 1 | 2 sheep | 100,000 steps | ret(last 50)=-13.44 sr=0%]
... [trial 1 | 2 sheep | 200,000 steps | ret(last 50)=-14.60 sr=0%]
... [trial 1 | 2 sheep | 300,000 steps | ret(last 50)=-17.36 sr=0%]
... [trial 1 | 2 sheep | 400,000 steps | ret(last 50)=-17.36 sr=0%]
... [trial 1 | 2 sheep | 500,000 steps | ret(last 50)=-17.92 sr=0%]
... [trial 1 | 2 sheep | 600,000 steps | ret(last 50)=-15.65 sr=0%]
... [trial 1 | 2 sheep | 700,000 steps | ret(last 50)=-17.69 sr=2%]
... [trial 1 | 2 sheep | 800,000 steps | ret(last 50)=-14.61 sr=2%]
... [trial 1 | 2 sheep | 900,000 steps | ret(last 50)=-17.36 sr=0%]
... [trial 1 | 2 sheep | 1,000,000 steps | ret(last 50)=-17.44 sr=0%]
... [trial 1 | 2 sheep | 1,100,000 steps | ret(last 50)=-15.91 sr=2%]
... [trial 1 | 2 sheep | 1,200,000 steps | ret(last 50)=-16.08 sr=0%]
... [trial 1 | 2 sheep | 1,300,000 steps | ret(last 50)=-14.34 sr=0%]
... [trial 1 | 2 sheep | 1,400,000 steps | ret(last 50)=-17.00 sr=2%]
... [trial 1 | 2 sheep | 1,500,000 steps | ret(last 50)=-18.52 sr=0%]
... [trial 1 | 2 sheep | 1,600,000 steps | ret(last 50)=-16.68 sr=0%]
... [trial 1 | 2 sheep | 1,700,000 steps | ret(last 50)=-17.52 sr=0%]
... [trial 1 | 2 sheep | 1,800,000 steps | ret(last 50)=-17.33 sr=0%]
... [trial 1 | 2 sheep | 1,900,000 steps | ret(last 50)=-14.96 sr=2%]
... [trial 1 | 2 sheep | 2,000,000 steps | ret(last 50)=-15.59 sr=0%]
[Stage n_sheep=2] evaluating 30 eps
[Stage n_sheep=2] sr=0% mean_len=1500 mean_min_pen=13.2m mean_act=0.96
============================================================
REPLAY SUMMARY
============================================================
n_sheep=2 sr= 0% len= 1500 min_pen= 13.2m act=0.96
Total time: 10.7 min
Artefacts: runs/expA_fresh2/
training/runs/expA_fresh2/config.json
-11
View File
@@ -1,11 +0,0 @@
{
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
}
training/runs/expA_fresh2/final_model.zip
BIN
View File
Binary file not shown.
training/runs/expA_fresh2/stage_results.json
-9
View File
@@ -1,9 +0,0 @@
[
{
"n_sheep": 2,
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 13.171057415008544,
"mean_act": 0.960968065615257
}
]
training/runs/expA_fresh2/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/expB_mixed.log
-51
View File
@@ -1,51 +0,0 @@
Config: {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
Run dir: runs/expB_mixed
MIXED training: random n_sheep ∈ [1, 3], 3,000,000 total steps
[Mixed] training 3,000,000 steps
... [trial 1 | mixed | 100,000 steps | ret(last 50)=-13.68 sr=2%]
... [trial 1 | mixed | 200,000 steps | ret(last 50)=-14.08 sr=0%]
... [trial 1 | mixed | 300,000 steps | ret(last 50)=-9.80 sr=0%]
... [trial 1 | mixed | 400,000 steps | ret(last 50)=-11.20 sr=0%]
... [trial 1 | mixed | 500,000 steps | ret(last 50)=-10.61 sr=0%]
... [trial 1 | mixed | 600,000 steps | ret(last 50)=-11.19 sr=0%]
... [trial 1 | mixed | 700,000 steps | ret(last 50)=-14.22 sr=0%]
... [trial 1 | mixed | 800,000 steps | ret(last 50)=-6.31 sr=0%]
... [trial 1 | mixed | 900,000 steps | ret(last 50)=-12.68 sr=0%]
... [trial 1 | mixed | 1,000,000 steps | ret(last 50)=-11.06 sr=0%]
... [trial 1 | mixed | 1,100,000 steps | ret(last 50)=-13.39 sr=0%]
... [trial 1 | mixed | 1,200,000 steps | ret(last 50)=-14.20 sr=0%]
... [trial 1 | mixed | 1,300,000 steps | ret(last 50)=-11.33 sr=0%]
... [trial 1 | mixed | 1,400,000 steps | ret(last 50)=-10.73 sr=0%]
... [trial 1 | mixed | 1,500,000 steps | ret(last 50)=-10.91 sr=0%]
... [trial 1 | mixed | 1,600,000 steps | ret(last 50)=-10.44 sr=0%]
... [trial 1 | mixed | 1,700,000 steps | ret(last 50)=-10.56 sr=0%]
... [trial 1 | mixed | 1,800,000 steps | ret(last 50)=-15.74 sr=0%]
... [trial 1 | mixed | 1,900,000 steps | ret(last 50)=-13.46 sr=0%]
... [trial 1 | mixed | 2,000,000 steps | ret(last 50)=-9.86 sr=0%]
... [trial 1 | mixed | 2,100,000 steps | ret(last 50)=-13.07 sr=0%]
... [trial 1 | mixed | 2,200,000 steps | ret(last 50)=-9.86 sr=0%]
... [trial 1 | mixed | 2,300,000 steps | ret(last 50)=-9.73 sr=2%]
... [trial 1 | mixed | 2,400,000 steps | ret(last 50)=-12.21 sr=0%]
... [trial 1 | mixed | 2,500,000 steps | ret(last 50)=-14.27 sr=0%]
... [trial 1 | mixed | 2,600,000 steps | ret(last 50)=-10.90 sr=2%]
... [trial 1 | mixed | 2,700,000 steps | ret(last 50)=-9.67 sr=0%]
... [trial 1 | mixed | 2,800,000 steps | ret(last 50)=-14.29 sr=0%]
... [trial 1 | mixed | 2,900,000 steps | ret(last 50)=-9.08 sr=0%]
... [trial 1 | mixed | 3,000,000 steps | ret(last 50)=-11.62 sr=6%]
[Mixed] evaluating n=1, 30 eps
[Mixed] n_sheep=1 sr=0% mean_len=1500 mean_min_pen=12.1m mean_act=0.64
[Mixed] evaluating n=2, 30 eps
[Mixed] n_sheep=2 sr=0% mean_len=1500 mean_min_pen=13.6m mean_act=1.12
[Mixed] evaluating n=3, 30 eps
[Mixed] n_sheep=3 sr=0% mean_len=1500 mean_min_pen=13.3m mean_act=1.02
============================================================
REPLAY SUMMARY
============================================================
n_sheep=1 sr= 0% len= 1500 min_pen= 12.1m act=0.64
n_sheep=2 sr= 0% len= 1500 min_pen= 13.6m act=1.12
n_sheep=3 sr= 0% len= 1500 min_pen= 13.3m act=1.02
Total time: 20.6 min
Artefacts: runs/expB_mixed/
training/runs/expB_mixed/config.json
-11
View File
@@ -1,11 +0,0 @@
{
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
}
training/runs/expB_mixed/final_model.zip
BIN
View File
Binary file not shown.
training/runs/expB_mixed/stage_results.json
-23
View File
@@ -1,23 +0,0 @@
[
{
"n_sheep": 1,
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 12.136781152089437,
"mean_act": 0.6380681545449439
},
{
"n_sheep": 2,
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 13.609641806284587,
"mean_act": 1.1225489819858792
},
{
"n_sheep": 3,
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 13.337443319956462,
"mean_act": 1.0186407331574738
}
]
training/runs/expB_mixed/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/expC_clustered.log
-57
View File
@@ -1,57 +0,0 @@
Config: {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
Run dir: runs/expC_clustered
Curriculum: 1 → 3 sheep, 1,000,000 steps/stage
[Stage n_sheep=1] training 1,000,000 steps
... [trial 1 | 1 sheep | 100,000 steps | ret(last 50)=-17.04 sr=6%]
... [trial 1 | 1 sheep | 200,000 steps | ret(last 50)=-17.39 sr=4%]
... [trial 1 | 1 sheep | 300,000 steps | ret(last 50)=-15.50 sr=4%]
... [trial 1 | 1 sheep | 400,000 steps | ret(last 50)=-2.07 sr=26%]
... [trial 1 | 1 sheep | 500,000 steps | ret(last 50)=+3.81 sr=52%]
... [trial 1 | 1 sheep | 600,000 steps | ret(last 50)=+8.03 sr=76%]
... [trial 1 | 1 sheep | 700,000 steps | ret(last 50)=+9.49 sr=86%]
... [trial 1 | 1 sheep | 800,000 steps | ret(last 50)=+9.42 sr=88%]
... [trial 1 | 1 sheep | 900,000 steps | ret(last 50)=+9.49 sr=88%]
... [trial 1 | 1 sheep | 1,000,000 steps | ret(last 50)=+10.34 sr=94%]
[Stage n_sheep=1] evaluating 30 eps
[Stage n_sheep=1] sr=83% mean_len=519 mean_min_pen=3.5m mean_act=0.25
[Stage n_sheep=2] training 1,000,000 steps
... [trial 1 | 2 sheep | 1,015,816 steps | ret(last 0)=+nan sr=nan%]
... [trial 1 | 2 sheep | 1,115,816 steps | ret(last 50)=-0.13 sr=10%]
... [trial 1 | 2 sheep | 1,215,816 steps | ret(last 50)=-1.23 sr=10%]
... [trial 1 | 2 sheep | 1,315,816 steps | ret(last 50)=-0.10 sr=6%]
... [trial 1 | 2 sheep | 1,415,816 steps | ret(last 50)=+4.10 sr=28%]
... [trial 1 | 2 sheep | 1,515,816 steps | ret(last 50)=+6.24 sr=32%]
... [trial 1 | 2 sheep | 1,615,816 steps | ret(last 50)=+8.48 sr=52%]
... [trial 1 | 2 sheep | 1,715,816 steps | ret(last 50)=+14.14 sr=98%]
... [trial 1 | 2 sheep | 1,815,816 steps | ret(last 50)=+14.33 sr=98%]
... [trial 1 | 2 sheep | 1,915,816 steps | ret(last 50)=+14.02 sr=100%]
... [trial 1 | 2 sheep | 2,015,816 steps | ret(last 50)=+14.05 sr=100%]
[Stage n_sheep=2] evaluating 30 eps
[Stage n_sheep=2] sr=100% mean_len=695 mean_min_pen=3.4m mean_act=0.58
[Stage n_sheep=3] training 1,000,000 steps
... [trial 1 | 3 sheep | 2,031,624 steps | ret(last 0)=+nan sr=nan%]
... [trial 1 | 3 sheep | 2,131,624 steps | ret(last 50)=+10.43 sr=56%]
... [trial 1 | 3 sheep | 2,231,624 steps | ret(last 50)=+13.91 sr=74%]
... [trial 1 | 3 sheep | 2,331,624 steps | ret(last 50)=+13.98 sr=76%]
... [trial 1 | 3 sheep | 2,431,624 steps | ret(last 50)=+12.67 sr=68%]
... [trial 1 | 3 sheep | 2,531,624 steps | ret(last 50)=+15.79 sr=90%]
... [trial 1 | 3 sheep | 2,631,624 steps | ret(last 50)=+16.29 sr=94%]
... [trial 1 | 3 sheep | 2,731,624 steps | ret(last 50)=+15.47 sr=90%]
... [trial 1 | 3 sheep | 2,831,624 steps | ret(last 50)=+16.67 sr=96%]
... [trial 1 | 3 sheep | 2,931,624 steps | ret(last 50)=+17.50 sr=100%]
... [trial 1 | 3 sheep | 3,031,624 steps | ret(last 50)=+16.49 sr=96%]
[Stage n_sheep=3] evaluating 30 eps
[Stage n_sheep=3] sr=90% mean_len=794 mean_min_pen=3.7m mean_act=0.47
============================================================
REPLAY SUMMARY
============================================================
n_sheep=1 sr= 83% len= 519 min_pen= 3.5m act=0.25
n_sheep=2 sr=100% len= 695 min_pen= 3.4m act=0.58
n_sheep=3 sr= 90% len= 794 min_pen= 3.7m act=0.47
Total time: 15.1 min
Artefacts: runs/expC_clustered/
training/runs/expC_clustered/config.json
-11
View File
@@ -1,11 +0,0 @@
{
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
}
training/runs/expC_clustered/final_model.zip
BIN
View File
Binary file not shown.
training/runs/expC_clustered/stage_results.json
-23
View File
@@ -1,23 +0,0 @@
[
{
"n_sheep": 1,
"sr": 0.8333333333333334,
"mean_len": 518.5333333333333,
"mean_min_pen": 3.5244259238243103,
"mean_act": 0.25044742608759274
},
{
"n_sheep": 2,
"sr": 1.0,
"mean_len": 694.9,
"mean_min_pen": 3.4314632336298625,
"mean_act": 0.5796192060058971
},
{
"n_sheep": 3,
"sr": 0.9,
"mean_len": 794.1333333333333,
"mean_min_pen": 3.6645382324854534,
"mean_act": 0.46590614892287907
}
]
training/runs/expC_clustered/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/final_v2.log
-219
View File
@@ -1,219 +0,0 @@
Config: {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
Run dir: runs/final_v2
Curriculum: 1 → 10 sheep, 1,500,000 steps/stage
[Stage n_sheep=1] training 1,500,000 steps
... [trial 1 | 1 sheep | 100,000 steps | ret(last 41)=-38.49 win_sr=10% cum_sr=10%]
... [trial 1 | 1 sheep | 200,000 steps | ret(last 50)=-32.87 win_sr=8% cum_sr=9%]
... [trial 1 | 1 sheep | 300,000 steps | ret(last 50)=-33.60 win_sr=4% cum_sr=7%]
... [trial 1 | 1 sheep | 400,000 steps | ret(last 50)=-34.78 win_sr=8% cum_sr=7%]
... [trial 1 | 1 sheep | 500,000 steps | ret(last 50)=-31.25 win_sr=12% cum_sr=8%]
... [trial 1 | 1 sheep | 600,000 steps | ret(last 50)=-32.87 win_sr=2% cum_sr=7%]
... [trial 1 | 1 sheep | 700,000 steps | ret(last 50)=-33.25 win_sr=6% cum_sr=7%]
... [trial 1 | 1 sheep | 800,000 steps | ret(last 50)=-27.80 win_sr=16% cum_sr=8%]
... [trial 1 | 1 sheep | 900,000 steps | ret(last 50)=-27.44 win_sr=14% cum_sr=9%]
... [trial 1 | 1 sheep | 1,000,000 steps | ret(last 50)=-30.52 win_sr=6% cum_sr=9%]
... [trial 1 | 1 sheep | 1,100,000 steps | ret(last 50)=-24.75 win_sr=20% cum_sr=10%]
... [trial 1 | 1 sheep | 1,200,000 steps | ret(last 50)=-29.94 win_sr=4% cum_sr=10%]
... [trial 1 | 1 sheep | 1,300,000 steps | ret(last 50)=-22.72 win_sr=22% cum_sr=11%]
... [trial 1 | 1 sheep | 1,400,000 steps | ret(last 50)=-9.84 win_sr=46% cum_sr=14%]
... [trial 1 | 1 sheep | 1,500,000 steps | ret(last 50)=+10.01 win_sr=96% cum_sr=24%]
[Stage n_sheep=1] evaluating 30 eps
[Stage n_sheep=1] sr=97% mean_len=351 mean_min_pen=3.9m mean_act=0.28
[Stage n_sheep=2] training 1,500,000 steps
... [trial 1 | 2 sheep | 1,507,336 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 2 sheep | 1,607,336 steps | ret(last 43)=-4.11 win_sr=33% cum_sr=33%]
... [trial 1 | 2 sheep | 1,707,336 steps | ret(last 50)=-0.34 win_sr=36% cum_sr=34%]
... [trial 1 | 2 sheep | 1,807,336 steps | ret(last 50)=+14.73 win_sr=92% cum_sr=62%]
... [trial 1 | 2 sheep | 1,907,336 steps | ret(last 50)=+17.38 win_sr=100% cum_sr=76%]
... [trial 1 | 2 sheep | 2,007,336 steps | ret(last 50)=+16.80 win_sr=100% cum_sr=83%]
... [trial 1 | 2 sheep | 2,107,336 steps | ret(last 50)=+15.67 win_sr=100% cum_sr=87%]
... [trial 1 | 2 sheep | 2,207,336 steps | ret(last 50)=+15.39 win_sr=100% cum_sr=90%]
... [trial 1 | 2 sheep | 2,307,336 steps | ret(last 50)=+15.58 win_sr=100% cum_sr=92%]
... [trial 1 | 2 sheep | 2,407,336 steps | ret(last 50)=+15.01 win_sr=100% cum_sr=93%]
... [trial 1 | 2 sheep | 2,507,336 steps | ret(last 50)=+15.50 win_sr=100% cum_sr=94%]
... [trial 1 | 2 sheep | 2,607,336 steps | ret(last 50)=+15.21 win_sr=100% cum_sr=95%]
... [trial 1 | 2 sheep | 2,707,336 steps | ret(last 50)=+15.22 win_sr=100% cum_sr=95%]
... [trial 1 | 2 sheep | 2,807,336 steps | ret(last 50)=+15.05 win_sr=100% cum_sr=96%]
... [trial 1 | 2 sheep | 2,907,336 steps | ret(last 50)=+14.37 win_sr=100% cum_sr=96%]
... [trial 1 | 2 sheep | 3,007,336 steps | ret(last 50)=+14.70 win_sr=100% cum_sr=97%]
[Stage n_sheep=2] evaluating 30 eps
[Stage n_sheep=2] sr=100% mean_len=421 mean_min_pen=3.5m mean_act=1.01
[Stage n_sheep=3] training 1,500,000 steps
... [trial 1 | 3 sheep | 3,014,664 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 3 sheep | 3,114,664 steps | ret(last 50)=+16.52 win_sr=100% cum_sr=99%]
... [trial 1 | 3 sheep | 3,214,664 steps | ret(last 50)=+16.74 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,314,664 steps | ret(last 50)=+17.09 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,414,664 steps | ret(last 50)=+16.90 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,514,664 steps | ret(last 50)=+16.97 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,614,664 steps | ret(last 50)=+17.20 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,714,664 steps | ret(last 50)=+17.09 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,814,664 steps | ret(last 50)=+17.12 win_sr=98% cum_sr=100%]
... [trial 1 | 3 sheep | 3,914,664 steps | ret(last 50)=+17.17 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,014,664 steps | ret(last 50)=+16.25 win_sr=98% cum_sr=100%]
... [trial 1 | 3 sheep | 4,114,664 steps | ret(last 50)=+17.04 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,214,664 steps | ret(last 50)=+16.31 win_sr=98% cum_sr=100%]
... [trial 1 | 3 sheep | 4,314,664 steps | ret(last 50)=+16.82 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,414,664 steps | ret(last 50)=+16.49 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,514,664 steps | ret(last 50)=+16.54 win_sr=100% cum_sr=100%]
[Stage n_sheep=3] evaluating 30 eps
[Stage n_sheep=3] sr=100% mean_len=608 mean_min_pen=3.5m mean_act=1.06
[Stage n_sheep=4] training 1,500,000 steps
... [trial 1 | 4 sheep | 4,521,992 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 4 sheep | 4,621,992 steps | ret(last 50)=+18.55 win_sr=98% cum_sr=94%]
... [trial 1 | 4 sheep | 4,721,992 steps | ret(last 50)=+19.17 win_sr=100% cum_sr=97%]
... [trial 1 | 4 sheep | 4,821,992 steps | ret(last 50)=+18.64 win_sr=100% cum_sr=98%]
... [trial 1 | 4 sheep | 4,921,992 steps | ret(last 50)=+19.06 win_sr=100% cum_sr=99%]
... [trial 1 | 4 sheep | 5,021,992 steps | ret(last 50)=+19.01 win_sr=100% cum_sr=99%]
... [trial 1 | 4 sheep | 5,121,992 steps | ret(last 50)=+19.23 win_sr=100% cum_sr=99%]
... [trial 1 | 4 sheep | 5,221,992 steps | ret(last 50)=+18.71 win_sr=100% cum_sr=99%]
... [trial 1 | 4 sheep | 5,321,992 steps | ret(last 50)=+18.81 win_sr=100% cum_sr=99%]
... [trial 1 | 4 sheep | 5,421,992 steps | ret(last 50)=+19.51 win_sr=100% cum_sr=99%]
... [trial 1 | 4 sheep | 5,521,992 steps | ret(last 50)=+19.01 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,621,992 steps | ret(last 50)=+19.21 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,721,992 steps | ret(last 50)=+18.62 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,821,992 steps | ret(last 50)=+18.57 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,921,992 steps | ret(last 50)=+19.22 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 6,021,992 steps | ret(last 50)=+18.73 win_sr=100% cum_sr=100%]
[Stage n_sheep=4] evaluating 30 eps
[Stage n_sheep=4] sr=100% mean_len=874 mean_min_pen=3.3m mean_act=1.23
[Stage n_sheep=5] training 1,500,000 steps
