Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Verify performance of baseline #9

Open
emilio-cartoni opened this issue Apr 29, 2020 · 11 comments
Open

Verify performance of baseline #9

emilio-cartoni opened this issue Apr 29, 2020 · 11 comments
Assignees

Comments

@emilio-cartoni
Copy link
Owner

Keep an eye on how the baseline performs during integration.

@emilio-cartoni
Copy link
Owner Author

On commit: 3cbe8d3

NOPLAN
100001steps [16:56, 1982.66steps /s]{'score_REAL2020': 0.09163811733270416

BASELINE
1500001steps 19:39, 2216.19steps /s 0.10718097341998632,
15000001steps 2:20:49, 2269.06steps /s 0.17869694419642482

@emilio-cartoni
Copy link
Owner Author

On commit: a68a9ca

BASELINE
with experience file filtered_transitions_15000000_12486_1_4170_466197_60x128_pos_noq_4_xy_push.npy
Extrinsic Phase: 100%|█| 50/50 [16:54<00:00, 20.59s/trials , score_REAL2020=0.09{'score_REAL2020': 0.09163811733270416

without:
Intrinsic Phase: 15000001steps [2:37:03, 2127.19steps /s]{'score_REAL2020': 0.1628960306426154,
Intrinsic Phase: 15000001steps [2:40:07, 2080.36steps /s]{'score_REAL2020': 0.12885043929342968,
Intrinsic Phase: 15000001steps [2:40:13, 1938.87steps /s] {'score_REAL2020': 0.15607540637037715,

NOPLAN
Extrinsic Phase: 100%|█| 50/50 [06:27<00:00, 6.91s/trials , score_REAL2020=0.04{'score_REAL2020': 0.0401959478225
Extrinsic Phase: 100%|█| 50/50 [06:11<00:00, 7.28s/trials , score_REAL2020=0.04{'score_REAL2020': 0.040195947822584996
Extrinsic Phase: 100%█| 50/50 [05:46<00:00, 6.61s/trials , score_REAL2020=0.040{'score_REAL2020': 0.040195947822584996,

Extrinsic Phase: 100%|█| 50/50 [05:21<00:00, 6.01s/trials , score_REAL2020=0.04{'score_REAL2020': 0.040195947822584996
Intrinsic Phase: 100001steps [06:06, 1935.13steps /s]{'score_REAL2020': 0.040195947822584996

@DavideMontella
Copy link
Contributor

14 tests done with an average of 0.173

@emilio-cartoni
Copy link
Owner Author

14 tests done with an average of 0.173
With a 15M timestep intrinsic phase?

Let's keep this issue open until we release the Starting Kit.

@emilio-cartoni emilio-cartoni reopened this May 5, 2020
@DavideMontella
Copy link
Contributor

Yes, with 15M

@emilio-cartoni
Copy link
Owner Author

15M, with master ba121b3 and real_robots REAL2020 436f174
0.141
0.167
0.149
0.186

@emilio-cartoni
Copy link
Owner Author

Starter kit: 5a95163
real_robots: 0036124

NOPLAN, 1 object - 0.009
score_3D': 0.0065647186052345107, 'score_total': 0.0093648237362823957, 'score_2.5D': 0.0042436305504342259, 'score_2D': 0.013557581700210451
array({'2D': [0.0019238853468477371, 0.014646116762107234, 0.048329430719743632, 2.5206305740121675e-05, 0.037496311650508624, 0.0052332436507228977, 0.020380452399946389, 0.059569587310970271, 0.00018916903745471261, 0.027199084161152232, 0.00013083856632348473, 0.00021494142303393095, 0.00018921974232565845, 0.00026355219019182904, 9.6823605526141426e-05, 0.00010775576218115716, 0.0082696999888048437, 0.017498710185933061, 5.1160499418949119e-05, 0.047631148890843966, 1.0097519259201269e-05, 0.0086437200647780325, 0.0088970949253597779, 0.031326907584173717, 0.00061538421191363027], '3D': [0.011075257460490333, 0.01813792883665712, 1.127737925253152e-05, 0.002186923870105011, 0.0043489824021749647, 0.0068826776417246546, 0.0005238670337095303, 5.8904818376775313e-05, 0.0093045396734613792, 0.013116826936392812], '2.5D': [2.0418761466028717e-05, 5.5672870882952706e-05, 1.2744765160688819e-05, 0.0082813724892642215, 0.00025559948778319992, 0.0027008181263109081, 0.0018580968837133597, 0.0061901565422679749, 0.0050607456863446185, 0.015931357834323388, 0.0023816306365511955, 0.019474455390238496, 0.00052397305214397041, 0.00016237765792980274, 0.00074503807213260008]}