... [trial 1 | 5 sheep | 6,029,320 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 5 sheep | 6,129,320 steps | ret(last 50)=+22.70 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,229,320 steps | ret(last 50)=+20.82 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,329,320 steps | ret(last 50)=+20.84 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,429,320 steps | ret(last 50)=+21.70 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,529,320 steps | ret(last 50)=+21.25 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,629,320 steps | ret(last 50)=+20.61 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,729,320 steps | ret(last 50)=+21.10 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,829,320 steps | ret(last 50)=+21.42 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,929,320 steps | ret(last 50)=+21.39 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,029,320 steps | ret(last 50)=+20.80 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,129,320 steps | ret(last 50)=+21.19 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,229,320 steps | ret(last 50)=+20.92 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,329,320 steps | ret(last 50)=+20.97 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,429,320 steps | ret(last 50)=+20.48 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,529,320 steps | ret(last 50)=+21.36 win_sr=100% cum_sr=100%]
[Stage n_sheep=5] evaluating 30 eps
[Stage n_sheep=5] sr=97% mean_len=945 mean_min_pen=3.4m mean_act=1.33
[Stage n_sheep=6] training 1,500,000 steps
... [trial 1 | 6 sheep | 7,536,648 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 6 sheep | 7,636,648 steps | ret(last 50)=+22.41 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 7,736,648 steps | ret(last 50)=+23.84 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 7,836,648 steps | ret(last 50)=+22.95 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 7,936,648 steps | ret(last 50)=+23.97 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,036,648 steps | ret(last 50)=+24.02 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,136,648 steps | ret(last 50)=+23.42 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,236,648 steps | ret(last 50)=+24.15 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,336,648 steps | ret(last 50)=+23.32 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,436,648 steps | ret(last 50)=+23.46 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,536,648 steps | ret(last 50)=+23.80 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,636,648 steps | ret(last 50)=+24.41 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,736,648 steps | ret(last 50)=+23.86 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,836,648 steps | ret(last 50)=+23.57 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,936,648 steps | ret(last 50)=+23.74 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 9,036,648 steps | ret(last 50)=+22.87 win_sr=100% cum_sr=100%]
[Stage n_sheep=6] evaluating 30 eps
[Stage n_sheep=6] sr=100% mean_len=1162 mean_min_pen=3.1m mean_act=1.36
[Stage n_sheep=7] training 1,500,000 steps
... [trial 1 | 7 sheep | 9,043,976 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 7 sheep | 9,143,976 steps | ret(last 50)=+24.46 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,243,976 steps | ret(last 50)=+25.47 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,343,976 steps | ret(last 50)=+25.10 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,443,976 steps | ret(last 50)=+24.85 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,543,976 steps | ret(last 50)=+26.01 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,643,976 steps | ret(last 50)=+26.26 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,743,976 steps | ret(last 50)=+26.44 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,843,976 steps | ret(last 50)=+26.08 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,943,976 steps | ret(last 50)=+25.00 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,043,976 steps | ret(last 50)=+26.22 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,143,976 steps | ret(last 50)=+24.79 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,243,976 steps | ret(last 50)=+26.33 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,343,976 steps | ret(last 50)=+26.36 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,443,976 steps | ret(last 50)=+25.68 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,543,976 steps | ret(last 50)=+26.75 win_sr=100% cum_sr=100%]
[Stage n_sheep=7] evaluating 30 eps
[Stage n_sheep=7] sr=100% mean_len=1253 mean_min_pen=2.7m mean_act=1.38
[Stage n_sheep=8] training 1,500,000 steps
... [trial 1 | 8 sheep | 10,551,304 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 8 sheep | 10,651,304 steps | ret(last 50)=+28.19 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 10,751,304 steps | ret(last 50)=+28.80 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 10,851,304 steps | ret(last 50)=+27.81 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 10,951,304 steps | ret(last 50)=+27.31 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,051,304 steps | ret(last 50)=+27.67 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,151,304 steps | ret(last 50)=+27.14 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,251,304 steps | ret(last 50)=+29.60 win_sr=98% cum_sr=100%]
... [trial 1 | 8 sheep | 11,351,304 steps | ret(last 50)=+28.81 win_sr=98% cum_sr=100%]
... [trial 1 | 8 sheep | 11,451,304 steps | ret(last 50)=+27.76 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,551,304 steps | ret(last 50)=+27.28 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,651,304 steps | ret(last 50)=+29.04 win_sr=98% cum_sr=99%]
... [trial 1 | 8 sheep | 11,751,304 steps | ret(last 50)=+28.75 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,851,304 steps | ret(last 50)=+29.04 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,951,304 steps | ret(last 50)=+28.27 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 12,051,304 steps | ret(last 50)=+27.90 win_sr=100% cum_sr=100%]
[Stage n_sheep=8] evaluating 30 eps
[Stage n_sheep=8] sr=93% mean_len=1495 mean_min_pen=2.6m mean_act=1.39
[Stage n_sheep=9] training 1,500,000 steps
... [trial 1 | 9 sheep | 12,058,632 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 9 sheep | 12,158,632 steps | ret(last 50)=+30.67 win_sr=98% cum_sr=98%]
... [trial 1 | 9 sheep | 12,258,632 steps | ret(last 50)=+28.78 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 12,358,632 steps | ret(last 50)=+30.08 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 12,458,632 steps | ret(last 50)=+29.61 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 12,558,632 steps | ret(last 50)=+30.34 win_sr=98% cum_sr=99%]
... [trial 1 | 9 sheep | 12,658,632 steps | ret(last 50)=+29.48 win_sr=98% cum_sr=99%]
... [trial 1 | 9 sheep | 12,758,632 steps | ret(last 50)=+29.92 win_sr=98% cum_sr=99%]
... [trial 1 | 9 sheep | 12,858,632 steps | ret(last 50)=+29.26 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 12,958,632 steps | ret(last 50)=+30.36 win_sr=96% cum_sr=98%]
... [trial 1 | 9 sheep | 13,058,632 steps | ret(last 50)=+30.19 win_sr=100% cum_sr=98%]
... [trial 1 | 9 sheep | 13,158,632 steps | ret(last 50)=+29.24 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 13,258,632 steps | ret(last 50)=+30.40 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 13,358,632 steps | ret(last 50)=+31.65 win_sr=100% cum_sr=99%]
... [trial 1 | 9 sheep | 13,458,632 steps | ret(last 50)=+30.77 win_sr=98% cum_sr=99%]
... [trial 1 | 9 sheep | 13,558,632 steps | ret(last 50)=+30.21 win_sr=94% cum_sr=98%]
[Stage n_sheep=9] evaluating 30 eps
[Stage n_sheep=9] sr=97% mean_len=1625 mean_min_pen=2.1m mean_act=1.39
[Stage n_sheep=10] training 1,500,000 steps
... [trial 1 | 10 sheep | 13,565,960 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 10 sheep | 13,665,960 steps | ret(last 50)=+30.13 win_sr=90% cum_sr=92%]
... [trial 1 | 10 sheep | 13,765,960 steps | ret(last 50)=+31.84 win_sr=96% cum_sr=92%]
... [trial 1 | 10 sheep | 13,865,960 steps | ret(last 50)=+32.66 win_sr=88% cum_sr=91%]
... [trial 1 | 10 sheep | 13,965,960 steps | ret(last 50)=+32.56 win_sr=90% cum_sr=91%]
... [trial 1 | 10 sheep | 14,065,960 steps | ret(last 50)=+31.29 win_sr=98% cum_sr=93%]
... [trial 1 | 10 sheep | 14,165,960 steps | ret(last 50)=+32.72 win_sr=94% cum_sr=93%]
... [trial 1 | 10 sheep | 14,265,960 steps | ret(last 50)=+32.42 win_sr=96% cum_sr=93%]
... [trial 1 | 10 sheep | 14,365,960 steps | ret(last 50)=+33.96 win_sr=92% cum_sr=93%]
... [trial 1 | 10 sheep | 14,465,960 steps | ret(last 50)=+33.17 win_sr=98% cum_sr=94%]
... [trial 1 | 10 sheep | 14,565,960 steps | ret(last 50)=+31.48 win_sr=96% cum_sr=94%]
... [trial 1 | 10 sheep | 14,665,960 steps | ret(last 50)=+31.19 win_sr=90% cum_sr=94%]
... [trial 1 | 10 sheep | 14,765,960 steps | ret(last 50)=+32.87 win_sr=98% cum_sr=94%]
... [trial 1 | 10 sheep | 14,865,960 steps | ret(last 50)=+32.36 win_sr=94% cum_sr=94%]
... [trial 1 | 10 sheep | 14,965,960 steps | ret(last 50)=+31.14 win_sr=94% cum_sr=94%]
... [trial 1 | 10 sheep | 15,065,960 steps | ret(last 50)=+32.18 win_sr=96% cum_sr=94%]
[Stage n_sheep=10] evaluating 30 eps
[Stage n_sheep=10] sr=97% mean_len=1816 mean_min_pen=2.0m mean_act=1.39
============================================================
REPLAY SUMMARY
============================================================
n_sheep=1 sr= 97% len= 351 min_pen= 3.9m act=0.28
n_sheep=2 sr=100% len= 421 min_pen= 3.5m act=1.01
n_sheep=3 sr=100% len= 608 min_pen= 3.5m act=1.06
n_sheep=4 sr=100% len= 874 min_pen= 3.3m act=1.23
n_sheep=5 sr= 97% len= 945 min_pen= 3.4m act=1.33
n_sheep=6 sr=100% len= 1162 min_pen= 3.1m act=1.36
n_sheep=7 sr=100% len= 1253 min_pen= 2.7m act=1.38
n_sheep=8 sr= 93% len= 1495 min_pen= 2.6m act=1.39
n_sheep=9 sr= 97% len= 1625 min_pen= 2.1m act=1.39
n_sheep=10 sr= 97% len= 1816 min_pen= 2.0m act=1.39
Total time: 90.3 min
Artefacts: runs/final_v2/
training/runs/final_v2/config.json
-11
View File
@@ -1,11 +0,0 @@
{
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
}
training/runs/final_v2/final_model.zip
BIN
View File
Binary file not shown.
training/runs/final_v2/stage_results.json
-72
View File
@@ -1,72 +0,0 @@
[
{
"n_sheep": 1,
"sr": 0.9666666666666667,
"mean_len": 350.96666666666664,
"mean_min_pen": 3.913520161310832,
"mean_act": 0.2797267940386975
},
{
"n_sheep": 2,
"sr": 1.0,
"mean_len": 421.46666666666664,
"mean_min_pen": 3.485754116376241,
"mean_act": 1.0053067604365706
},
{
"n_sheep": 3,
"sr": 1.0,
"mean_len": 608.5,
"mean_min_pen": 3.52824010848999,
"mean_act": 1.0576287743527575
},
{
"n_sheep": 4,
"sr": 1.0,
"mean_len": 874.1333333333333,
"mean_min_pen": 3.2648465514183043,
"mean_act": 1.2302308682249101
},
{
"n_sheep": 5,
"sr": 0.9666666666666667,
"mean_len": 945.1333333333333,
"mean_min_pen": 3.390091093381246,
"mean_act": 1.328577256075333
},
{
"n_sheep": 6,
"sr": 1.0,
"mean_len": 1162.1,
"mean_min_pen": 3.0996540347735086,
"mean_act": 1.3581346810990618
},
{
"n_sheep": 7,
"sr": 1.0,
"mean_len": 1252.6,
"mean_min_pen": 2.6753984689712524,
"mean_act": 1.3753795162019462
},
{
"n_sheep": 8,
"sr": 0.9333333333333333,
"mean_len": 1495.2333333333333,
"mean_min_pen": 2.560386610031128,
"mean_act": 1.3861974064434042
},
{
"n_sheep": 9,
"sr": 0.9666666666666667,
"mean_len": 1624.9,
"mean_min_pen": 2.130835851033529,
"mean_act": 1.387693840600181
},
{
"n_sheep": 10,
"sr": 0.9666666666666667,
"mean_len": 1816.5,
"mean_min_pen": 1.9940622925758362,
"mean_act": 1.3946097864970635
}
]
training/runs/final_v2/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/final_v3.log
-253
View File
@@ -1,253 +0,0 @@
Config: {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
Run dir: runs/final_v3
Curriculum: 1 → 10 sheep, 1,500,000 steps/stage
[Stage n_sheep=1] training 1,500,000 steps
... [trial 1 | 1 sheep | 100,000 steps | ret(last 40)=-28.61 win_sr=10% cum_sr=10%]
... [trial 1 | 1 sheep | 200,000 steps | ret(last 50)=-29.25 win_sr=12% cum_sr=11%]
... [trial 1 | 1 sheep | 300,000 steps | ret(last 50)=-31.55 win_sr=6% cum_sr=9%]
... [trial 1 | 1 sheep | 400,000 steps | ret(last 50)=-30.74 win_sr=10% cum_sr=9%]
... [trial 1 | 1 sheep | 500,000 steps | ret(last 50)=-32.89 win_sr=4% cum_sr=8%]
... [trial 1 | 1 sheep | 600,000 steps | ret(last 50)=-34.66 win_sr=4% cum_sr=7%]
... [trial 1 | 1 sheep | 700,000 steps | ret(last 50)=-31.44 win_sr=12% cum_sr=8%]
... [trial 1 | 1 sheep | 800,000 steps | ret(last 50)=-32.70 win_sr=6% cum_sr=8%]
... [trial 1 | 1 sheep | 900,000 steps | ret(last 50)=-35.48 win_sr=2% cum_sr=7%]
... [trial 1 | 1 sheep | 1,000,000 steps | ret(last 50)=-31.81 win_sr=10% cum_sr=8%]
... [trial 1 | 1 sheep | 1,100,000 steps | ret(last 50)=-28.53 win_sr=10% cum_sr=8%]
... [trial 1 | 1 sheep | 1,200,000 steps | ret(last 50)=-5.61 win_sr=62% cum_sr=13%]
... [trial 1 | 1 sheep | 1,300,000 steps | ret(last 50)=+11.97 win_sr=100% cum_sr=34%]
... [trial 1 | 1 sheep | 1,400,000 steps | ret(last 50)=+10.92 win_sr=96% cum_sr=50%]
... [trial 1 | 1 sheep | 1,500,000 steps | ret(last 50)=+11.97 win_sr=100% cum_sr=63%]
[Stage n_sheep=1] evaluating 30 eps
[Stage n_sheep=1] sr=100% mean_len=249 mean_min_pen=3.7m mean_act=0.41
[Stage n_sheep=2] training 1,500,000 steps
... [trial 1 | 2 sheep | 1,507,336 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 2 sheep | 1,607,336 steps | ret(last 47)=-1.11 win_sr=45% cum_sr=45%]
... [trial 1 | 2 sheep | 1,707,336 steps | ret(last 50)=-8.90 win_sr=8% cum_sr=27%]
... [trial 1 | 2 sheep | 1,807,336 steps | ret(last 50)=-5.28 win_sr=16% cum_sr=24%]
... [trial 1 | 2 sheep | 1,907,336 steps | ret(last 50)=+3.16 win_sr=58% cum_sr=33%]
... [trial 1 | 2 sheep | 2,007,336 steps | ret(last 50)=+10.26 win_sr=84% cum_sr=48%]
... [trial 1 | 2 sheep | 2,107,336 steps | ret(last 50)=+14.27 win_sr=100% cum_sr=64%]
... [trial 1 | 2 sheep | 2,207,336 steps | ret(last 50)=+14.08 win_sr=100% cum_sr=72%]
... [trial 1 | 2 sheep | 2,307,336 steps | ret(last 50)=+14.38 win_sr=100% cum_sr=77%]
... [trial 1 | 2 sheep | 2,407,336 steps | ret(last 50)=+14.27 win_sr=100% cum_sr=81%]
... [trial 1 | 2 sheep | 2,507,336 steps | ret(last 50)=+14.37 win_sr=100% cum_sr=84%]
... [trial 1 | 2 sheep | 2,607,336 steps | ret(last 50)=+14.33 win_sr=100% cum_sr=86%]
... [trial 1 | 2 sheep | 2,707,336 steps | ret(last 50)=+14.04 win_sr=100% cum_sr=87%]
... [trial 1 | 2 sheep | 2,807,336 steps | ret(last 50)=+14.25 win_sr=100% cum_sr=89%]
... [trial 1 | 2 sheep | 2,907,336 steps | ret(last 50)=+14.61 win_sr=100% cum_sr=90%]
... [trial 1 | 2 sheep | 3,007,336 steps | ret(last 50)=+13.98 win_sr=98% cum_sr=91%]
[Stage n_sheep=2] evaluating 30 eps
[Stage n_sheep=2] sr=100% mean_len=548 mean_min_pen=3.5m mean_act=0.92
[Stage n_sheep=3] training 1,500,000 steps
... [trial 1 | 3 sheep | 3,014,664 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 3 sheep | 3,114,664 steps | ret(last 50)=+16.10 win_sr=100% cum_sr=99%]
... [trial 1 | 3 sheep | 3,214,664 steps | ret(last 50)=+17.27 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,314,664 steps | ret(last 50)=+16.86 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,414,664 steps | ret(last 50)=+16.86 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,514,664 steps | ret(last 50)=+17.46 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,614,664 steps | ret(last 50)=+17.43 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,714,664 steps | ret(last 50)=+16.76 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,814,664 steps | ret(last 50)=+16.97 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 3,914,664 steps | ret(last 50)=+16.97 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,014,664 steps | ret(last 50)=+17.19 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,114,664 steps | ret(last 50)=+17.23 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,214,664 steps | ret(last 50)=+16.45 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,314,664 steps | ret(last 50)=+17.18 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,414,664 steps | ret(last 50)=+16.42 win_sr=100% cum_sr=100%]
... [trial 1 | 3 sheep | 4,514,664 steps | ret(last 50)=+16.32 win_sr=100% cum_sr=100%]
[Stage n_sheep=3] evaluating 30 eps
[Stage n_sheep=3] sr=100% mean_len=640 mean_min_pen=3.5m mean_act=1.06
[Stage n_sheep=4] training 1,500,000 steps
... [trial 1 | 4 sheep | 4,521,992 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 4 sheep | 4,621,992 steps | ret(last 50)=+18.61 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 4,721,992 steps | ret(last 50)=+18.82 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 4,821,992 steps | ret(last 50)=+18.91 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 4,921,992 steps | ret(last 50)=+18.55 win_sr=98% cum_sr=100%]
... [trial 1 | 4 sheep | 5,021,992 steps | ret(last 50)=+18.99 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,121,992 steps | ret(last 50)=+18.76 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,221,992 steps | ret(last 50)=+18.46 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,321,992 steps | ret(last 50)=+19.21 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,421,992 steps | ret(last 50)=+17.86 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,521,992 steps | ret(last 50)=+19.19 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,621,992 steps | ret(last 50)=+18.83 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,721,992 steps | ret(last 50)=+18.51 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,821,992 steps | ret(last 50)=+18.38 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 5,921,992 steps | ret(last 50)=+18.56 win_sr=100% cum_sr=100%]
... [trial 1 | 4 sheep | 6,021,992 steps | ret(last 50)=+18.82 win_sr=100% cum_sr=100%]
[Stage n_sheep=4] evaluating 30 eps
[Stage n_sheep=4] sr=100% mean_len=762 mean_min_pen=3.5m mean_act=1.26
[Stage n_sheep=5] training 1,500,000 steps
... [trial 1 | 5 sheep | 6,029,320 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 5 sheep | 6,129,320 steps | ret(last 50)=+20.46 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,229,320 steps | ret(last 50)=+20.41 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,329,320 steps | ret(last 50)=+20.58 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,429,320 steps | ret(last 50)=+21.10 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,529,320 steps | ret(last 50)=+20.48 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,629,320 steps | ret(last 50)=+20.56 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,729,320 steps | ret(last 50)=+20.51 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,829,320 steps | ret(last 50)=+20.70 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 6,929,320 steps | ret(last 50)=+20.83 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,029,320 steps | ret(last 50)=+21.52 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,129,320 steps | ret(last 50)=+21.62 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,229,320 steps | ret(last 50)=+21.22 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,329,320 steps | ret(last 50)=+21.17 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,429,320 steps | ret(last 50)=+21.00 win_sr=100% cum_sr=100%]
... [trial 1 | 5 sheep | 7,529,320 steps | ret(last 50)=+20.48 win_sr=100% cum_sr=100%]
[Stage n_sheep=5] evaluating 30 eps
[Stage n_sheep=5] sr=100% mean_len=931 mean_min_pen=3.6m mean_act=1.31
[Stage n_sheep=6] training 1,500,000 steps
... [trial 1 | 6 sheep | 7,536,648 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 6 sheep | 7,636,648 steps | ret(last 50)=+21.89 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 7,736,648 steps | ret(last 50)=+22.98 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 7,836,648 steps | ret(last 50)=+22.66 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 7,936,648 steps | ret(last 50)=+23.23 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,036,648 steps | ret(last 50)=+22.83 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,136,648 steps | ret(last 50)=+22.65 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,236,648 steps | ret(last 50)=+22.22 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,336,648 steps | ret(last 50)=+22.45 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,436,648 steps | ret(last 50)=+22.55 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,536,648 steps | ret(last 50)=+22.99 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,636,648 steps | ret(last 50)=+21.99 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,736,648 steps | ret(last 50)=+22.30 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,836,648 steps | ret(last 50)=+23.06 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 8,936,648 steps | ret(last 50)=+23.32 win_sr=100% cum_sr=100%]
... [trial 1 | 6 sheep | 9,036,648 steps | ret(last 50)=+21.80 win_sr=100% cum_sr=100%]
[Stage n_sheep=6] evaluating 30 eps
[Stage n_sheep=6] sr=100% mean_len=1082 mean_min_pen=3.6m mean_act=1.35
[Stage n_sheep=7] training 1,500,000 steps
... [trial 1 | 7 sheep | 9,043,976 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 7 sheep | 9,143,976 steps | ret(last 50)=+25.57 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,243,976 steps | ret(last 50)=+24.76 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,343,976 steps | ret(last 50)=+24.69 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,443,976 steps | ret(last 50)=+26.12 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,543,976 steps | ret(last 50)=+25.53 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,643,976 steps | ret(last 50)=+25.39 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,743,976 steps | ret(last 50)=+24.45 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,843,976 steps | ret(last 50)=+26.45 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 9,943,976 steps | ret(last 50)=+24.51 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,043,976 steps | ret(last 50)=+24.80 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,143,976 steps | ret(last 50)=+25.56 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,243,976 steps | ret(last 50)=+25.75 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,343,976 steps | ret(last 50)=+25.64 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,443,976 steps | ret(last 50)=+26.45 win_sr=100% cum_sr=100%]
... [trial 1 | 7 sheep | 10,543,976 steps | ret(last 50)=+25.19 win_sr=100% cum_sr=100%]
[Stage n_sheep=7] evaluating 30 eps
[Stage n_sheep=7] sr=100% mean_len=1081 mean_min_pen=3.5m mean_act=1.37
[Stage n_sheep=8] training 1,500,000 steps
... [trial 1 | 8 sheep | 10,551,304 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 8 sheep | 10,651,304 steps | ret(last 50)=+26.63 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 10,751,304 steps | ret(last 50)=+27.63 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 10,851,304 steps | ret(last 50)=+27.53 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 10,951,304 steps | ret(last 50)=+27.43 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,051,304 steps | ret(last 50)=+27.70 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,151,304 steps | ret(last 50)=+26.53 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,251,304 steps | ret(last 50)=+27.24 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,351,304 steps | ret(last 50)=+27.14 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,451,304 steps | ret(last 50)=+27.43 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,551,304 steps | ret(last 50)=+27.25 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,651,304 steps | ret(last 50)=+27.40 win_sr=98% cum_sr=100%]
... [trial 1 | 8 sheep | 11,751,304 steps | ret(last 50)=+27.35 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,851,304 steps | ret(last 50)=+26.33 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 11,951,304 steps | ret(last 50)=+26.89 win_sr=100% cum_sr=100%]
... [trial 1 | 8 sheep | 12,051,304 steps | ret(last 50)=+27.86 win_sr=100% cum_sr=100%]
[Stage n_sheep=8] evaluating 30 eps
[Stage n_sheep=8] sr=100% mean_len=1311 mean_min_pen=3.5m mean_act=1.38
[Stage n_sheep=9] training 1,500,000 steps
... [trial 1 | 9 sheep | 12,058,632 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 9 sheep | 12,158,632 steps | ret(last 50)=+29.62 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,258,632 steps | ret(last 50)=+31.32 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,358,632 steps | ret(last 50)=+30.30 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,458,632 steps | ret(last 50)=+29.33 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,558,632 steps | ret(last 50)=+28.83 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,658,632 steps | ret(last 50)=+29.02 win_sr=98% cum_sr=100%]
... [trial 1 | 9 sheep | 12,758,632 steps | ret(last 50)=+29.60 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,858,632 steps | ret(last 50)=+29.88 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 12,958,632 steps | ret(last 50)=+30.12 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 13,058,632 steps | ret(last 50)=+28.80 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 13,158,632 steps | ret(last 50)=+30.33 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 13,258,632 steps | ret(last 50)=+27.85 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 13,358,632 steps | ret(last 50)=+28.21 win_sr=96% cum_sr=100%]
... [trial 1 | 9 sheep | 13,458,632 steps | ret(last 50)=+29.88 win_sr=100% cum_sr=100%]
... [trial 1 | 9 sheep | 13,558,632 steps | ret(last 50)=+29.06 win_sr=98% cum_sr=100%]
[Stage n_sheep=9] evaluating 30 eps
[Stage n_sheep=9] sr=100% mean_len=1435 mean_min_pen=3.6m mean_act=1.39
[Stage n_sheep=10] training 1,500,000 steps
... [trial 1 | 10 sheep | 13,565,960 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | 10 sheep | 13,665,960 steps | ret(last 50)=+30.42 win_sr=96% cum_sr=96%]
... [trial 1 | 10 sheep | 13,765,960 steps | ret(last 50)=+29.97 win_sr=92% cum_sr=95%]
... [trial 1 | 10 sheep | 13,865,960 steps | ret(last 50)=+30.45 win_sr=82% cum_sr=90%]
... [trial 1 | 10 sheep | 13,965,960 steps | ret(last 50)=+29.82 win_sr=90% cum_sr=91%]
... [trial 1 | 10 sheep | 14,065,960 steps | ret(last 50)=+29.66 win_sr=90% cum_sr=91%]
... [trial 1 | 10 sheep | 14,165,960 steps | ret(last 50)=+31.57 win_sr=98% cum_sr=92%]
... [trial 1 | 10 sheep | 14,265,960 steps | ret(last 50)=+31.71 win_sr=96% cum_sr=93%]
... [trial 1 | 10 sheep | 14,365,960 steps | ret(last 50)=+31.75 win_sr=94% cum_sr=93%]
... [trial 1 | 10 sheep | 14,465,960 steps | ret(last 50)=+29.46 win_sr=88% cum_sr=93%]
... [trial 1 | 10 sheep | 14,565,960 steps | ret(last 50)=+29.62 win_sr=94% cum_sr=93%]
... [trial 1 | 10 sheep | 14,665,960 steps | ret(last 50)=+31.64 win_sr=98% cum_sr=93%]
... [trial 1 | 10 sheep | 14,765,960 steps | ret(last 50)=+30.86 win_sr=90% cum_sr=93%]
... [trial 1 | 10 sheep | 14,865,960 steps | ret(last 50)=+31.65 win_sr=90% cum_sr=93%]
... [trial 1 | 10 sheep | 14,965,960 steps | ret(last 50)=+31.75 win_sr=92% cum_sr=93%]
... [trial 1 | 10 sheep | 15,065,960 steps | ret(last 50)=+30.24 win_sr=100% cum_sr=93%]
[Stage n_sheep=10] evaluating 30 eps
[Stage n_sheep=10] sr=90% mean_len=1841 mean_min_pen=3.6m mean_act=1.39
[Consolidation] mixed n_sheep ∈ [1, 10], 2,000,000 steps
... [trial 1 | consolidate | 15,073,288 steps | ret(last 0)=+nan win_sr=nan% cum_sr=nan%]
... [trial 1 | consolidate | 15,173,288 steps | ret(last 50)=+20.69 win_sr=94% cum_sr=95%]
... [trial 1 | consolidate | 15,273,288 steps | ret(last 50)=+20.62 win_sr=90% cum_sr=92%]
... [trial 1 | consolidate | 15,373,288 steps | ret(last 50)=+20.25 win_sr=94% cum_sr=93%]
... [trial 1 | consolidate | 15,473,288 steps | ret(last 50)=+19.82 win_sr=96% cum_sr=94%]
... [trial 1 | consolidate | 15,573,288 steps | ret(last 50)=+20.56 win_sr=94% cum_sr=94%]
... [trial 1 | consolidate | 15,673,288 steps | ret(last 50)=+20.56 win_sr=92% cum_sr=94%]
... [trial 1 | consolidate | 15,773,288 steps | ret(last 50)=+19.43 win_sr=94% cum_sr=95%]
... [trial 1 | consolidate | 15,873,288 steps | ret(last 50)=+21.85 win_sr=98% cum_sr=95%]
... [trial 1 | consolidate | 15,973,288 steps | ret(last 50)=+21.84 win_sr=94% cum_sr=95%]
... [trial 1 | consolidate | 16,073,288 steps | ret(last 50)=+22.13 win_sr=98% cum_sr=95%]
... [trial 1 | consolidate | 16,173,288 steps | ret(last 50)=+21.89 win_sr=94% cum_sr=95%]
... [trial 1 | consolidate | 16,273,288 steps | ret(last 50)=+21.88 win_sr=98% cum_sr=95%]
... [trial 1 | consolidate | 16,373,288 steps | ret(last 50)=+20.81 win_sr=94% cum_sr=95%]
... [trial 1 | consolidate | 16,473,288 steps | ret(last 50)=+20.91 win_sr=98% cum_sr=95%]
... [trial 1 | consolidate | 16,573,288 steps | ret(last 50)=+21.13 win_sr=98% cum_sr=95%]
... [trial 1 | consolidate | 16,673,288 steps | ret(last 50)=+19.85 win_sr=100% cum_sr=95%]
... [trial 1 | consolidate | 16,773,288 steps | ret(last 50)=+22.30 win_sr=92% cum_sr=95%]
... [trial 1 | consolidate | 16,873,288 steps | ret(last 50)=+20.61 win_sr=96% cum_sr=95%]
... [trial 1 | consolidate | 16,973,288 steps | ret(last 50)=+21.93 win_sr=98% cum_sr=96%]
... [trial 1 | consolidate | 17,073,288 steps | ret(last 50)=+21.86 win_sr=98% cum_sr=96%]
[Consolidation] re-evaluating all sheep counts
[Consolidation] n_sheep=1 sr=97% mean_len=377 mean_min_pen=3.5m mean_act=1.39
[Consolidation] n_sheep=2 sr=47% mean_len=1718 mean_min_pen=2.4m mean_act=1.39
[Consolidation] n_sheep=3 sr=93% mean_len=970 mean_min_pen=3.2m mean_act=1.39
[Consolidation] n_sheep=4 sr=97% mean_len=1008 mean_min_pen=3.3m mean_act=1.39
[Consolidation] n_sheep=5 sr=100% mean_len=1176 mean_min_pen=3.3m mean_act=1.39
[Consolidation] n_sheep=6 sr=100% mean_len=1305 mean_min_pen=3.3m mean_act=1.39
[Consolidation] n_sheep=7 sr=100% mean_len=1300 mean_min_pen=3.4m mean_act=1.39
[Consolidation] n_sheep=8 sr=100% mean_len=1461 mean_min_pen=3.5m mean_act=1.39
[Consolidation] n_sheep=9 sr=87% mean_len=1607 mean_min_pen=3.8m mean_act=1.39
[Consolidation] n_sheep=10 sr=80% mean_len=1801 mean_min_pen=3.7m mean_act=1.39
============================================================
REPLAY SUMMARY
============================================================
n_sheep=1 sr= 97% len= 377 min_pen= 3.5m act=1.39
n_sheep=2 sr= 47% len= 1718 min_pen= 2.4m act=1.39
n_sheep=3 sr= 93% len= 970 min_pen= 3.2m act=1.39
n_sheep=4 sr= 97% len= 1008 min_pen= 3.3m act=1.39
n_sheep=5 sr=100% len= 1176 min_pen= 3.3m act=1.39
n_sheep=6 sr=100% len= 1305 min_pen= 3.3m act=1.39
n_sheep=7 sr=100% len= 1300 min_pen= 3.4m act=1.39
n_sheep=8 sr=100% len= 1461 min_pen= 3.5m act=1.39
n_sheep=9 sr= 87% len= 1607 min_pen= 3.8m act=1.39
n_sheep=10 sr= 80% len= 1801 min_pen= 3.7m act=1.39
Total time: 110.1 min
Artefacts: runs/final_v3/
training/runs/final_v3/config.json
-11
View File
@@ -1,11 +0,0 @@
{
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
}
training/runs/final_v3/final_model.zip
BIN
View File
Binary file not shown.
training/runs/final_v3/stage_results.json
-72
View File
@@ -1,72 +0,0 @@
[
{
"n_sheep": 1,
"sr": 0.9666666666666667,
"mean_len": 377.3666666666667,
"mean_min_pen": 3.5389957586924234,
"mean_act": 1.3908841227086732
},
{
"n_sheep": 2,
"sr": 0.4666666666666667,
"mean_len": 1717.6333333333334,
"mean_min_pen": 2.4164488633473713,
"mean_act": 1.3922284740020803
},
{
"n_sheep": 3,
"sr": 0.9333333333333333,
"mean_len": 970.2666666666667,
"mean_min_pen": 3.203955141703288,
"mean_act": 1.3945290882248416
},
{
"n_sheep": 4,
"sr": 0.9666666666666667,
"mean_len": 1008.0,
"mean_min_pen": 3.279213563601176,
"mean_act": 1.3918021049325862
},
{
"n_sheep": 5,
"sr": 1.0,
"mean_len": 1175.8666666666666,
"mean_min_pen": 3.3209743976593016,
"mean_act": 1.3925684957666513
},
{
"n_sheep": 6,
"sr": 1.0,
"mean_len": 1305.0,
"mean_min_pen": 3.312229561805725,
"mean_act": 1.391130207932886
},
{
"n_sheep": 7,
"sr": 1.0,
"mean_len": 1300.0,
"mean_min_pen": 3.363971138000488,
"mean_act": 1.392986050516367
},
{
"n_sheep": 8,
"sr": 1.0,
"mean_len": 1461.3666666666666,
"mean_min_pen": 3.4741388003031415,
"mean_act": 1.392040583461347
},
{
"n_sheep": 9,
"sr": 0.8666666666666667,
"mean_len": 1606.7333333333333,
"mean_min_pen": 3.835897175470988,
"mean_act": 1.3907199496534952
},
{
"n_sheep": 10,
"sr": 0.8,
"mean_len": 1800.9666666666667,
"mean_min_pen": 3.741190282503764,
"mean_act": 1.392501896076031
}
]
training/runs/final_v3/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/ppo_debug.log
-5569
View File
File diff suppressed because it is too large Load Diff
training/runs/ppo_debug/best_model/best_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_debug/evaluations.npz
BIN
View File
Binary file not shown.
training/runs/ppo_debug/final_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_debug/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check.log
-3388
View File
File diff suppressed because it is too large Load Diff
training/runs/ppo_fix_check/best_model/best_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check/evaluations.npz
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check/final_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check2.log
-3391
View File
File diff suppressed because it is too large Load Diff
training/runs/ppo_fix_check2/best_model/best_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check2/evaluations.npz
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check2/final_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_fix_check2/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/ppo_v2/best_model/best_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_v2/evaluations.npz
BIN
View File
Binary file not shown.
training/runs/ppo_v2/final_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_v2/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/ppo_v2_cont/best_model/best_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_v2_cont/evaluations.npz
BIN
View File
Binary file not shown.
training/runs/ppo_v2_cont/final_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_v2_cont/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/ppo_v3/best_model/best_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_v3/evaluations.npz
BIN
View File
Binary file not shown.
training/runs/ppo_v3/final_model.zip
BIN
View File
Binary file not shown.
training/runs/ppo_v3/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/replay_20260425_152857/config.json
-11
View File
@@ -1,11 +0,0 @@
{
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
}
training/runs/replay_20260425_152857/final_model.zip
BIN
View File
Binary file not shown.
training/runs/replay_20260425_152857/stage_results.json
-23
View File
@@ -1,23 +0,0 @@
[
{
"n_sheep": 1,
"sr": 1.0,
"mean_len": 267.6333333333333,
"mean_min_pen": 3.7235233147939044,
"mean_act": 0.3746675180125346
},
{
"n_sheep": 2,
"sr": 0.06666666666666667,
"mean_len": 1458.6666666666667,
"mean_min_pen": 14.14484707514445,
"mean_act": 0.284232099657656
},
{
"n_sheep": 3,
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 12.514182837804158,
"mean_act": 1.2590703022670828
}
]
training/runs/replay_20260425_152857/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/replay_best.log
-72
View File
@@ -1,72 +0,0 @@
Config: {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
Run dir: runs/replay_20260425_152857
Curriculum: 1 → 3 sheep, 1,500,000 steps/stage
[Stage n_sheep=1] training 1,500,000 steps
... [trial 1 | 1 sheep | 100,000 steps | ret(last 50)=-20.83 sr=6%]
... [trial 1 | 1 sheep | 200,000 steps | ret(last 50)=-21.40 sr=4%]
... [trial 1 | 1 sheep | 300,000 steps | ret(last 50)=-22.31 sr=0%]
... [trial 1 | 1 sheep | 400,000 steps | ret(last 50)=-19.13 sr=4%]
... [trial 1 | 1 sheep | 500,000 steps | ret(last 50)=-18.79 sr=8%]
... [trial 1 | 1 sheep | 600,000 steps | ret(last 50)=-10.15 sr=8%]
... [trial 1 | 1 sheep | 700,000 steps | ret(last 50)=+10.14 sr=82%]
... [trial 1 | 1 sheep | 800,000 steps | ret(last 50)=+11.90 sr=100%]
... [trial 1 | 1 sheep | 900,000 steps | ret(last 50)=+11.32 sr=100%]
... [trial 1 | 1 sheep | 1,000,000 steps | ret(last 50)=+11.36 sr=100%]
... [trial 1 | 1 sheep | 1,100,000 steps | ret(last 50)=+11.18 sr=100%]
... [trial 1 | 1 sheep | 1,200,000 steps | ret(last 50)=+11.08 sr=100%]
... [trial 1 | 1 sheep | 1,300,000 steps | ret(last 50)=+11.14 sr=100%]
... [trial 1 | 1 sheep | 1,400,000 steps | ret(last 50)=+11.10 sr=100%]
... [trial 1 | 1 sheep | 1,500,000 steps | ret(last 50)=+10.99 sr=100%]
[Stage n_sheep=1] evaluating 30 eps
[Stage n_sheep=1] sr=100% mean_len=268 mean_min_pen=3.7m mean_act=0.37
[Stage n_sheep=2] training 1,500,000 steps
... [trial 1 | 2 sheep | 1,507,336 steps | ret(last 0)=+nan sr=nan%]
... [trial 1 | 2 sheep | 1,607,336 steps | ret(last 50)=-3.10 sr=2%]
... [trial 1 | 2 sheep | 1,707,336 steps | ret(last 50)=-3.41 sr=2%]
... [trial 1 | 2 sheep | 1,807,336 steps | ret(last 50)=-3.11 sr=6%]
... [trial 1 | 2 sheep | 1,907,336 steps | ret(last 50)=-2.65 sr=8%]
... [trial 1 | 2 sheep | 2,007,336 steps | ret(last 50)=-4.11 sr=2%]
... [trial 1 | 2 sheep | 2,107,336 steps | ret(last 50)=-3.19 sr=6%]
... [trial 1 | 2 sheep | 2,207,336 steps | ret(last 50)=-3.45 sr=4%]
... [trial 1 | 2 sheep | 2,307,336 steps | ret(last 50)=-4.13 sr=0%]
... [trial 1 | 2 sheep | 2,407,336 steps | ret(last 50)=-3.47 sr=8%]
... [trial 1 | 2 sheep | 2,507,336 steps | ret(last 50)=-3.83 sr=4%]
... [trial 1 | 2 sheep | 2,607,336 steps | ret(last 50)=-4.58 sr=0%]
... [trial 1 | 2 sheep | 2,707,336 steps | ret(last 50)=-3.94 sr=2%]
... [trial 1 | 2 sheep | 2,807,336 steps | ret(last 50)=-4.15 sr=2%]
... [trial 1 | 2 sheep | 2,907,336 steps | ret(last 50)=-3.95 sr=4%]
... [trial 1 | 2 sheep | 3,007,336 steps | ret(last 50)=-4.44 sr=0%]
[Stage n_sheep=2] evaluating 30 eps
[Stage n_sheep=2] sr=7% mean_len=1459 mean_min_pen=14.1m mean_act=0.28
[Stage n_sheep=3] training 1,500,000 steps
... [trial 1 | 3 sheep | 3,014,664 steps | ret(last 0)=+nan sr=nan%]
... [trial 1 | 3 sheep | 3,114,664 steps | ret(last 50)=-4.16 sr=0%]
... [trial 1 | 3 sheep | 3,214,664 steps | ret(last 50)=-4.94 sr=0%]
... [trial 1 | 3 sheep | 3,314,664 steps | ret(last 50)=-4.42 sr=0%]
... [trial 1 | 3 sheep | 3,414,664 steps | ret(last 50)=-4.69 sr=0%]
... [trial 1 | 3 sheep | 3,514,664 steps | ret(last 50)=-3.72 sr=0%]
... [trial 1 | 3 sheep | 3,614,664 steps | ret(last 50)=-5.04 sr=0%]
... [trial 1 | 3 sheep | 3,714,664 steps | ret(last 50)=-4.26 sr=0%]
... [trial 1 | 3 sheep | 3,814,664 steps | ret(last 50)=-4.70 sr=0%]
... [trial 1 | 3 sheep | 3,914,664 steps | ret(last 50)=-4.61 sr=0%]
... [trial 1 | 3 sheep | 4,014,664 steps | ret(last 50)=-4.19 sr=0%]
... [trial 1 | 3 sheep | 4,114,664 steps | ret(last 50)=-4.35 sr=0%]
... [trial 1 | 3 sheep | 4,214,664 steps | ret(last 50)=-4.41 sr=0%]
... [trial 1 | 3 sheep | 4,314,664 steps | ret(last 50)=-4.42 sr=0%]
... [trial 1 | 3 sheep | 4,414,664 steps | ret(last 50)=-4.77 sr=0%]
... [trial 1 | 3 sheep | 4,514,664 steps | ret(last 50)=-4.49 sr=0%]
[Stage n_sheep=3] evaluating 30 eps
[Stage n_sheep=3] sr=0% mean_len=1500 mean_min_pen=12.5m mean_act=1.26
============================================================
REPLAY SUMMARY
============================================================
n_sheep=1 sr=100% len= 268 min_pen= 3.7m act=0.37
n_sheep=2 sr= 7% len= 1459 min_pen= 14.1m act=0.28
n_sheep=3 sr= 0% len= 1500 min_pen= 12.5m act=1.26
Total time: 26.9 min
Artefacts: runs/replay_20260425_152857/
training/runs/smoke_stage1/model.zip
BIN
View File
Binary file not shown.
training/runs/smoke_stage1/timeseries.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 55 KiB

training/runs/smoke_stage1/trajectory.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 28 KiB

training/runs/smoke_stage1/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/smoke_stage2/model.zip
BIN
View File
Binary file not shown.
training/runs/smoke_stage2/timeseries.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 74 KiB

training/runs/smoke_stage2/trajectory.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 41 KiB

training/runs/smoke_stage2/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/smoke_stage3/model.zip
BIN
View File
Binary file not shown.
training/runs/smoke_stage3/timeseries.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 93 KiB

training/runs/smoke_stage3/trajectory.png
BIN
View File
Binary file not shown.
Side by Side

Before

Width:  |  Height:  |  Size: 40 KiB

training/runs/smoke_stage3/vecnorm.pkl
BIN
View File
Binary file not shown.
training/runs/sweep_20260425_124021/best.json
-41
View File
@@ -1,41 +0,0 @@
{
"trial": 0,
"config": {
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.1,
"W_PEN_BONUS": 10.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 100.0,
"W_COMPACT": 3.0,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.005
},
"score": 0.06,
"sr": {
"1": 0.3,
"2": 0.0,
"3": 0.0
},
"details": {
"1": {
"sr": 0.3,
"mean_len": 1252.2,
"mean_min_pen": 2.1085331559181215,
"mean_act": 0.07743233270979732
},
"2": {
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 12.107558453083039,
"mean_act": 0.15608626089841424
},
"3": {
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 13.675278377532958,
"mean_act": 0.10535904271739319
}
},
"elapsed_s": 307.773992061615
}
training/runs/sweep_20260425_124021/results.jsonl
-1
View File
@@ -1 +0,0 @@
{"trial": 0, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.02, "W_COMPLETE": 100.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.005}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1252.2, "mean_min_pen": 2.1085331559181215, "mean_act": 0.07743233270979732}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.107558453083039, "mean_act": 0.15608626089841424}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.675278377532958, "mean_act": 0.10535904271739319}}, "elapsed_s": 307.773992061615}
training/runs/sweep_20260425_124630/best.json
-41
View File
@@ -1,41 +0,0 @@
{
"trial": 13,
"config": {
"W_PER_SHEEP": 1.0,
"W_ALIGN": 0.0,
"W_PEN_BONUS": 5.0,
"W_STEP_COST": 0.02,
"W_COMPLETE": 200.0,
"W_COMPACT": 1.5,
"ALIGN_SHAPE": "standoff",
"ALIGN_GATED": false,
"ent_coef": 0.02
},
"score": 0.35,
"sr": {
"1": 1.0,
"2": 0.3,
"3": 0.0
},
"details": {
"1": {
"sr": 1.0,
"mean_len": 428.9,
"mean_min_pen": 3.731236696243286,
"mean_act": 0.33429858573849425
},
"2": {
"sr": 0.3,
"mean_len": 1242.7,
"mean_min_pen": 8.937442195415496,
"mean_act": 0.3998076917437125
},
"3": {
"sr": 0.0,
"mean_len": 1500.0,
"mean_min_pen": 14.061083602905274,
"mean_act": 0.5966902794524755
}
},
"elapsed_s": 313.8281009197235
}
training/runs/sweep_20260425_124630/results.jsonl
-25
View File
@@ -1,25 +0,0 @@
{"trial": 0, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.02, "W_COMPLETE": 100.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.005}, "score": 0.15000000000000002, "sr": {"1": 0.5, "2": 0.1, "3": 0.0}, "details": {"1": {"sr": 0.5, "mean_len": 1051.6, "mean_min_pen": 3.0551586985588073, "mean_act": 0.0887192903536989}, "2": {"sr": 0.1, "mean_len": 1438.1, "mean_min_pen": 10.993862140178681, "mean_act": 0.1723056222816755}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 11.92835488319397, "mean_act": 0.15403316749989074}}, "elapsed_s": 316.9084241390228}
{"trial": 1, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.05, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.05, "W_COMPLETE": 200.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.005}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1153.8, "mean_min_pen": 3.8145030617713926, "mean_act": 0.15146865127462797}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.058024168014526, "mean_act": 0.10904584494279744}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.5988187789917, "mean_act": 0.09578829008591905}}, "elapsed_s": 310.8732409477234}
{"trial": 2, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.025, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.02, "W_COMPLETE": 50.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.01}, "score": 0.27, "sr": {"1": 0.7, "2": 0.2, "3": 0.1}, "details": {"1": {"sr": 0.7, "mean_len": 772.1, "mean_min_pen": 2.92204372882843, "mean_act": 0.1583604314471399}, "2": {"sr": 0.2, "mean_len": 1390.6, "mean_min_pen": 12.992859578132629, "mean_act": 0.16090679360424953}, "3": {"sr": 0.1, "mean_len": 1403.7, "mean_min_pen": 13.045468378067017, "mean_act": 0.07991531561051667}}, "elapsed_s": 303.7708294391632}
{"trial": 3, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.05, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.02, "W_COMPLETE": 50.0, "W_COMPACT": 0.0, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.005}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1137.5, "mean_min_pen": 2.1229824781417848, "mean_act": 0.08172097406143335}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 11.521494126319885, "mean_act": 0.16864279503144788}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.317158126831055, "mean_act": 0.05537428615499472}}, "elapsed_s": 301.6172459125519}
{"trial": 4, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.02, "W_COMPLETE": 50.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "near", "ALIGN_GATED": true, "ent_coef": 0.005}, "score": 0.2, "sr": {"1": 1.0, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 1.0, "mean_len": 567.0, "mean_min_pen": 3.2795117855072022, "mean_act": 0.1855437107780058}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 9.976170372962951, "mean_act": 0.2074074002778701}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.89306182861328, "mean_act": 0.21666522849385267}}, "elapsed_s": 313.525591135025}
{"trial": 5, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.025, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.05, "W_COMPLETE": 200.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": true, "ent_coef": 0.01}, "score": 0.16000000000000003, "sr": {"1": 0.8, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.8, "mean_len": 675.5, "mean_min_pen": 3.1338732481002807, "mean_act": 0.11691584614814514}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 9.693846690654755, "mean_act": 0.19984676872865814}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.684805488586425, "mean_act": 0.06430307933471292}}, "elapsed_s": 312.4476580619812}
{"trial": 6, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.0, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.005, "W_COMPLETE": 200.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.01}, "score": 0.08000000000000002, "sr": {"1": 0.4, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.4, "mean_len": 1343.9, "mean_min_pen": 4.092962062358856, "mean_act": 0.07675616785431166}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.157618689537049, "mean_act": 0.13906600509098352}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.079688358306885, "mean_act": 0.07073271389845953}}, "elapsed_s": 337.7615342140198}
{"trial": 7, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.025, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.05, "W_COMPLETE": 100.0, "W_COMPACT": 0.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.005}, "score": 0.11, "sr": {"1": 0.3, "2": 0.1, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1177.5, "mean_min_pen": 2.261639392375946, "mean_act": 0.11013885321646562}, "2": {"sr": 0.1, "mean_len": 1437.5, "mean_min_pen": 5.9263048529624935, "mean_act": 0.16420815230170227}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.130784749984741, "mean_act": 0.20303070502222206}}, "elapsed_s": 451.2424490451813}
{"trial": 8, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.0, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.02, "W_COMPLETE": 50.0, "W_COMPACT": 0.0, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.05}, "score": 0.19, "sr": {"1": 0.7, "2": 0.1, "3": 0.0}, "details": {"1": {"sr": 0.7, "mean_len": 874.2, "mean_min_pen": 4.152815592288971, "mean_act": 0.1303976929043709}, "2": {"sr": 0.1, "mean_len": 1381.4, "mean_min_pen": 12.115124177932739, "mean_act": 0.3749806733317197}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.596546864509582, "mean_act": 0.10082290474528718}}, "elapsed_s": 349.3926422595978}
{"trial": 9, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.0, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.02, "W_COMPLETE": 200.0, "W_COMPACT": 0.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.05}, "score": 0.0, "sr": {"1": 0.0, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 8.404254817962647, "mean_act": 0.6749623541596586}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 11.970247220993041, "mean_act": 0.45562502020561796}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.029277420043945, "mean_act": 0.1599790089856222}}, "elapsed_s": 319.38924622535706}
{"trial": 10, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.02, "W_COMPLETE": 200.0, "W_COMPACT": 0.5, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.02}, "score": 0.16000000000000003, "sr": {"1": 0.8, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.8, "mean_len": 690.7, "mean_min_pen": 3.1264367938041686, "mean_act": 0.13493279961414406}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.040377330780029, "mean_act": 0.20203861368317985}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.379706478118896, "mean_act": 0.05979441475490263}}, "elapsed_s": 310.1806254386902}
{"trial": 11, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.05, "W_COMPLETE": 50.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.02}, "score": 0.24, "sr": {"1": 0.7, "2": 0.2, "3": 0.0}, "details": {"1": {"sr": 0.7, "mean_len": 727.5, "mean_min_pen": 2.933144009113312, "mean_act": 0.11888058594495643}, "2": {"sr": 0.2, "mean_len": 1317.8, "mean_min_pen": 10.2599928855896, "mean_act": 0.14370172662258304}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.231103086471558, "mean_act": 0.0614644922383149}}, "elapsed_s": 330.0620620250702}
{"trial": 12, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.05, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.05, "W_COMPLETE": 100.0, "W_COMPACT": 0.5, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.005}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1244.8, "mean_min_pen": 2.1193889737129212, "mean_act": 0.08216679023110932}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 10.745809042453766, "mean_act": 0.16497857472260813}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.016976690292358, "mean_act": 0.09897869050660908}}, "elapsed_s": 323.27931213378906}
{"trial": 13, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.0, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.02, "W_COMPLETE": 200.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.02}, "score": 0.35, "sr": {"1": 1.0, "2": 0.3, "3": 0.0}, "details": {"1": {"sr": 1.0, "mean_len": 428.9, "mean_min_pen": 3.731236696243286, "mean_act": 0.33429858573849425}, "2": {"sr": 0.3, "mean_len": 1242.7, "mean_min_pen": 8.937442195415496, "mean_act": 0.3998076917437125}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.061083602905274, "mean_act": 0.5966902794524755}}, "elapsed_s": 313.8281009197235}
{"trial": 14, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.02, "W_COMPLETE": 100.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.05}, "score": 0.13999999999999999, "sr": {"1": 0.7, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.7, "mean_len": 912.4, "mean_min_pen": 2.940706562995911, "mean_act": 1.3471978399000248}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 9.901372599601746, "mean_act": 0.9463685217667609}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.291404342651367, "mean_act": 0.08601266834173493}}, "elapsed_s": 322.57220220565796}
{"trial": 15, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.05, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.02, "W_COMPLETE": 100.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": true, "ent_coef": 0.01}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1210.5, "mean_min_pen": 2.107759189605713, "mean_act": 0.08131515106917063}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 10.824185514450074, "mean_act": 0.20362997558291535}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.042323064804076, "mean_act": 0.17125511734669563}}, "elapsed_s": 312.3465087413788}
{"trial": 16, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 5.0, "W_STEP_COST": 0.005, "W_COMPLETE": 200.0, "W_COMPACT": 0.0, "ALIGN_SHAPE": "near", "ALIGN_GATED": true, "ent_coef": 0.05}, "score": 0.24, "sr": {"1": 0.7, "2": 0.2, "3": 0.0}, "details": {"1": {"sr": 0.7, "mean_len": 650.1, "mean_min_pen": 2.981771671772003, "mean_act": 0.1621352170537764}, "2": {"sr": 0.2, "mean_len": 1435.5, "mean_min_pen": 8.686615812778474, "mean_act": 0.3279171284351484}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.374159717559815, "mean_act": 0.04937917392927017}}, "elapsed_s": 303.71519470214844}
{"trial": 17, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.025, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.005, "W_COMPLETE": 100.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "near", "ALIGN_GATED": false, "ent_coef": 0.02}, "score": 0.16, "sr": {"1": 0.3, "2": 0.2, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1088.1, "mean_min_pen": 3.4793057322502134, "mean_act": 0.09515179877670824}, "2": {"sr": 0.2, "mean_len": 1428.5, "mean_min_pen": 10.024536824226379, "mean_act": 0.4135459636897354}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.302330660820008, "mean_act": 0.34973196326509737}}, "elapsed_s": 315.76633620262146}
{"trial": 18, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.025, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.05, "W_COMPLETE": 50.0, "W_COMPACT": 0.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": true, "ent_coef": 0.005}, "score": 0.16000000000000003, "sr": {"1": 0.8, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.8, "mean_len": 645.4, "mean_min_pen": 3.1326077818870544, "mean_act": 0.15081361126264722}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 10.723365247249603, "mean_act": 0.10806036127302399}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.303192138671875, "mean_act": 0.08246586098832388}}, "elapsed_s": 318.483638048172}
{"trial": 19, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.05, "W_COMPLETE": 100.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.02}, "score": 0.13, "sr": {"1": 0.4, "2": 0.1, "3": 0.0}, "details": {"1": {"sr": 0.4, "mean_len": 1231.4, "mean_min_pen": 2.6246669054031373, "mean_act": 0.07338090033141094}, "2": {"sr": 0.1, "mean_len": 1420.2, "mean_min_pen": 8.371916389465332, "mean_act": 0.16944798908643302}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 14.287557554244994, "mean_act": 0.09957915147298428}}, "elapsed_s": 315.07627868652344}
{"trial": 20, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.0, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.05, "W_COMPLETE": 100.0, "W_COMPACT": 0.5, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": true, "ent_coef": 0.005}, "score": 0.05, "sr": {"1": 0.0, "2": 0.1, "3": 0.0}, "details": {"1": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 1.5734932541847229, "mean_act": 0.08394606926547861}, "2": {"sr": 0.1, "mean_len": 1498.9, "mean_min_pen": 6.444609999656677, "mean_act": 0.2938110977638972}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 11.258054113388061, "mean_act": 0.16288984295733971}}, "elapsed_s": 309.5854580402374}
{"trial": 21, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.05, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.02, "W_COMPLETE": 100.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": true, "ent_coef": 0.005}, "score": 0.11, "sr": {"1": 0.3, "2": 0.1, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1324.6, "mean_min_pen": 3.3425565361976624, "mean_act": 0.1115106962044226}, "2": {"sr": 0.1, "mean_len": 1443.0, "mean_min_pen": 11.069470012187958, "mean_act": 0.17271345215252376}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.845431709289551, "mean_act": 0.13337391122176}}, "elapsed_s": 315.54923272132874}
{"trial": 22, "config": {"W_PER_SHEEP": 2.0, "W_ALIGN": 0.1, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.05, "W_COMPLETE": 100.0, "W_COMPACT": 1.5, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": true, "ent_coef": 0.05}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1220.2, "mean_min_pen": 2.1276236534118653, "mean_act": 0.4312911105166665}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 8.770305395126343, "mean_act": 0.6047595652043354}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.12634140253067, "mean_act": 0.14348885283676113}}, "elapsed_s": 471.740927696228}
{"trial": 23, "config": {"W_PER_SHEEP": 6.0, "W_ALIGN": 0.025, "W_PEN_BONUS": 20.0, "W_STEP_COST": 0.005, "W_COMPLETE": 200.0, "W_COMPACT": 3.0, "ALIGN_SHAPE": "standoff", "ALIGN_GATED": false, "ent_coef": 0.01}, "score": 0.06, "sr": {"1": 0.3, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.3, "mean_len": 1209.4, "mean_min_pen": 3.811609184741974, "mean_act": 0.08888363576016632}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 10.143073177337646, "mean_act": 0.27062979487000655}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 15.135865116119385, "mean_act": 0.3670903712440903}}, "elapsed_s": 335.26912212371826}
{"trial": 24, "config": {"W_PER_SHEEP": 1.0, "W_ALIGN": 0.0, "W_PEN_BONUS": 10.0, "W_STEP_COST": 0.05, "W_COMPLETE": 50.0, "W_COMPACT": 0.5, "ALIGN_SHAPE": "near", "ALIGN_GATED": true, "ent_coef": 0.02}, "score": 0.0, "sr": {"1": 0.0, "2": 0.0, "3": 0.0}, "details": {"1": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 10.014724779129029, "mean_act": 1.024556803444028}, "2": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 12.734652400016785, "mean_act": 1.0186923123559604}, "3": {"sr": 0.0, "mean_len": 1500.0, "mean_min_pen": 13.690151166915893, "mean_act": 1.000638129701217}}, "elapsed_s": 306.1110165119171}
training/runs/sweep_full.log
-681
View File
@@ -1,681 +0,0 @@
Sweep dir: runs/sweep_20260425_124630
Search space: ['W_PER_SHEEP', 'W_ALIGN', 'W_PEN_BONUS', 'W_STEP_COST', 'W_COMPLETE', 'W_COMPACT', 'ALIGN_SHAPE', 'ALIGN_GATED', 'ent_coef']
Per-trial: 1,000,000 steps train + 30 eval eps
Time budget: 7.5h
[Trial 1] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 100.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.005}
... [trial 1 | 1 sheep | 50,000 steps | ret(last 33)=-7.72 sr=6%]
... [trial 1 | 1 sheep | 100,000 steps | ret(last 50)=-10.07 sr=2%]
... [trial 1 | 1 sheep | 150,000 steps | ret(last 50)=-9.89 sr=2%]
... [trial 1 | 1 sheep | 200,000 steps | ret(last 50)=-7.94 sr=4%]
... [trial 1 | 1 sheep | 250,000 steps | ret(last 50)=+2.69 sr=2%]
... [trial 1 | 1 sheep | 300,000 steps | ret(last 50)=+18.25 sr=24%]
... [trial 1 | 1 sheep | 350,000 steps | ret(last 50)=+24.63 sr=20%]
... [trial 1 | 1 sheep | 400,000 steps | ret(last 50)=+24.83 sr=26%]
... [trial 1 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 1 | 2 sheep | 459,608 steps | ret(last 32)=+10.08 sr=0%]
... [trial 1 | 2 sheep | 509,608 steps | ret(last 50)=+11.51 sr=0%]
... [trial 1 | 2 sheep | 559,608 steps | ret(last 50)=+12.82 sr=0%]
... [trial 1 | 2 sheep | 609,608 steps | ret(last 50)=+14.39 sr=0%]
... [trial 1 | 2 sheep | 659,608 steps | ret(last 50)=+14.14 sr=0%]
... [trial 1 | 2 sheep | 709,608 steps | ret(last 50)=+12.36 sr=2%]
... [trial 1 | 2 sheep | 759,608 steps | ret(last 50)=+13.08 sr=0%]
... [trial 1 | 2 sheep | 809,608 steps | ret(last 50)=+13.24 sr=0%]
... [trial 1 | 2 sheep | 859,608 steps | ret(last 50)=+13.23 sr=0%]
... [trial 1 | 2 sheep | 909,608 steps | ret(last 50)=+14.23 sr=2%]
... [trial 1 | 2 sheep | 959,608 steps | ret(last 50)=+14.69 sr=0%]
... [trial 1 | 2 sheep | 1,009,608 steps | ret(last 50)=+20.23 sr=0%]
... [trial 1 | eval n=1]
... [trial 1 | eval n=2]
... [trial 1 | eval n=3]
→ score=0.150 sr1=0.50 sr2=0.10 sr3=0.00 [317s]
[Trial 2] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.05, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.005}
... [trial 2 | 1 sheep | 50,000 steps | ret(last 34)=-24.61 sr=9%]
... [trial 2 | 1 sheep | 100,000 steps | ret(last 50)=-28.20 sr=10%]
... [trial 2 | 1 sheep | 150,000 steps | ret(last 50)=-28.14 sr=8%]
... [trial 2 | 1 sheep | 200,000 steps | ret(last 50)=-31.36 sr=2%]
... [trial 2 | 1 sheep | 250,000 steps | ret(last 50)=-31.38 sr=6%]
... [trial 2 | 1 sheep | 300,000 steps | ret(last 50)=-32.89 sr=4%]
... [trial 2 | 1 sheep | 350,000 steps | ret(last 50)=-29.11 sr=8%]
... [trial 2 | 1 sheep | 400,000 steps | ret(last 50)=-19.16 sr=30%]
... [trial 2 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 2 | 2 sheep | 459,608 steps | ret(last 34)=-17.61 sr=9%]
... [trial 2 | 2 sheep | 509,608 steps | ret(last 50)=-18.59 sr=2%]
... [trial 2 | 2 sheep | 559,608 steps | ret(last 50)=-16.92 sr=0%]
... [trial 2 | 2 sheep | 609,608 steps | ret(last 50)=-17.40 sr=0%]
... [trial 2 | 2 sheep | 659,608 steps | ret(last 50)=-18.13 sr=0%]
... [trial 2 | 2 sheep | 709,608 steps | ret(last 50)=-17.45 sr=0%]
... [trial 2 | 2 sheep | 759,608 steps | ret(last 50)=-16.06 sr=0%]
... [trial 2 | 2 sheep | 809,608 steps | ret(last 50)=-15.35 sr=0%]
... [trial 2 | 2 sheep | 859,608 steps | ret(last 50)=-12.63 sr=0%]
... [trial 2 | 2 sheep | 909,608 steps | ret(last 50)=-12.41 sr=0%]
... [trial 2 | 2 sheep | 959,608 steps | ret(last 50)=-12.91 sr=0%]
... [trial 2 | 2 sheep | 1,009,608 steps | ret(last 50)=-10.94 sr=0%]
... [trial 2 | eval n=1]
... [trial 2 | eval n=2]
... [trial 2 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [311s]
[Trial 3] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.025, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 50.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.01}
... [trial 3 | 1 sheep | 50,000 steps | ret(last 32)=-1.75 sr=0%]
... [trial 3 | 1 sheep | 100,000 steps | ret(last 50)=-3.70 sr=0%]
... [trial 3 | 1 sheep | 150,000 steps | ret(last 50)=-6.09 sr=2%]
... [trial 3 | 1 sheep | 200,000 steps | ret(last 50)=-3.44 sr=4%]
... [trial 3 | 1 sheep | 250,000 steps | ret(last 50)=+6.68 sr=8%]
... [trial 3 | 1 sheep | 300,000 steps | ret(last 50)=+14.58 sr=22%]
... [trial 3 | 1 sheep | 350,000 steps | ret(last 50)=+15.28 sr=64%]
... [trial 3 | 1 sheep | 400,000 steps | ret(last 50)=+14.70 sr=74%]
... [trial 3 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 3 | 2 sheep | 459,608 steps | ret(last 35)=+0.82 sr=9%]
... [trial 3 | 2 sheep | 509,608 steps | ret(last 50)=-0.66 sr=2%]
... [trial 3 | 2 sheep | 559,608 steps | ret(last 50)=-0.02 sr=0%]
... [trial 3 | 2 sheep | 609,608 steps | ret(last 50)=-0.02 sr=0%]
... [trial 3 | 2 sheep | 659,608 steps | ret(last 50)=+1.37 sr=4%]
... [trial 3 | 2 sheep | 709,608 steps | ret(last 50)=+2.75 sr=8%]
... [trial 3 | 2 sheep | 759,608 steps | ret(last 50)=+1.25 sr=6%]
... [trial 3 | 2 sheep | 809,608 steps | ret(last 50)=+4.20 sr=10%]
... [trial 3 | 2 sheep | 859,608 steps | ret(last 50)=+2.14 sr=4%]
... [trial 3 | 2 sheep | 909,608 steps | ret(last 50)=+3.13 sr=8%]
... [trial 3 | 2 sheep | 959,608 steps | ret(last 50)=+5.16 sr=6%]
... [trial 3 | 2 sheep | 1,009,608 steps | ret(last 50)=+5.95 sr=8%]
... [trial 3 | eval n=1]
... [trial 3 | eval n=2]
... [trial 3 | eval n=3]
→ score=0.270 sr1=0.70 sr2=0.20 sr3=0.10 [304s]
[Trial 4] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.05, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 50.0, 'W_COMPACT': 0.0, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.005}
... [trial 4 | 1 sheep | 50,000 steps | ret(last 33)=-2.86 sr=9%]
... [trial 4 | 1 sheep | 100,000 steps | ret(last 50)=-3.54 sr=6%]
... [trial 4 | 1 sheep | 150,000 steps | ret(last 50)=-2.76 sr=8%]
... [trial 4 | 1 sheep | 200,000 steps | ret(last 50)=-1.56 sr=8%]
... [trial 4 | 1 sheep | 250,000 steps | ret(last 50)=+9.18 sr=24%]
... [trial 4 | 1 sheep | 300,000 steps | ret(last 50)=+18.46 sr=46%]
... [trial 4 | 1 sheep | 350,000 steps | ret(last 50)=+15.01 sr=34%]
... [trial 4 | 1 sheep | 400,000 steps | ret(last 50)=+14.44 sr=42%]
... [trial 4 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 4 | 2 sheep | 459,608 steps | ret(last 35)=+6.77 sr=9%]
... [trial 4 | 2 sheep | 509,608 steps | ret(last 50)=+5.50 sr=6%]
... [trial 4 | 2 sheep | 559,608 steps | ret(last 50)=+4.39 sr=0%]
... [trial 4 | 2 sheep | 609,608 steps | ret(last 50)=+4.54 sr=0%]
... [trial 4 | 2 sheep | 659,608 steps | ret(last 50)=+6.97 sr=0%]
... [trial 4 | 2 sheep | 709,608 steps | ret(last 50)=+4.28 sr=4%]
... [trial 4 | 2 sheep | 759,608 steps | ret(last 50)=+4.30 sr=2%]
... [trial 4 | 2 sheep | 809,608 steps | ret(last 50)=+6.34 sr=4%]
... [trial 4 | 2 sheep | 859,608 steps | ret(last 50)=+7.27 sr=2%]
... [trial 4 | 2 sheep | 909,608 steps | ret(last 50)=+8.22 sr=4%]
... [trial 4 | 2 sheep | 959,608 steps | ret(last 50)=+7.23 sr=6%]
... [trial 4 | 2 sheep | 1,009,608 steps | ret(last 50)=+7.24 sr=2%]
... [trial 4 | eval n=1]
... [trial 4 | eval n=2]
... [trial 4 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [302s]
[Trial 5] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 50.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': True, 'ent_coef': 0.005}
... [trial 5 | 1 sheep | 50,000 steps | ret(last 33)=+3.70 sr=6%]
... [trial 5 | 1 sheep | 100,000 steps | ret(last 50)=-2.32 sr=0%]
... [trial 5 | 1 sheep | 150,000 steps | ret(last 50)=-4.36 sr=4%]
... [trial 5 | 1 sheep | 200,000 steps | ret(last 50)=-4.30 sr=6%]
... [trial 5 | 1 sheep | 250,000 steps | ret(last 50)=-0.15 sr=14%]
... [trial 5 | 1 sheep | 300,000 steps | ret(last 50)=+1.39 sr=8%]
... [trial 5 | 1 sheep | 350,000 steps | ret(last 50)=+11.40 sr=36%]
... [trial 5 | 1 sheep | 400,000 steps | ret(last 50)=+11.08 sr=24%]
... [trial 5 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 5 | 2 sheep | 459,608 steps | ret(last 34)=+6.85 sr=6%]
... [trial 5 | 2 sheep | 509,608 steps | ret(last 50)=+7.35 sr=8%]
... [trial 5 | 2 sheep | 559,608 steps | ret(last 50)=+7.57 sr=4%]
... [trial 5 | 2 sheep | 609,608 steps | ret(last 50)=+6.64 sr=2%]
... [trial 5 | 2 sheep | 659,608 steps | ret(last 50)=+9.15 sr=10%]
... [trial 5 | 2 sheep | 709,608 steps | ret(last 50)=+14.27 sr=10%]
... [trial 5 | 2 sheep | 759,608 steps | ret(last 50)=+10.93 sr=6%]
... [trial 5 | 2 sheep | 809,608 steps | ret(last 50)=+10.17 sr=12%]
... [trial 5 | 2 sheep | 859,608 steps | ret(last 50)=+8.20 sr=8%]
... [trial 5 | 2 sheep | 909,608 steps | ret(last 50)=+9.61 sr=14%]
... [trial 5 | 2 sheep | 959,608 steps | ret(last 50)=+11.14 sr=10%]
... [trial 5 | 2 sheep | 1,009,608 steps | ret(last 50)=+10.75 sr=12%]
... [trial 5 | eval n=1]
... [trial 5 | eval n=2]
... [trial 5 | eval n=3]
→ score=0.200 sr1=1.00 sr2=0.00 sr3=0.00 [314s]
[Trial 6] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.025, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 200.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': True, 'ent_coef': 0.01}
... [trial 6 | 1 sheep | 50,000 steps | ret(last 32)=-13.18 sr=9%]
... [trial 6 | 1 sheep | 100,000 steps | ret(last 50)=-10.28 sr=16%]
... [trial 6 | 1 sheep | 150,000 steps | ret(last 50)=+5.28 sr=44%]
... [trial 6 | 1 sheep | 200,000 steps | ret(last 50)=+9.40 sr=38%]
... [trial 6 | 1 sheep | 250,000 steps | ret(last 50)=+8.62 sr=32%]
... [trial 6 | 1 sheep | 300,000 steps | ret(last 50)=+9.14 sr=34%]
... [trial 6 | 1 sheep | 350,000 steps | ret(last 50)=+12.59 sr=60%]
... [trial 6 | 1 sheep | 400,000 steps | ret(last 50)=+14.10 sr=72%]
... [trial 6 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 6 | 2 sheep | 459,608 steps | ret(last 34)=+0.12 sr=9%]
... [trial 6 | 2 sheep | 509,608 steps | ret(last 50)=-2.84 sr=4%]
... [trial 6 | 2 sheep | 559,608 steps | ret(last 50)=-2.11 sr=10%]
... [trial 6 | 2 sheep | 609,608 steps | ret(last 50)=-1.91 sr=14%]
... [trial 6 | 2 sheep | 659,608 steps | ret(last 50)=-2.14 sr=14%]
... [trial 6 | 2 sheep | 709,608 steps | ret(last 50)=-4.30 sr=6%]
... [trial 6 | 2 sheep | 759,608 steps | ret(last 50)=-1.89 sr=10%]
... [trial 6 | 2 sheep | 809,608 steps | ret(last 50)=-3.47 sr=8%]
... [trial 6 | 2 sheep | 859,608 steps | ret(last 50)=-1.45 sr=8%]
... [trial 6 | 2 sheep | 909,608 steps | ret(last 50)=-3.55 sr=2%]
... [trial 6 | 2 sheep | 959,608 steps | ret(last 50)=-2.93 sr=4%]
... [trial 6 | 2 sheep | 1,009,608 steps | ret(last 50)=-1.45 sr=10%]
... [trial 6 | eval n=1]
... [trial 6 | eval n=2]
... [trial 6 | eval n=3]
→ score=0.160 sr1=0.80 sr2=0.00 sr3=0.00 [312s]
[Trial 7] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.005, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.01}
... [trial 7 | 1 sheep | 50,000 steps | ret(last 32)=-8.47 sr=0%]
... [trial 7 | 1 sheep | 100,000 steps | ret(last 50)=-5.40 sr=4%]
... [trial 7 | 1 sheep | 150,000 steps | ret(last 50)=-2.72 sr=10%]
... [trial 7 | 1 sheep | 200,000 steps | ret(last 50)=-1.59 sr=10%]
... [trial 7 | 1 sheep | 250,000 steps | ret(last 50)=-1.58 sr=6%]
... [trial 7 | 1 sheep | 300,000 steps | ret(last 50)=-3.68 sr=2%]
... [trial 7 | 1 sheep | 350,000 steps | ret(last 50)=+4.82 sr=10%]
... [trial 7 | 1 sheep | 400,000 steps | ret(last 50)=+15.81 sr=54%]
... [trial 7 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 7 | 2 sheep | 459,608 steps | ret(last 32)=-2.50 sr=6%]
... [trial 7 | 2 sheep | 509,608 steps | ret(last 50)=-2.32 sr=2%]
... [trial 7 | 2 sheep | 559,608 steps | ret(last 50)=+0.76 sr=4%]
... [trial 7 | 2 sheep | 609,608 steps | ret(last 50)=+0.45 sr=0%]
... [trial 7 | 2 sheep | 659,608 steps | ret(last 50)=+1.03 sr=8%]
... [trial 7 | 2 sheep | 709,608 steps | ret(last 50)=+0.62 sr=6%]
... [trial 7 | 2 sheep | 759,608 steps | ret(last 50)=+0.36 sr=8%]
... [trial 7 | 2 sheep | 809,608 steps | ret(last 50)=+2.27 sr=10%]
... [trial 7 | 2 sheep | 859,608 steps | ret(last 50)=+2.31 sr=6%]
... [trial 7 | 2 sheep | 909,608 steps | ret(last 50)=+3.78 sr=4%]
... [trial 7 | 2 sheep | 959,608 steps | ret(last 50)=+2.21 sr=10%]
... [trial 7 | 2 sheep | 1,009,608 steps | ret(last 50)=+2.66 sr=4%]
... [trial 7 | eval n=1]
... [trial 7 | eval n=2]
... [trial 7 | eval n=3]
→ score=0.080 sr1=0.40 sr2=0.00 sr3=0.00 [338s]
[Trial 8] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.025, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 100.0, 'W_COMPACT': 0.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.005}
... [trial 8 | 1 sheep | 50,000 steps | ret(last 32)=-7.73 sr=6%]
... [trial 8 | 1 sheep | 100,000 steps | ret(last 50)=-9.58 sr=8%]
... [trial 8 | 1 sheep | 150,000 steps | ret(last 50)=-10.87 sr=8%]
... [trial 8 | 1 sheep | 200,000 steps | ret(last 50)=-9.79 sr=6%]
... [trial 8 | 1 sheep | 250,000 steps | ret(last 50)=-7.19 sr=8%]
... [trial 8 | 1 sheep | 300,000 steps | ret(last 50)=-3.84 sr=18%]
... [trial 8 | 1 sheep | 350,000 steps | ret(last 50)=-0.03 sr=26%]
... [trial 8 | 1 sheep | 400,000 steps | ret(last 50)=+6.80 sr=44%]
... [trial 8 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 8 | 2 sheep | 459,608 steps | ret(last 35)=-3.00 sr=9%]
... [trial 8 | 2 sheep | 509,608 steps | ret(last 50)=-4.26 sr=4%]
... [trial 8 | 2 sheep | 559,608 steps | ret(last 50)=+1.91 sr=14%]
... [trial 8 | 2 sheep | 609,608 steps | ret(last 50)=-0.57 sr=16%]
... [trial 8 | 2 sheep | 659,608 steps | ret(last 50)=+1.65 sr=14%]
... [trial 8 | 2 sheep | 709,608 steps | ret(last 50)=+2.90 sr=8%]
... [trial 8 | 2 sheep | 759,608 steps | ret(last 50)=+0.98 sr=2%]
... [trial 8 | 2 sheep | 809,608 steps | ret(last 50)=-2.52 sr=4%]
... [trial 8 | 2 sheep | 859,608 steps | ret(last 50)=-1.11 sr=2%]
... [trial 8 | 2 sheep | 909,608 steps | ret(last 50)=+2.74 sr=2%]
... [trial 8 | 2 sheep | 959,608 steps | ret(last 50)=+2.94 sr=0%]
... [trial 8 | 2 sheep | 1,009,608 steps | ret(last 50)=+5.13 sr=0%]
... [trial 8 | eval n=1]
... [trial 8 | eval n=2]
... [trial 8 | eval n=3]
→ score=0.110 sr1=0.30 sr2=0.10 sr3=0.00 [451s]
[Trial 9] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 50.0, 'W_COMPACT': 0.0, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.05}
... [trial 9 | 1 sheep | 50,000 steps | ret(last 34)=-11.25 sr=15%]
... [trial 9 | 1 sheep | 100,000 steps | ret(last 50)=-11.98 sr=8%]
... [trial 9 | 1 sheep | 150,000 steps | ret(last 50)=-10.46 sr=14%]
... [trial 9 | 1 sheep | 200,000 steps | ret(last 50)=-2.86 sr=14%]
... [trial 9 | 1 sheep | 250,000 steps | ret(last 50)=+8.65 sr=60%]
... [trial 9 | 1 sheep | 300,000 steps | ret(last 50)=+10.48 sr=58%]
... [trial 9 | 1 sheep | 350,000 steps | ret(last 50)=+8.65 sr=56%]
... [trial 9 | 1 sheep | 400,000 steps | ret(last 50)=+10.25 sr=68%]
... [trial 9 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 9 | 2 sheep | 459,608 steps | ret(last 35)=-0.75 sr=20%]
... [trial 9 | 2 sheep | 509,608 steps | ret(last 50)=-6.64 sr=2%]
... [trial 9 | 2 sheep | 559,608 steps | ret(last 50)=-7.43 sr=4%]
... [trial 9 | 2 sheep | 609,608 steps | ret(last 50)=-4.32 sr=6%]
... [trial 9 | 2 sheep | 659,608 steps | ret(last 50)=-3.64 sr=6%]
... [trial 9 | 2 sheep | 709,608 steps | ret(last 50)=-7.09 sr=0%]
... [trial 9 | 2 sheep | 759,608 steps | ret(last 50)=-5.60 sr=4%]
... [trial 9 | 2 sheep | 809,608 steps | ret(last 50)=-5.70 sr=6%]
... [trial 9 | 2 sheep | 859,608 steps | ret(last 50)=-4.99 sr=4%]
... [trial 9 | 2 sheep | 909,608 steps | ret(last 50)=-4.60 sr=6%]
... [trial 9 | 2 sheep | 959,608 steps | ret(last 50)=-6.53 sr=4%]
... [trial 9 | 2 sheep | 1,009,608 steps | ret(last 50)=-7.46 sr=2%]
... [trial 9 | eval n=1]
... [trial 9 | eval n=2]
... [trial 9 | eval n=3]
→ score=0.190 sr1=0.70 sr2=0.10 sr3=0.00 [349s]
[Trial 10] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 0.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.05}
... [trial 10 | 1 sheep | 50,000 steps | ret(last 32)=-13.35 sr=3%]
... [trial 10 | 1 sheep | 100,000 steps | ret(last 50)=-12.49 sr=4%]
... [trial 10 | 1 sheep | 150,000 steps | ret(last 50)=-13.24 sr=8%]
... [trial 10 | 1 sheep | 200,000 steps | ret(last 50)=-12.73 sr=10%]
... [trial 10 | 1 sheep | 250,000 steps | ret(last 50)=-15.27 sr=4%]
... [trial 10 | 1 sheep | 300,000 steps | ret(last 50)=-9.43 sr=8%]
... [trial 10 | 1 sheep | 350,000 steps | ret(last 50)=-2.65 sr=22%]
... [trial 10 | 1 sheep | 400,000 steps | ret(last 50)=+5.12 sr=46%]
... [trial 10 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 10 | 2 sheep | 459,608 steps | ret(last 34)=-4.93 sr=6%]
... [trial 10 | 2 sheep | 509,608 steps | ret(last 50)=-6.25 sr=2%]
... [trial 10 | 2 sheep | 559,608 steps | ret(last 50)=-5.57 sr=4%]
... [trial 10 | 2 sheep | 609,608 steps | ret(last 50)=-6.24 sr=4%]
... [trial 10 | 2 sheep | 659,608 steps | ret(last 50)=-9.34 sr=0%]
... [trial 10 | 2 sheep | 709,608 steps | ret(last 50)=-8.23 sr=0%]
... [trial 10 | 2 sheep | 759,608 steps | ret(last 50)=-8.34 sr=0%]
... [trial 10 | 2 sheep | 809,608 steps | ret(last 50)=-5.27 sr=0%]
... [trial 10 | 2 sheep | 859,608 steps | ret(last 50)=-8.24 sr=0%]
... [trial 10 | 2 sheep | 909,608 steps | ret(last 50)=-8.75 sr=0%]
... [trial 10 | 2 sheep | 959,608 steps | ret(last 50)=-9.15 sr=0%]
... [trial 10 | 2 sheep | 1,009,608 steps | ret(last 50)=-9.75 sr=0%]
... [trial 10 | eval n=1]
... [trial 10 | eval n=2]
... [trial 10 | eval n=3]
→ score=0.000 sr1=0.00 sr2=0.00 sr3=0.00 [319s]
[Trial 11] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 0.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
... [trial 11 | 1 sheep | 50,000 steps | ret(last 32)=-3.50 sr=12%]
... [trial 11 | 1 sheep | 100,000 steps | ret(last 50)=-5.79 sr=6%]
... [trial 11 | 1 sheep | 150,000 steps | ret(last 50)=-2.10 sr=18%]
... [trial 11 | 1 sheep | 200,000 steps | ret(last 50)=+2.60 sr=8%]
... [trial 11 | 1 sheep | 250,000 steps | ret(last 50)=+11.49 sr=8%]
... [trial 11 | 1 sheep | 300,000 steps | ret(last 50)=+21.73 sr=26%]
... [trial 11 | 1 sheep | 350,000 steps | ret(last 50)=+20.73 sr=36%]
... [trial 11 | 1 sheep | 400,000 steps | ret(last 50)=+19.77 sr=62%]
... [trial 11 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 11 | 2 sheep | 459,608 steps | ret(last 36)=+10.19 sr=11%]
... [trial 11 | 2 sheep | 509,608 steps | ret(last 50)=+11.56 sr=6%]
... [trial 11 | 2 sheep | 559,608 steps | ret(last 50)=+13.61 sr=2%]
... [trial 11 | 2 sheep | 609,608 steps | ret(last 50)=+15.44 sr=4%]
... [trial 11 | 2 sheep | 659,608 steps | ret(last 50)=+15.61 sr=10%]
... [trial 11 | 2 sheep | 709,608 steps | ret(last 50)=+16.30 sr=6%]
... [trial 11 | 2 sheep | 759,608 steps | ret(last 50)=+17.33 sr=4%]
... [trial 11 | 2 sheep | 809,608 steps | ret(last 50)=+18.36 sr=2%]
... [trial 11 | 2 sheep | 859,608 steps | ret(last 50)=+19.78 sr=8%]
... [trial 11 | 2 sheep | 909,608 steps | ret(last 50)=+20.12 sr=14%]
... [trial 11 | 2 sheep | 959,608 steps | ret(last 50)=+18.93 sr=8%]
... [trial 11 | 2 sheep | 1,009,608 steps | ret(last 50)=+18.16 sr=2%]
... [trial 11 | eval n=1]
... [trial 11 | eval n=2]
... [trial 11 | eval n=3]
→ score=0.160 sr1=0.80 sr2=0.00 sr3=0.00 [310s]
[Trial 12] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 50.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.02}
... [trial 12 | 1 sheep | 50,000 steps | ret(last 32)=-42.77 sr=0%]
... [trial 12 | 1 sheep | 100,000 steps | ret(last 50)=-39.16 sr=2%]
... [trial 12 | 1 sheep | 150,000 steps | ret(last 50)=-35.02 sr=6%]
... [trial 12 | 1 sheep | 200,000 steps | ret(last 50)=-31.49 sr=4%]
... [trial 12 | 1 sheep | 250,000 steps | ret(last 50)=-8.31 sr=16%]
... [trial 12 | 1 sheep | 300,000 steps | ret(last 50)=+7.97 sr=36%]
... [trial 12 | 1 sheep | 350,000 steps | ret(last 50)=+11.77 sr=68%]
... [trial 12 | 1 sheep | 400,000 steps | ret(last 50)=+12.47 sr=74%]
... [trial 12 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 12 | 2 sheep | 459,608 steps | ret(last 34)=-9.76 sr=0%]
... [trial 12 | 2 sheep | 509,608 steps | ret(last 50)=-4.85 sr=0%]
... [trial 12 | 2 sheep | 559,608 steps | ret(last 50)=-2.81 sr=8%]
... [trial 12 | 2 sheep | 609,608 steps | ret(last 50)=+2.27 sr=10%]
... [trial 12 | 2 sheep | 659,608 steps | ret(last 50)=+1.66 sr=6%]
... [trial 12 | 2 sheep | 709,608 steps | ret(last 50)=+3.42 sr=4%]
... [trial 12 | 2 sheep | 759,608 steps | ret(last 50)=+4.08 sr=2%]
... [trial 12 | 2 sheep | 809,608 steps | ret(last 50)=+5.49 sr=2%]
... [trial 12 | 2 sheep | 859,608 steps | ret(last 50)=+7.12 sr=10%]
... [trial 12 | 2 sheep | 909,608 steps | ret(last 50)=+7.91 sr=6%]
... [trial 12 | 2 sheep | 959,608 steps | ret(last 50)=+6.87 sr=2%]
... [trial 12 | 2 sheep | 1,009,608 steps | ret(last 50)=+5.83 sr=2%]
... [trial 12 | eval n=1]
... [trial 12 | eval n=2]
... [trial 12 | eval n=3]
→ score=0.240 sr1=0.70 sr2=0.20 sr3=0.00 [330s]
[Trial 13] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.05, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 100.0, 'W_COMPACT': 0.5, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.005}
... [trial 13 | 1 sheep | 50,000 steps | ret(last 34)=-31.15 sr=9%]
... [trial 13 | 1 sheep | 100,000 steps | ret(last 50)=-32.34 sr=4%]
... [trial 13 | 1 sheep | 150,000 steps | ret(last 50)=-33.16 sr=0%]
... [trial 13 | 1 sheep | 200,000 steps | ret(last 50)=-29.98 sr=6%]
... [trial 13 | 1 sheep | 250,000 steps | ret(last 50)=-28.64 sr=4%]
... [trial 13 | 1 sheep | 300,000 steps | ret(last 50)=-17.91 sr=14%]
... [trial 13 | 1 sheep | 350,000 steps | ret(last 50)=-15.27 sr=22%]
... [trial 13 | 1 sheep | 400,000 steps | ret(last 50)=-11.36 sr=16%]
... [trial 13 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 13 | 2 sheep | 459,608 steps | ret(last 34)=-16.78 sr=0%]
... [trial 13 | 2 sheep | 509,608 steps | ret(last 50)=-16.84 sr=2%]
... [trial 13 | 2 sheep | 559,608 steps | ret(last 50)=-14.28 sr=0%]
... [trial 13 | 2 sheep | 609,608 steps | ret(last 50)=-12.35 sr=6%]
... [trial 13 | 2 sheep | 659,608 steps | ret(last 50)=-14.50 sr=2%]
... [trial 13 | 2 sheep | 709,608 steps | ret(last 50)=-12.96 sr=2%]
... [trial 13 | 2 sheep | 759,608 steps | ret(last 50)=-9.86 sr=4%]
... [trial 13 | 2 sheep | 809,608 steps | ret(last 50)=-13.88 sr=2%]
... [trial 13 | 2 sheep | 859,608 steps | ret(last 50)=-14.76 sr=0%]
... [trial 13 | 2 sheep | 909,608 steps | ret(last 50)=-12.79 sr=0%]
... [trial 13 | 2 sheep | 959,608 steps | ret(last 50)=-12.54 sr=0%]
... [trial 13 | 2 sheep | 1,009,608 steps | ret(last 50)=-12.11 sr=8%]
... [trial 13 | eval n=1]
... [trial 13 | eval n=2]
... [trial 13 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [323s]
[Trial 14] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 200.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
... [trial 14 | 1 sheep | 50,000 steps | ret(last 32)=-20.15 sr=9%]
... [trial 14 | 1 sheep | 100,000 steps | ret(last 50)=-15.28 sr=8%]
... [trial 14 | 1 sheep | 150,000 steps | ret(last 50)=-8.87 sr=26%]
... [trial 14 | 1 sheep | 200,000 steps | ret(last 50)=-9.94 sr=8%]
... [trial 14 | 1 sheep | 250,000 steps | ret(last 50)=-9.04 sr=8%]
... [trial 14 | 1 sheep | 300,000 steps | ret(last 50)=-7.40 sr=14%]
... [trial 14 | 1 sheep | 350,000 steps | ret(last 50)=+2.22 sr=50%]
... [trial 14 | 1 sheep | 400,000 steps | ret(last 50)=+4.06 sr=58%]
... [trial 14 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 14 | 2 sheep | 459,608 steps | ret(last 33)=-5.93 sr=3%]
... [trial 14 | 2 sheep | 509,608 steps | ret(last 50)=-6.85 sr=4%]
... [trial 14 | 2 sheep | 559,608 steps | ret(last 50)=-6.81 sr=6%]
... [trial 14 | 2 sheep | 609,608 steps | ret(last 50)=-4.80 sr=4%]
... [trial 14 | 2 sheep | 659,608 steps | ret(last 50)=-6.55 sr=4%]
... [trial 14 | 2 sheep | 709,608 steps | ret(last 50)=-4.81 sr=12%]
... [trial 14 | 2 sheep | 759,608 steps | ret(last 50)=-5.41 sr=10%]
... [trial 14 | 2 sheep | 809,608 steps | ret(last 50)=-0.00 sr=30%]
... [trial 14 | 2 sheep | 859,608 steps | ret(last 50)=+1.17 sr=26%]
... [trial 14 | 2 sheep | 909,608 steps | ret(last 50)=+0.17 sr=20%]
... [trial 14 | 2 sheep | 959,608 steps | ret(last 50)=-0.96 sr=18%]
... [trial 14 | 2 sheep | 1,009,608 steps | ret(last 50)=-1.33 sr=20%]
... [trial 14 | eval n=1]
... [trial 14 | eval n=2]
... [trial 14 | eval n=3]
→ score=0.350 sr1=1.00 sr2=0.30 sr3=0.00 [314s]
[Trial 15] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 100.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.05}
... [trial 15 | 1 sheep | 50,000 steps | ret(last 32)=-6.83 sr=3%]
... [trial 15 | 1 sheep | 100,000 steps | ret(last 50)=-7.59 sr=4%]
... [trial 15 | 1 sheep | 150,000 steps | ret(last 50)=-5.74 sr=6%]
... [trial 15 | 1 sheep | 200,000 steps | ret(last 50)=-5.92 sr=6%]
... [trial 15 | 1 sheep | 250,000 steps | ret(last 50)=+8.14 sr=22%]
... [trial 15 | 1 sheep | 300,000 steps | ret(last 50)=+15.51 sr=22%]
... [trial 15 | 1 sheep | 350,000 steps | ret(last 50)=+21.46 sr=20%]
... [trial 15 | 1 sheep | 400,000 steps | ret(last 50)=+22.52 sr=16%]
... [trial 15 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 15 | 2 sheep | 459,608 steps | ret(last 35)=+6.28 sr=0%]
... [trial 15 | 2 sheep | 509,608 steps | ret(last 50)=+13.19 sr=2%]
... [trial 15 | 2 sheep | 559,608 steps | ret(last 50)=+15.58 sr=4%]
... [trial 15 | 2 sheep | 609,608 steps | ret(last 50)=+18.78 sr=10%]
... [trial 15 | 2 sheep | 659,608 steps | ret(last 50)=+22.71 sr=10%]
... [trial 15 | 2 sheep | 709,608 steps | ret(last 50)=+23.95 sr=6%]
... [trial 15 | 2 sheep | 759,608 steps | ret(last 50)=+24.84 sr=14%]
... [trial 15 | 2 sheep | 809,608 steps | ret(last 50)=+24.00 sr=8%]
... [trial 15 | 2 sheep | 859,608 steps | ret(last 50)=+23.91 sr=2%]
... [trial 15 | 2 sheep | 909,608 steps | ret(last 50)=+23.73 sr=4%]
... [trial 15 | 2 sheep | 959,608 steps | ret(last 50)=+24.23 sr=4%]
... [trial 15 | 2 sheep | 1,009,608 steps | ret(last 50)=+24.77 sr=4%]
... [trial 15 | eval n=1]
... [trial 15 | eval n=2]
... [trial 15 | eval n=3]
→ score=0.140 sr1=0.70 sr2=0.00 sr3=0.00 [323s]
[Trial 16] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.05, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 100.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': True, 'ent_coef': 0.01}
... [trial 16 | 1 sheep | 50,000 steps | ret(last 32)=-7.14 sr=9%]
... [trial 16 | 1 sheep | 100,000 steps | ret(last 50)=-5.58 sr=12%]
... [trial 16 | 1 sheep | 150,000 steps | ret(last 50)=+5.93 sr=26%]
... [trial 16 | 1 sheep | 200,000 steps | ret(last 50)=+15.53 sr=68%]
... [trial 16 | 1 sheep | 250,000 steps | ret(last 50)=+14.88 sr=56%]
... [trial 16 | 1 sheep | 300,000 steps | ret(last 50)=+13.86 sr=36%]
... [trial 16 | 1 sheep | 350,000 steps | ret(last 50)=+14.84 sr=54%]
... [trial 16 | 1 sheep | 400,000 steps | ret(last 50)=+15.15 sr=70%]
... [trial 16 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 16 | 2 sheep | 459,608 steps | ret(last 34)=-1.47 sr=6%]
... [trial 16 | 2 sheep | 509,608 steps | ret(last 50)=-1.63 sr=2%]
... [trial 16 | 2 sheep | 559,608 steps | ret(last 50)=-3.78 sr=2%]
... [trial 16 | 2 sheep | 609,608 steps | ret(last 50)=-2.17 sr=4%]
... [trial 16 | 2 sheep | 659,608 steps | ret(last 50)=+1.25 sr=6%]
... [trial 16 | 2 sheep | 709,608 steps | ret(last 50)=+0.28 sr=4%]
... [trial 16 | 2 sheep | 759,608 steps | ret(last 50)=+2.74 sr=4%]
... [trial 16 | 2 sheep | 809,608 steps | ret(last 50)=+7.19 sr=6%]
... [trial 16 | 2 sheep | 859,608 steps | ret(last 50)=+7.68 sr=4%]
... [trial 16 | 2 sheep | 909,608 steps | ret(last 50)=+2.38 sr=0%]
... [trial 16 | 2 sheep | 959,608 steps | ret(last 50)=+3.43 sr=0%]
... [trial 16 | 2 sheep | 1,009,608 steps | ret(last 50)=+11.11 sr=0%]
... [trial 16 | eval n=1]
... [trial 16 | eval n=2]
... [trial 16 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [312s]
[Trial 17] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 5.0, 'W_STEP_COST': 0.005, 'W_COMPLETE': 200.0, 'W_COMPACT': 0.0, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': True, 'ent_coef': 0.05}
... [trial 17 | 1 sheep | 50,000 steps | ret(last 32)=+2.15 sr=6%]
... [trial 17 | 1 sheep | 100,000 steps | ret(last 50)=-0.51 sr=2%]
... [trial 17 | 1 sheep | 150,000 steps | ret(last 50)=+0.84 sr=6%]
... [trial 17 | 1 sheep | 200,000 steps | ret(last 50)=+2.96 sr=6%]
... [trial 17 | 1 sheep | 250,000 steps | ret(last 50)=+3.04 sr=4%]
... [trial 17 | 1 sheep | 300,000 steps | ret(last 50)=+10.58 sr=10%]
... [trial 17 | 1 sheep | 350,000 steps | ret(last 50)=+21.95 sr=36%]
... [trial 17 | 1 sheep | 400,000 steps | ret(last 50)=+19.20 sr=16%]
... [trial 17 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 17 | 2 sheep | 459,608 steps | ret(last 32)=+10.27 sr=16%]
... [trial 17 | 2 sheep | 509,608 steps | ret(last 50)=+12.25 sr=6%]
... [trial 17 | 2 sheep | 559,608 steps | ret(last 50)=+12.94 sr=6%]
... [trial 17 | 2 sheep | 609,608 steps | ret(last 50)=+11.82 sr=4%]
... [trial 17 | 2 sheep | 659,608 steps | ret(last 50)=+13.45 sr=4%]
... [trial 17 | 2 sheep | 709,608 steps | ret(last 50)=+13.03 sr=4%]
... [trial 17 | 2 sheep | 759,608 steps | ret(last 50)=+10.69 sr=6%]
... [trial 17 | 2 sheep | 809,608 steps | ret(last 50)=+7.79 sr=6%]
... [trial 17 | 2 sheep | 859,608 steps | ret(last 50)=+12.16 sr=16%]
... [trial 17 | 2 sheep | 909,608 steps | ret(last 50)=+11.75 sr=12%]
... [trial 17 | 2 sheep | 959,608 steps | ret(last 50)=+13.65 sr=16%]
... [trial 17 | 2 sheep | 1,009,608 steps | ret(last 50)=+12.43 sr=10%]
... [trial 17 | eval n=1]
... [trial 17 | eval n=2]
... [trial 17 | eval n=3]
→ score=0.240 sr1=0.70 sr2=0.20 sr3=0.00 [304s]
[Trial 18] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.025, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.005, 'W_COMPLETE': 100.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': False, 'ent_coef': 0.02}
... [trial 18 | 1 sheep | 50,000 steps | ret(last 32)=-3.63 sr=3%]
... [trial 18 | 1 sheep | 100,000 steps | ret(last 50)=-2.28 sr=12%]
... [trial 18 | 1 sheep | 150,000 steps | ret(last 50)=-3.15 sr=10%]
... [trial 18 | 1 sheep | 200,000 steps | ret(last 50)=-3.31 sr=6%]
... [trial 18 | 1 sheep | 250,000 steps | ret(last 50)=-3.23 sr=2%]
... [trial 18 | 1 sheep | 300,000 steps | ret(last 50)=+3.55 sr=22%]
... [trial 18 | 1 sheep | 350,000 steps | ret(last 50)=+8.15 sr=28%]
... [trial 18 | 1 sheep | 400,000 steps | ret(last 50)=+10.56 sr=18%]
... [trial 18 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 18 | 2 sheep | 459,608 steps | ret(last 34)=+3.80 sr=0%]
... [trial 18 | 2 sheep | 509,608 steps | ret(last 50)=+7.30 sr=4%]
... [trial 18 | 2 sheep | 559,608 steps | ret(last 50)=+9.61 sr=10%]
... [trial 18 | 2 sheep | 609,608 steps | ret(last 50)=+7.70 sr=8%]
... [trial 18 | 2 sheep | 659,608 steps | ret(last 50)=+6.01 sr=2%]
... [trial 18 | 2 sheep | 709,608 steps | ret(last 50)=+8.28 sr=6%]
... [trial 18 | 2 sheep | 759,608 steps | ret(last 50)=+6.74 sr=0%]
... [trial 18 | 2 sheep | 809,608 steps | ret(last 50)=+10.61 sr=0%]
... [trial 18 | 2 sheep | 859,608 steps | ret(last 50)=+12.20 sr=0%]
... [trial 18 | 2 sheep | 909,608 steps | ret(last 50)=+11.25 sr=2%]
... [trial 18 | 2 sheep | 959,608 steps | ret(last 50)=+13.58 sr=4%]
... [trial 18 | 2 sheep | 1,009,608 steps | ret(last 50)=+16.61 sr=20%]
... [trial 18 | eval n=1]
... [trial 18 | eval n=2]
... [trial 18 | eval n=3]
→ score=0.160 sr1=0.30 sr2=0.20 sr3=0.00 [316s]
[Trial 19] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.025, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 50.0, 'W_COMPACT': 0.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': True, 'ent_coef': 0.005}
... [trial 19 | 1 sheep | 50,000 steps | ret(last 32)=-36.89 sr=3%]
... [trial 19 | 1 sheep | 100,000 steps | ret(last 50)=-30.93 sr=4%]
... [trial 19 | 1 sheep | 150,000 steps | ret(last 50)=-28.35 sr=12%]
... [trial 19 | 1 sheep | 200,000 steps | ret(last 50)=-30.73 sr=8%]
... [trial 19 | 1 sheep | 250,000 steps | ret(last 50)=-29.54 sr=4%]
... [trial 19 | 1 sheep | 300,000 steps | ret(last 50)=-20.15 sr=20%]
... [trial 19 | 1 sheep | 350,000 steps | ret(last 50)=-0.07 sr=68%]
... [trial 19 | 1 sheep | 400,000 steps | ret(last 50)=+1.66 sr=52%]
... [trial 19 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 19 | 2 sheep | 459,608 steps | ret(last 36)=-12.82 sr=19%]
... [trial 19 | 2 sheep | 509,608 steps | ret(last 50)=-20.66 sr=0%]
... [trial 19 | 2 sheep | 559,608 steps | ret(last 50)=-16.54 sr=4%]
... [trial 19 | 2 sheep | 609,608 steps | ret(last 50)=-17.11 sr=4%]
... [trial 19 | 2 sheep | 659,608 steps | ret(last 50)=-19.32 sr=0%]
... [trial 19 | 2 sheep | 709,608 steps | ret(last 50)=-16.20 sr=0%]
... [trial 19 | 2 sheep | 759,608 steps | ret(last 50)=-13.12 sr=2%]
... [trial 19 | 2 sheep | 809,608 steps | ret(last 50)=-17.18 sr=4%]
... [trial 19 | 2 sheep | 859,608 steps | ret(last 50)=-18.16 sr=2%]
... [trial 19 | 2 sheep | 909,608 steps | ret(last 50)=-18.12 sr=4%]
... [trial 19 | 2 sheep | 959,608 steps | ret(last 50)=-17.79 sr=2%]
... [trial 19 | 2 sheep | 1,009,608 steps | ret(last 50)=-17.58 sr=0%]
... [trial 19 | eval n=1]
... [trial 19 | eval n=2]
... [trial 19 | eval n=3]
→ score=0.160 sr1=0.80 sr2=0.00 sr3=0.00 [318s]
[Trial 20] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 100.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.02}
... [trial 20 | 1 sheep | 50,000 steps | ret(last 33)=-15.83 sr=9%]
... [trial 20 | 1 sheep | 100,000 steps | ret(last 50)=-18.74 sr=10%]
... [trial 20 | 1 sheep | 150,000 steps | ret(last 50)=-22.88 sr=6%]
... [trial 20 | 1 sheep | 200,000 steps | ret(last 50)=-23.86 sr=4%]
... [trial 20 | 1 sheep | 250,000 steps | ret(last 50)=-21.10 sr=6%]
... [trial 20 | 1 sheep | 300,000 steps | ret(last 50)=-18.42 sr=6%]
... [trial 20 | 1 sheep | 350,000 steps | ret(last 50)=+1.74 sr=14%]
... [trial 20 | 1 sheep | 400,000 steps | ret(last 50)=+7.62 sr=34%]
... [trial 20 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 20 | 2 sheep | 459,608 steps | ret(last 34)=-2.63 sr=3%]
... [trial 20 | 2 sheep | 509,608 steps | ret(last 50)=+1.10 sr=2%]
... [trial 20 | 2 sheep | 559,608 steps | ret(last 50)=+5.57 sr=4%]
... [trial 20 | 2 sheep | 609,608 steps | ret(last 50)=+8.54 sr=8%]
... [trial 20 | 2 sheep | 659,608 steps | ret(last 50)=+12.02 sr=8%]
... [trial 20 | 2 sheep | 709,608 steps | ret(last 50)=+11.28 sr=4%]
... [trial 20 | 2 sheep | 759,608 steps | ret(last 50)=+11.45 sr=2%]
... [trial 20 | 2 sheep | 809,608 steps | ret(last 50)=+9.52 sr=0%]
... [trial 20 | 2 sheep | 859,608 steps | ret(last 50)=+9.07 sr=2%]
... [trial 20 | 2 sheep | 909,608 steps | ret(last 50)=+12.06 sr=8%]
... [trial 20 | 2 sheep | 959,608 steps | ret(last 50)=+12.77 sr=8%]
... [trial 20 | 2 sheep | 1,009,608 steps | ret(last 50)=+11.55 sr=2%]
... [trial 20 | eval n=1]
... [trial 20 | eval n=2]
... [trial 20 | eval n=3]
→ score=0.130 sr1=0.40 sr2=0.10 sr3=0.00 [315s]
[Trial 21] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 100.0, 'W_COMPACT': 0.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': True, 'ent_coef': 0.005}
... [trial 21 | 1 sheep | 50,000 steps | ret(last 32)=-14.94 sr=6%]
... [trial 21 | 1 sheep | 100,000 steps | ret(last 50)=-12.47 sr=4%]
... [trial 21 | 1 sheep | 150,000 steps | ret(last 50)=-12.65 sr=6%]
... [trial 21 | 1 sheep | 200,000 steps | ret(last 50)=-12.44 sr=2%]
... [trial 21 | 1 sheep | 250,000 steps | ret(last 50)=-12.95 sr=6%]
... [trial 21 | 1 sheep | 300,000 steps | ret(last 50)=-13.04 sr=6%]
... [trial 21 | 1 sheep | 350,000 steps | ret(last 50)=-5.14 sr=8%]
... [trial 21 | 1 sheep | 400,000 steps | ret(last 50)=-0.46 sr=8%]
... [trial 21 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 21 | 2 sheep | 459,608 steps | ret(last 33)=-7.10 sr=0%]
... [trial 21 | 2 sheep | 509,608 steps | ret(last 50)=-8.26 sr=0%]
... [trial 21 | 2 sheep | 559,608 steps | ret(last 50)=-6.17 sr=4%]
... [trial 21 | 2 sheep | 609,608 steps | ret(last 50)=-4.23 sr=4%]
... [trial 21 | 2 sheep | 659,608 steps | ret(last 50)=-5.62 sr=0%]
... [trial 21 | 2 sheep | 709,608 steps | ret(last 50)=-3.72 sr=0%]
... [trial 21 | 2 sheep | 759,608 steps | ret(last 50)=-2.06 sr=0%]
... [trial 21 | 2 sheep | 809,608 steps | ret(last 50)=-1.23 sr=0%]
... [trial 21 | 2 sheep | 859,608 steps | ret(last 50)=-0.14 sr=0%]
... [trial 21 | 2 sheep | 909,608 steps | ret(last 50)=+1.30 sr=2%]
... [trial 21 | 2 sheep | 959,608 steps | ret(last 50)=+0.64 sr=2%]
... [trial 21 | 2 sheep | 1,009,608 steps | ret(last 50)=+2.62 sr=6%]
... [trial 21 | eval n=1]
... [trial 21 | eval n=2]
... [trial 21 | eval n=3]
→ score=0.050 sr1=0.00 sr2=0.10 sr3=0.00 [310s]
[Trial 22] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.05, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 100.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': True, 'ent_coef': 0.005}
... [trial 22 | 1 sheep | 50,000 steps | ret(last 32)=-11.10 sr=6%]
... [trial 22 | 1 sheep | 100,000 steps | ret(last 50)=-10.61 sr=8%]
... [trial 22 | 1 sheep | 150,000 steps | ret(last 50)=-11.16 sr=4%]
... [trial 22 | 1 sheep | 200,000 steps | ret(last 50)=-11.15 sr=4%]
... [trial 22 | 1 sheep | 250,000 steps | ret(last 50)=-10.56 sr=6%]
... [trial 22 | 1 sheep | 300,000 steps | ret(last 50)=-14.90 sr=0%]
... [trial 22 | 1 sheep | 350,000 steps | ret(last 50)=-5.11 sr=14%]
... [trial 22 | 1 sheep | 400,000 steps | ret(last 50)=+2.22 sr=24%]
... [trial 22 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 22 | 2 sheep | 459,608 steps | ret(last 35)=-4.69 sr=6%]
... [trial 22 | 2 sheep | 509,608 steps | ret(last 50)=-3.17 sr=0%]
... [trial 22 | 2 sheep | 559,608 steps | ret(last 50)=+2.18 sr=2%]
... [trial 22 | 2 sheep | 609,608 steps | ret(last 50)=+4.53 sr=8%]
... [trial 22 | 2 sheep | 659,608 steps | ret(last 50)=+4.97 sr=10%]
... [trial 22 | 2 sheep | 709,608 steps | ret(last 50)=+5.06 sr=8%]
... [trial 22 | 2 sheep | 759,608 steps | ret(last 50)=+6.04 sr=4%]
... [trial 22 | 2 sheep | 809,608 steps | ret(last 50)=+5.95 sr=4%]
... [trial 22 | 2 sheep | 859,608 steps | ret(last 50)=+3.34 sr=2%]
... [trial 22 | 2 sheep | 909,608 steps | ret(last 50)=+6.80 sr=8%]
... [trial 22 | 2 sheep | 959,608 steps | ret(last 50)=+4.13 sr=8%]
... [trial 22 | 2 sheep | 1,009,608 steps | ret(last 50)=+4.17 sr=2%]
... [trial 22 | eval n=1]
... [trial 22 | eval n=2]
... [trial 22 | eval n=3]
→ score=0.110 sr1=0.30 sr2=0.10 sr3=0.00 [316s]
[Trial 23] {'W_PER_SHEEP': 2.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 100.0, 'W_COMPACT': 1.5, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': True, 'ent_coef': 0.05}
... [trial 23 | 1 sheep | 50,000 steps | ret(last 32)=-22.59 sr=9%]
... [trial 23 | 1 sheep | 100,000 steps | ret(last 50)=-21.14 sr=6%]
... [trial 23 | 1 sheep | 150,000 steps | ret(last 50)=-20.75 sr=6%]
... [trial 23 | 1 sheep | 200,000 steps | ret(last 50)=-20.37 sr=8%]
... [trial 23 | 1 sheep | 250,000 steps | ret(last 50)=-5.04 sr=18%]
... [trial 23 | 1 sheep | 300,000 steps | ret(last 50)=+7.25 sr=12%]
... [trial 23 | 1 sheep | 350,000 steps | ret(last 50)=+11.34 sr=32%]
... [trial 23 | 1 sheep | 400,000 steps | ret(last 50)=+13.02 sr=24%]
... [trial 23 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 23 | 2 sheep | 459,608 steps | ret(last 32)=+0.29 sr=3%]
... [trial 23 | 2 sheep | 509,608 steps | ret(last 50)=-0.39 sr=4%]
... [trial 23 | 2 sheep | 559,608 steps | ret(last 50)=+6.56 sr=2%]
... [trial 23 | 2 sheep | 609,608 steps | ret(last 50)=+10.45 sr=2%]
... [trial 23 | 2 sheep | 659,608 steps | ret(last 50)=+9.75 sr=2%]
... [trial 23 | 2 sheep | 709,608 steps | ret(last 50)=+7.98 sr=6%]
... [trial 23 | 2 sheep | 759,608 steps | ret(last 50)=+9.20 sr=4%]
... [trial 23 | 2 sheep | 809,608 steps | ret(last 50)=+11.03 sr=6%]
... [trial 23 | 2 sheep | 859,608 steps | ret(last 50)=+12.53 sr=6%]
... [trial 23 | 2 sheep | 909,608 steps | ret(last 50)=+10.86 sr=6%]
... [trial 23 | 2 sheep | 959,608 steps | ret(last 50)=+13.16 sr=14%]
... [trial 23 | 2 sheep | 1,009,608 steps | ret(last 50)=+12.36 sr=12%]
... [trial 23 | eval n=1]
... [trial 23 | eval n=2]
... [trial 23 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [472s]
[Trial 24] {'W_PER_SHEEP': 6.0, 'W_ALIGN': 0.025, 'W_PEN_BONUS': 20.0, 'W_STEP_COST': 0.005, 'W_COMPLETE': 200.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.01}
... [trial 24 | 1 sheep | 50,000 steps | ret(last 32)=-1.97 sr=0%]
... [trial 24 | 1 sheep | 100,000 steps | ret(last 50)=-1.86 sr=2%]
... [trial 24 | 1 sheep | 150,000 steps | ret(last 50)=-2.97 sr=4%]
... [trial 24 | 1 sheep | 200,000 steps | ret(last 50)=-0.45 sr=8%]
... [trial 24 | 1 sheep | 250,000 steps | ret(last 50)=-1.73 sr=4%]
... [trial 24 | 1 sheep | 300,000 steps | ret(last 50)=+0.64 sr=4%]
... [trial 24 | 1 sheep | 350,000 steps | ret(last 50)=+1.35 sr=2%]
... [trial 24 | 1 sheep | 400,000 steps | ret(last 50)=+0.95 sr=4%]
... [trial 24 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 24 | 2 sheep | 459,608 steps | ret(last 33)=+1.34 sr=0%]
... [trial 24 | 2 sheep | 509,608 steps | ret(last 50)=+1.48 sr=0%]
... [trial 24 | 2 sheep | 559,608 steps | ret(last 50)=+6.05 sr=0%]
... [trial 24 | 2 sheep | 609,608 steps | ret(last 50)=+3.58 sr=0%]
... [trial 24 | 2 sheep | 659,608 steps | ret(last 50)=+2.33 sr=0%]
... [trial 24 | 2 sheep | 709,608 steps | ret(last 50)=+4.05 sr=2%]
... [trial 24 | 2 sheep | 759,608 steps | ret(last 50)=+0.93 sr=0%]
... [trial 24 | 2 sheep | 809,608 steps | ret(last 50)=-0.39 sr=0%]
... [trial 24 | 2 sheep | 859,608 steps | ret(last 50)=-2.68 sr=0%]
... [trial 24 | 2 sheep | 909,608 steps | ret(last 50)=+0.90 sr=0%]
... [trial 24 | 2 sheep | 959,608 steps | ret(last 50)=+2.63 sr=0%]
... [trial 24 | 2 sheep | 1,009,608 steps | ret(last 50)=+2.88 sr=0%]
... [trial 24 | eval n=1]
... [trial 24 | eval n=2]
... [trial 24 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [335s]
[Trial 25] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.0, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.05, 'W_COMPLETE': 50.0, 'W_COMPACT': 0.5, 'ALIGN_SHAPE': 'near', 'ALIGN_GATED': True, 'ent_coef': 0.02}
... [trial 25 | 1 sheep | 50,000 steps | ret(last 32)=-56.03 sr=3%]
... [trial 25 | 1 sheep | 100,000 steps | ret(last 50)=-53.61 sr=4%]
... [trial 25 | 1 sheep | 150,000 steps | ret(last 50)=-54.50 sr=4%]
... [trial 25 | 1 sheep | 200,000 steps | ret(last 50)=-57.55 sr=4%]
... [trial 25 | 1 sheep | 250,000 steps | ret(last 50)=-54.77 sr=8%]
... [trial 25 | 1 sheep | 300,000 steps | ret(last 50)=-55.53 sr=4%]
... [trial 25 | 1 sheep | 350,000 steps | ret(last 50)=-55.26 sr=4%]
... [trial 25 | 1 sheep | 400,000 steps | ret(last 50)=-56.11 sr=4%]
... [trial 25 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 25 | 2 sheep | 459,608 steps | ret(last 32)=-48.36 sr=0%]
... [trial 25 | 2 sheep | 509,608 steps | ret(last 50)=-54.87 sr=0%]
... [trial 25 | 2 sheep | 559,608 steps | ret(last 50)=-56.08 sr=0%]
... [trial 25 | 2 sheep | 609,608 steps | ret(last 50)=-54.86 sr=0%]
... [trial 25 | 2 sheep | 659,608 steps | ret(last 50)=-50.62 sr=0%]
... [trial 25 | 2 sheep | 709,608 steps | ret(last 50)=-49.92 sr=0%]
... [trial 25 | 2 sheep | 759,608 steps | ret(last 50)=-50.11 sr=0%]
... [trial 25 | 2 sheep | 809,608 steps | ret(last 50)=-51.41 sr=0%]
... [trial 25 | 2 sheep | 859,608 steps | ret(last 50)=-51.02 sr=0%]
... [trial 25 | 2 sheep | 909,608 steps | ret(last 50)=-50.80 sr=0%]
... [trial 25 | 2 sheep | 959,608 steps | ret(last 50)=-50.01 sr=0%]
... [trial 25 | 2 sheep | 1,009,608 steps | ret(last 50)=-49.71 sr=0%]
... [trial 25 | eval n=1]
... [trial 25 | eval n=2]
... [trial 25 | eval n=3]
→ score=0.000 sr1=0.00 sr2=0.00 sr3=0.00 [306s]
============================================================================================
LEADERBOARD
============================================================================================
rank score sr1 sr2 sr3 config
----------------------------------------------------------------------------------------
1 0.350 1.00 0.30 0.00 W_PER_SHEEP=1.0 W_ALIGN=0.0 W_PEN_BONUS=5.0 W_STEP_COST=0.02 W_COMPLETE=200.0 W_COMPACT=1.5 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.02
2 0.270 0.70 0.20 0.10 W_PER_SHEEP=6.0 W_ALIGN=0.025 W_PEN_BONUS=10.0 W_STEP_COST=0.02 W_COMPLETE=50.0 W_COMPACT=3.0 ALIGN_SHAPE=near ALIGN_GATED=False ent_coef=0.01
3 0.240 0.70 0.20 0.00 W_PER_SHEEP=1.0 W_ALIGN=0.1 W_PEN_BONUS=5.0 W_STEP_COST=0.05 W_COMPLETE=50.0 W_COMPACT=3.0 ALIGN_SHAPE=near ALIGN_GATED=False ent_coef=0.02
4 0.240 0.70 0.20 0.00 W_PER_SHEEP=6.0 W_ALIGN=0.1 W_PEN_BONUS=5.0 W_STEP_COST=0.005 W_COMPLETE=200.0 W_COMPACT=0.0 ALIGN_SHAPE=near ALIGN_GATED=True ent_coef=0.05
5 0.200 1.00 0.00 0.00 W_PER_SHEEP=6.0 W_ALIGN=0.1 W_PEN_BONUS=5.0 W_STEP_COST=0.02 W_COMPLETE=50.0 W_COMPACT=3.0 ALIGN_SHAPE=near ALIGN_GATED=True ent_coef=0.005
6 0.190 0.70 0.10 0.00 W_PER_SHEEP=2.0 W_ALIGN=0.0 W_PEN_BONUS=20.0 W_STEP_COST=0.02 W_COMPLETE=50.0 W_COMPACT=0.0 ALIGN_SHAPE=near ALIGN_GATED=False ent_coef=0.05
7 0.160 0.80 0.00 0.00 W_PER_SHEEP=6.0 W_ALIGN=0.025 W_PEN_BONUS=20.0 W_STEP_COST=0.05 W_COMPLETE=200.0 W_COMPACT=3.0 ALIGN_SHAPE=standoff ALIGN_GATED=True ent_coef=0.01
8 0.160 0.80 0.00 0.00 W_PER_SHEEP=2.0 W_ALIGN=0.1 W_PEN_BONUS=20.0 W_STEP_COST=0.02 W_COMPLETE=200.0 W_COMPACT=0.5 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.02
9 0.160 0.80 0.00 0.00 W_PER_SHEEP=2.0 W_ALIGN=0.025 W_PEN_BONUS=10.0 W_STEP_COST=0.05 W_COMPLETE=50.0 W_COMPACT=0.0 ALIGN_SHAPE=standoff ALIGN_GATED=True ent_coef=0.005
10 0.160 0.30 0.20 0.00 W_PER_SHEEP=2.0 W_ALIGN=0.025 W_PEN_BONUS=10.0 W_STEP_COST=0.005 W_COMPLETE=100.0 W_COMPACT=1.5 ALIGN_SHAPE=near ALIGN_GATED=False ent_coef=0.02
11 0.150 0.50 0.10 0.00 W_PER_SHEEP=1.0 W_ALIGN=0.1 W_PEN_BONUS=10.0 W_STEP_COST=0.02 W_COMPLETE=100.0 W_COMPACT=3.0 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.005
12 0.140 0.70 0.00 0.00 W_PER_SHEEP=1.0 W_ALIGN=0.1 W_PEN_BONUS=10.0 W_STEP_COST=0.02 W_COMPLETE=100.0 W_COMPACT=1.5 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.05
13 0.130 0.40 0.10 0.00 W_PER_SHEEP=1.0 W_ALIGN=0.1 W_PEN_BONUS=20.0 W_STEP_COST=0.05 W_COMPLETE=100.0 W_COMPACT=1.5 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.02
14 0.110 0.30 0.10 0.00 W_PER_SHEEP=6.0 W_ALIGN=0.025 W_PEN_BONUS=5.0 W_STEP_COST=0.05 W_COMPLETE=100.0 W_COMPACT=0.0 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.005
15 0.110 0.30 0.10 0.00 W_PER_SHEEP=2.0 W_ALIGN=0.05 W_PEN_BONUS=10.0 W_STEP_COST=0.02 W_COMPLETE=100.0 W_COMPACT=3.0 ALIGN_SHAPE=standoff ALIGN_GATED=True ent_coef=0.005
Best config saved to runs/sweep_20260425_124630/best.json
Total trials: 25 (25 successful, 0 failed)
Total time: 2.28h
training/runs/sweep_smoke.log
-43
View File
@@ -1,43 +0,0 @@
Sweep dir: runs/sweep_20260425_124021
Search space: ['W_PER_SHEEP', 'W_ALIGN', 'W_PEN_BONUS', 'W_STEP_COST', 'W_COMPLETE', 'W_COMPACT', 'ALIGN_SHAPE', 'ALIGN_GATED', 'ent_coef']
Per-trial: 1,000,000 steps train + 30 eval eps
Time budget: 0.5h
[Trial 1] {'W_PER_SHEEP': 1.0, 'W_ALIGN': 0.1, 'W_PEN_BONUS': 10.0, 'W_STEP_COST': 0.02, 'W_COMPLETE': 100.0, 'W_COMPACT': 3.0, 'ALIGN_SHAPE': 'standoff', 'ALIGN_GATED': False, 'ent_coef': 0.005}
... [trial 1 | 1 sheep | 50,000 steps | ret(last 32)=-8.33 sr=6%]
... [trial 1 | 1 sheep | 100,000 steps | ret(last 50)=-2.95 sr=6%]
... [trial 1 | 1 sheep | 150,000 steps | ret(last 50)=+12.68 sr=10%]
... [trial 1 | 1 sheep | 200,000 steps | ret(last 50)=+22.15 sr=22%]
... [trial 1 | 1 sheep | 250,000 steps | ret(last 50)=+22.47 sr=18%]
... [trial 1 | 1 sheep | 300,000 steps | ret(last 50)=+23.58 sr=24%]
... [trial 1 | 1 sheep | 350,000 steps | ret(last 50)=+23.42 sr=18%]
... [trial 1 | 1 sheep | 400,000 steps | ret(last 50)=+24.39 sr=32%]
... [trial 1 | 2 sheep | 409,608 steps | ret(last 0)=+nan sr=nan%]
... [trial 1 | 2 sheep | 459,608 steps | ret(last 35)=+15.39 sr=3%]
... [trial 1 | 2 sheep | 509,608 steps | ret(last 50)=+20.25 sr=0%]
... [trial 1 | 2 sheep | 559,608 steps | ret(last 50)=+23.24 sr=4%]
... [trial 1 | 2 sheep | 609,608 steps | ret(last 50)=+23.36 sr=4%]
... [trial 1 | 2 sheep | 659,608 steps | ret(last 50)=+25.32 sr=2%]
... [trial 1 | 2 sheep | 709,608 steps | ret(last 50)=+24.02 sr=4%]
... [trial 1 | 2 sheep | 759,608 steps | ret(last 50)=+24.66 sr=4%]
... [trial 1 | 2 sheep | 809,608 steps | ret(last 50)=+25.41 sr=4%]
... [trial 1 | 2 sheep | 859,608 steps | ret(last 50)=+24.27 sr=4%]
... [trial 1 | 2 sheep | 909,608 steps | ret(last 50)=+25.13 sr=8%]
... [trial 1 | 2 sheep | 959,608 steps | ret(last 50)=+25.10 sr=2%]
... [trial 1 | 2 sheep | 1,009,608 steps | ret(last 50)=+26.02 sr=2%]
... [trial 1 | eval n=1]
... [trial 1 | eval n=2]
... [trial 1 | eval n=3]
→ score=0.060 sr1=0.30 sr2=0.00 sr3=0.00 [308s]
============================================================================================
LEADERBOARD
============================================================================================
rank score sr1 sr2 sr3 config
----------------------------------------------------------------------------------------
1 0.060 0.30 0.00 0.00 W_PER_SHEEP=1.0 W_ALIGN=0.1 W_PEN_BONUS=10.0 W_STEP_COST=0.02 W_COMPLETE=100.0 W_COMPACT=3.0 ALIGN_SHAPE=standoff ALIGN_GATED=False ent_coef=0.005
Best config saved to runs/sweep_20260425_124021/best.json
Total trials: 1 (1 successful, 0 failed)
Total time: 0.09h