NOPLAN, 2 objects - 0.024
{'score_2D': 0.024226881473714795, 'score_2.5D': 0.016296566911933392, 'score_3D': 0.035530497474211886, 'score_total': 0.024108510305279798}
Extrinsic Phase: 100%|████████████████████████████████████████████████████████████████████████| 50/50 [03:44<00:00, 4.85s/trials , score_2D=0.0242, score_2.5D=0.0163, score_3D=0.0355, score_total=0.0241]{'score_2D': 0.024226881473714795, 'score_2.5D': 0.016296566911933392, 'score_3D': 0.035530497474211886, 'score_total': 0.024108510305279798}
{'2D': [0.04458546856689659, 0.05246327490573012, 0.002295162957392198, 0.011582211060486238, 0.06499581044790674, 0.0022169912438870856, 0.03536981266533803, 0.006013415339186251, 0.0019519270054632037, 0.013626183798269045, 0.028359984499848338, 0.04129984093848884, 0.004407287278455697, 0.007074204490545617, 0.03927022132766552, 0.04483449291349705, 0.0014029232885688107, 0.0027401458938103975, 0.026705049077939143, 0.04298895698912243, 0.03617689778436479, 0.002891152177613947, 0.05803325151931604, 0.030651966858386726, 0.003735403814691079], '2.5D': [0.05215592364978039, 0.0074181760122556525, 0.034489967643338955, 0.0035723360067399035, 0.026273182886867412, 0.001724153946449026, 0.0005874009228253276, 0.05010126750226719, 0.039925112194027695, 6.598735483097037e-05, 0.0035658941640465073, 0.008000714040828185, 0.007544086446980664, 0.00816672415604296, 0.0008575767517200522], '3D': [0.08368598908699233, 0.0014508271592979475, 0.0012270512967539368, 0.04189886900227845, 0.06732884570142209, 0.00577885972551587, 0.06961957360965254, 0.027153361437618997, 0.037816011408767385, 0.01934558631381934]}

NOPLAN, 3 objects - 0.032
{'score_2D': 0.03051578470205347, 'score_2.5D': 0.025829095121291883, 'score_3D': 0.044327031915719795, 'score_total': 0.031872027270558254}
Extrinsic Phase: 100%|█| 50/50 [04:09<00:00, 4.80s/trials , score_2D=0.0305, sc{'score_2D': 0.03051578470205347, 'score_2.5D': 0.025829095121291883, 'score_3D': 0.044327031915719795, 'score_total': 0.031872027270558254}
{'2D': [0.023407293743322385, 0.009036193117741823, 0.007650961482459576, 0.042759588832695414, 0.035179160612573104, 0.0651086841112424, 0.025996322135482287, 0.11957366830714725, 0.04627149998643088, 0.005769387505534589, 0.030040779020443747, 0.03247219049392795, 0.011275983111967536, 0.004955345708258985, 0.00233862474616395, 0.012993308139938467, 0.032381644175570726, 0.013654299989028625, 0.004969002424046994, 0.03704709211113489, 0.05262638189922506, 0.06495127496239983, 0.05462342895434255, 0.02327771445287137, 0.0045347875273863], '2.5D': [0.004944019771538153, 0.02209958920744585, 0.01885953302519892, 0.06425402535085971, 0.024458757733159434, 0.0370570225578504, 0.023645182439674096, 0.08186653809662953, 0.019330573353242675, 0.009067143684545797, 0.019197877821694215, 0.005652205992977563, 0.04177477892279089, 0.010182815239401531, 0.005046363622369453], '3D': [0.042532479418119995, 0.009111523146351275, 0.04610965053003911, 0.031354290541203436, 0.11846204819099314, 0.010562747726175289, 0.03885998336239692, 0.014964287752484338, 0.01896246643937177, 0.11235084205006265]}

RANDOM -- in progress

BASELINE, 1 object (0.201 avg)
8 runs:
0.189
0.168
0.224
0.205
0.184
0.190
0.259
0.189

@emilio-cartoni
Copy link
Owner Author

Starter kit: 5a95163
real_robots: 0036124

  RND 1 obj RND 2 obj RND 3 obj BASE 1 obj BASE 2 obj BASE 3 obj
N. Sim 108 108 224 110 100 40
Avg 0.022 0.060 0.101 0.211 0.095 0.127
Max 0.044 0.128 0.179 0.289 0.160 0.184
Min 0.008 0.029 0.055 0.143 0.052 0.088

@emilio-cartoni
Copy link
Owner Author

emilio-cartoni commented Jul 22, 2020

With Fixes on planner and environment (double action):
REAL2020: 56e4b3d
real_robots: e09c89e

  BASE 1 obj with fix
N. Sim 21
Avg 0.211
Max 0.278
Min 0.143

@emilio-cartoni
Copy link
Owner Author

emilio-cartoni commented Jul 27, 2020

Note on memory usage:
3 simulations on the server running the extrinsinc phase after 60M intrinsic timesteps are using 30Gb RAM each.
So a raw estimate is 0.5 GB per 1M timesteps.

@emilio-cartoni
Copy link
Owner Author

f1bba9c macro_action
N. Sim 30
Avg 0.223
Max 0.319
Min 0.149

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants