Upload folder using huggingface_hub

norm_stats/scripted_vla/norm_stats.json

@@ -0,0 +1,424 @@
+{
+  "norm_stats": {
+    "state": {
+      "mean": [
+        -0.27309971270031724,
+        0.004905484382909437,
+        -0.13386459361547315,
+        0.48529911062984543,
+        -0.24865932751361425,
+        -0.28930512389760094,
+        -0.26178404685784235,
+        0.007461422507885806,
+        0.12676572204031833,
+        0.4601092997667164,
+        -0.23190117247840364,
+        -0.2735857277985144,
+        0.04392772250580358,
+        -0.24261526971723624,
+        0.2821214551212597,
+        -0.23993505560240053,
+        -0.7534921790772784,
+        -0.06688426872173536,
+        -0.346920352787036,
+        -0.46651141848397837,
+        0.4678971890349292,
+        -1.0872951825608759,
+        0.40763205749387255,
+        0.031776551335986,
+        0.22981767207020984,
+        0.8621538012564458,
+        0.007689642312380222,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -6.099089526730122e-05,
+        -0.010616776017162911,
+        0.04959205657357659,
+        0.7226000129558786,
+        0.6749624679232243,
+        0.5265897695946505,
+        0.4290600538110013,
+        0.07485820517115171,
+        0.37351606782726327,
+        0.08587537491353883,
+        -0.001158502526194975,
+        0.6493250797872192,
+        0.5838886477178913,
+        0.4235756599253438,
+        0.3878975016698497,
+        0.059125604888923246,
+        0.31539595396573583,
+        0.028261964051776924,
+        0.15689737227153616
+      ],
+      "std": [
+        0.22562385404694818,
+        0.055878635865324044,
+        0.11234611163931828,
+        0.20572244782653704,
+        0.08756771289433983,
+        0.08513699150974097,
+        0.22775815369978478,
+        0.055078512484418043,
+        0.11216106893700081,
+        0.21553065161456558,
+        0.09835439849842058,
+        0.09209628382942274,
+        0.44223412297890485,
+        0.5712298808635958,
+        0.3595397295235674,
+        0.7411274928418973,
+        0.6541956625676839,
+        0.7456575901180093,
+        0.6385217794191931,
+        0.41663016449668683,
+        0.6122139413208634,
+        0.6754705545001012,
+        0.6830047851632526,
+        0.38484146865892843,
+        0.17581577768764567,
+        0.03609169534029174,
+        0.10848019315210757,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        0.03504540757209845,
+        0.005696929912171045,
+        0.11267938674932129,
+        0.750775289393508,
+        0.7220469743412248,
+        0.5720833739840286,
+        0.46821964698881857,
+        0.09148323105547447,
+        0.4757183687817587,
+        0.37992175289511193,
+        0.2720497988284101,
+        0.7364352681906882,
+        0.686456022329896,
+        0.5219177561859668,
+        0.4772706844578171,
+        0.08361291948244559,
+        0.47594066003330476,
+        0.37641978075057586,
+        0.31355244909942276
+      ],
+      "q01": [
+        -1.0214722156524658,
+        -0.21407538652420044,
+        -0.5572712421417236,
+        4.3452386307762936e-05,
+        -0.6661560535430908,
+        -0.7587461471557617,
+        -1.0142823457717896,
+        -0.20346559584140778,
+        -0.321945458650589,
+        -2.60770320892334e-08,
+        -0.700141429901123,
+        -0.7670596837997437,
+        -1.2210134267807007,
+        -2.3868658542633057,
+        -0.09071097522974014,
+        -2.5455121994018555,
+        -2.054269552230835,
+        -2.015564441680908,
+        -2.413947105407715,
+        -2.5310006141662598,
+        -2.3441874980926514,
+        -2.0856223106384277,
+        -2.070560932159424,
+        -0.7432714700698853,
+        -0.37406808137893677,
+        0.3849436640739441,
+        -0.8728468418121338,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.6000000238418579,
+        -0.28109481930732727,
+        -0.19398415088653564,
+        -0.0018468251219019294,
+        -0.001302360906265676,
+        -0.0014056512154638767,
+        -0.0018274497706443071,
+        -0.0057253967970609665,
+        -0.10030312091112137,
+        -0.8740310668945312,
+        -0.7850599884986877,
+        -1.0377142700690456e-08,
+        -1.202005783795812e-08,
+        -2.129973211140168e-07,
+        -0.001454970333725214,
+        -0.0012382633285596967,
+        -0.1021040678024292,
+        -0.8720249533653259,
+        -0.7765935063362122
+      ],
+      "q99": [
+        0.34363824129104614,
+        0.24427013099193573,
+        0.3466421067714691,
+        1.3648877143859863,
+        0.8783490657806396,
+        0.32081493735313416,
+        0.3106701672077179,
+        0.2807782292366028,
+        0.5225675106048584,
+        1.2994697093963623,
+        0.15771573781967163,
+        0.11682254076004028,
+        1.3657941818237305,
+        1.8729076385498047,
+        2.5310020446777344,
+        2.577188491821289,
+        0.0002469148894306272,
+        2.107147216796875,
+        1.9396065473556519,
+        0.07780399918556213,
+        2.8447015285491943,
+        3.1130126444622874e-05,
+        2.04402232170105,
+        0.7775254249572754,
+        0.4236878752708435,
+        1.0139862298965454,
+        1.0972763299942017,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.6000000238418579,
+        0.09346038103103638,
+        0.2844342291355133,
+        1.6837126016616821,
+        1.6669416427612305,
+        1.6397093534469604,
+        1.6586565971374512,
+        0.3591395318508148,
+        1.3000001907348633,
+        0.8720000386238098,
+        0.7932991981506348,
+        1.6883161067962646,
+        1.6584186553955078,
+        1.6332004070281982,
+        1.661732792854309,
+        0.3594921827316284,
+        1.2999998331069946,
+        0.8732497096061707,
+        0.7862652540206909
+      ]
+    },
+    "actions": {
+      "mean": [
+        -0.000607518626098651,
+        9.575069243256767e-05,
+        -0.000523338400647283,
+        0.000482771380118743,
+        -7.412217619873375e-05,
+        -0.00024609707860446763,
+        -0.000552559865617615,
+        0.0002407901014150179,
+        0.000543595079605525,
+        0.0003058344346116087,
+        -1.8898035218716592e-05,
+        3.6310099859471206e-05,
+        0.0013687167934632021,
+        -0.010381158162729697,
+        0.009917275193115813,
+        -0.014365902052916227,
+        -0.02248508214224262,
+        -0.0012185345812013814,
+        -0.007900193931466092,
+        -0.009066601298482352,
+        0.027083508855787097,
+        -0.01863679441305178,
+        0.0026241407535363298,
+        0.002243262217304221,
+        0.003117638482182673,
+        0.8618672409479884,
+        0.007371336264507068,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -8.619638886817377e-05,
+        -0.010698499906645962,
+        0.050735584301645945,
+        -0.002087475270811249,
+        -0.00268913590802174,
+        -0.0010528136314527396,
+        -0.00064308506056111,
+        -0.00044263735871118186,
+        -0.003074053635696678,
+        0.010917825340279057,
+        -0.003578306567450126,
+        -0.002154808376316006,
+        -0.002309330380840657,
+        -0.0007215027575303472,
+        0.00016753795150093823,
+        -0.00018784873080990766,
+        -0.0038529724185196097,
+        -0.005488420650965587,
+        0.0006026267762309557
+      ],
+      "std": [
+        0.04000467416885481,
+        0.010674814805417958,
+        0.020620868982444418,
+        0.04933512212102485,
+        0.02397405403358811,
+        0.02456253850739016,
+        0.0407766522341802,
+        0.01059710863606209,
+        0.019486522918856962,
+        0.050391572490416196,
+        0.024536011794656188,
+        0.024526979462767204,
+        0.1556866514902653,
+        0.17922227589366765,
+        0.1205800341923606,
+        0.27865972678791473,
+        0.2563142950091918,
+        0.30705078479060643,
+        0.239109891385619,
+        0.17354854197106687,
+        0.3078373844062837,
+        0.2891465136687483,
+        0.30754918704721346,
+        0.12319142801992716,
+        0.03899527201215157,
+        0.036316285109359865,
+        0.10224630432020966,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        0.03343863902282533,
+        0.00596110384538068,
+        0.11346820734262787,
+        0.2825175110982551,
+        0.25695545500180544,
+        0.2259041594927186,
+        0.1922561099427415,
+        0.04423342112194952,
+        0.19382834536217233,
+        0.22473327752896358,
+        0.14476567032454835,
+        0.3814632413550959,
+        0.3312109934802243,
+        0.25871339676411914,
+        0.24692851740271055,
+        0.040948035930502766,
+        0.23448390819949425,
+        0.21804886668409365,
+        0.15103017257922885
+      ],
+      "q01": [
+        -0.7147952318191528,
+        -0.2836029827594757,
+        -0.6649898290634155,
+        -1.0648843050003052,
+        -1.3761959075927734,
+        -1.0490152835845947,
+        -0.7667303085327148,
+        -0.37227416038513184,
+        -0.7743780016899109,
+        -0.9116590023040771,
+        -0.7954862713813782,
+        -0.6387381553649902,
+        -1.0536649227142334,
+        -2.2260069847106934,
+        -1.9504908323287964,
+        -3.5244388580322266,
+        -1.8612771034240723,
+        -3.2712697982788086,
+        -2.776780128479004,
+        -1.951221227645874,
+        -3.6829452514648438,
+        -1.7789580821990967,
+        -3.261068105697632,
+        -0.9868412017822266,
+        -0.7457020282745361,
+        -0.4304202198982239,
+        -0.9550415277481079,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.6000000238418579,
+        -0.31246742606163025,
+        -0.2484288364648819,
+        -1.6826945543289185,
+        -1.6629345417022705,
+        -1.6377326250076294,
+        -1.6301100254058838,
+        -0.3591177463531494,
+        -1.305860161781311,
+        -1.7238388061523438,
+        -1.4127142429351807,
+        -1.6834478378295898,
+        -1.647825837135315,
+        -1.6030843257904053,
+        -1.661732792854309,
+        -0.3594921827316284,
+        -1.2992538213729858,
+        -1.7440000772476196,
+        -1.2653837203979492
+      ],
+      "q99": [
+        1.011109709739685,
+        0.8514817953109741,
+        0.46904006600379944,
+        1.3868345022201538,
+        1.3376530408859253,
+        1.370640754699707,
+        0.9320402145385742,
+        0.5613572597503662,
+        0.48606204986572266,
+        1.1962518692016602,
+        1.6274943351745605,
+        1.1420226097106934,
+        1.01481032371521,
+        1.457144021987915,
+        1.7250022888183594,
+        3.2938408851623535,
+        1.7842456102371216,
+        3.5122900009155273,
+        3.1115174293518066,
+        2.2157771587371826,
+        3.351306438446045,
+        1.7572286128997803,
+        3.541489362716675,
+        1.4048994779586792,
+        0.5177275538444519,
+        1.0427541732788086,
+        1.1433892250061035,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.6000000238418579,
+        1.1213501691818237,
+        0.5709735751152039,
+        1.6837127208709717,
+        1.6343809366226196,
+        1.5609186887741089,
+        1.6257437467575073,
+        0.361497163772583,
+        1.2990154027938843,
+        1.7391327619552612,
+        1.495734691619873,
+        1.6834479570388794,
+        1.656484603881836,
+        1.6322323083877563,
+        1.6257740259170532,
+        0.3592170178890228,
+        1.3917657136917114,
+        1.690704584121704,
+        1.245269775390625
+      ]
+    }
+  }
+}

norm_stats/simxr_merged_all_20260904/norm_stats.json

@@ -0,0 +1,424 @@
+{
+  "norm_stats": {
+    "state": {
+      "mean": [
+        -0.27309971270031724,
+        0.004905484382909437,
+        -0.13386459361547315,
+        0.48529911062984543,
+        -0.24865932751361425,
+        -0.28930512389760094,
+        -0.26178404685784235,
+        0.007461422507885806,
+        0.12676572204031833,
+        0.4601092997667164,
+        -0.23190117247840364,
+        -0.2735857277985144,
+        0.04392772250580358,
+        -0.24261526971723624,
+        0.2821214551212597,
+        -0.23993505560240053,
+        -0.7534921790772784,
+        -0.06688426872173536,
+        -0.346920352787036,
+        -0.46651141848397837,
+        0.4678971890349292,
+        -1.0872951825608759,
+        0.40763205749387255,
+        0.031776551335986,
+        0.22981767207020984,
+        0.8621538012564458,
+        0.007689642312380222,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -6.099089526730122e-05,
+        -0.010616776017162911,
+        0.04959205657357659,
+        0.7226000129558786,
+        0.6749624679232243,
+        0.5265897695946505,
+        0.4290600538110013,
+        0.07485820517115171,
+        0.37351606782726327,
+        0.08587537491353883,
+        -0.001158502526194975,
+        0.6493250797872192,
+        0.5838886477178913,
+        0.4235756599253438,
+        0.3878975016698497,
+        0.059125604888923246,
+        0.31539595396573583,
+        0.028261964051776924,
+        0.15689737227153616
+      ],
+      "std": [
+        0.22562385404694818,
+        0.055878635865324044,
+        0.11234611163931828,
+        0.20572244782653704,
+        0.08756771289433983,
+        0.08513699150974097,
+        0.22775815369978478,
+        0.055078512484418043,
+        0.11216106893700081,
+        0.21553065161456558,
+        0.09835439849842058,
+        0.09209628382942274,
+        0.44223412297890485,
+        0.5712298808635958,
+        0.3595397295235674,
+        0.7411274928418973,
+        0.6541956625676839,
+        0.7456575901180093,
+        0.6385217794191931,
+        0.41663016449668683,
+        0.6122139413208634,
+        0.6754705545001012,
+        0.6830047851632526,
+        0.38484146865892843,
+        0.17581577768764567,
+        0.03609169534029174,
+        0.10848019315210757,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        0.03504540757209845,
+        0.005696929912171045,
+        0.11267938674932129,
+        0.750775289393508,
+        0.7220469743412248,
+        0.5720833739840286,
+        0.46821964698881857,
+        0.09148323105547447,
+        0.4757183687817587,
+        0.37992175289511193,
+        0.2720497988284101,
+        0.7364352681906882,
+        0.686456022329896,
+        0.5219177561859668,
+        0.4772706844578171,
+        0.08361291948244559,
+        0.47594066003330476,
+        0.37641978075057586,
+        0.31355244909942276
+      ],
+      "q01": [
+        -1.0214722156524658,
+        -0.21407538652420044,
+        -0.5572712421417236,
+        4.3452386307762936e-05,
+        -0.6661560535430908,
+        -0.7587461471557617,
+        -1.0142823457717896,
+        -0.20346559584140778,
+        -0.321945458650589,
+        -2.60770320892334e-08,
+        -0.700141429901123,
+        -0.7670596837997437,
+        -1.2210134267807007,
+        -2.3868658542633057,
+        -0.09071097522974014,
+        -2.5455121994018555,
+        -2.054269552230835,
+        -2.015564441680908,
+        -2.413947105407715,
+        -2.5310006141662598,
+        -2.3441874980926514,
+        -2.0856223106384277,
+        -2.070560932159424,
+        -0.7432714700698853,
+        -0.37406808137893677,
+        0.3849436640739441,
+        -0.8728468418121338,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.6000000238418579,
+        -0.28109481930732727,
+        -0.19398415088653564,
+        -0.0018468251219019294,
+        -0.001302360906265676,
+        -0.0014056512154638767,
+        -0.0018274497706443071,
+        -0.0057253967970609665,
+        -0.10030312091112137,
+        -0.8740310668945312,
+        -0.7850599884986877,
+        -1.0377142700690456e-08,
+        -1.202005783795812e-08,
+        -2.129973211140168e-07,
+        -0.001454970333725214,
+        -0.0012382633285596967,
+        -0.1021040678024292,
+        -0.8720249533653259,
+        -0.7765935063362122
+      ],
+      "q99": [
+        0.34363824129104614,
+        0.24427013099193573,
+        0.3466421067714691,
+        1.3648877143859863,
+        0.8783490657806396,
+        0.32081493735313416,
+        0.3106701672077179,
+        0.2807782292366028,
+        0.5225675106048584,
+        1.2994697093963623,
+        0.15771573781967163,
+        0.11682254076004028,
+        1.3657941818237305,
+        1.8729076385498047,
+        2.5310020446777344,
+        2.577188491821289,
+        0.0002469148894306272,
+        2.107147216796875,
+        1.9396065473556519,
+        0.07780399918556213,
+        2.8447015285491943,
+        3.1130126444622874e-05,
+        2.04402232170105,
+        0.7775254249572754,
+        0.4236878752708435,
+        1.0139862298965454,
+        1.0972763299942017,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.6000000238418579,
+        0.09346038103103638,
+        0.2844342291355133,
+        1.6837126016616821,
+        1.6669416427612305,
+        1.6397093534469604,
+        1.6586565971374512,
+        0.3591395318508148,
+        1.3000001907348633,
+        0.8720000386238098,
+        0.7932991981506348,
+        1.6883161067962646,
+        1.6584186553955078,
+        1.6332004070281982,
+        1.661732792854309,
+        0.3594921827316284,
+        1.2999998331069946,
+        0.8732497096061707,
+        0.7862652540206909
+      ]
+    },
+    "actions": {
+      "mean": [
+        -0.000607518626098651,
+        9.575069243256767e-05,
+        -0.000523338400647283,
+        0.000482771380118743,
+        -7.412217619873375e-05,
+        -0.00024609707860446763,
+        -0.000552559865617615,
+        0.0002407901014150179,
+        0.000543595079605525,
+        0.0003058344346116087,
+        -1.8898035218716592e-05,
+        3.6310099859471206e-05,
+        0.0013687167934632021,
+        -0.010381158162729697,
+        0.009917275193115813,
+        -0.014365902052916227,
+        -0.02248508214224262,
+        -0.0012185345812013814,
+        -0.007900193931466092,
+        -0.009066601298482352,
+        0.027083508855787097,
+        -0.01863679441305178,
+        0.0026241407535363298,
+        0.002243262217304221,
+        0.003117638482182673,
+        0.8618672409479884,
+        0.007371336264507068,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -8.619638886817377e-05,
+        -0.010698499906645962,
+        0.050735584301645945,
+        -0.002087475270811249,
+        -0.00268913590802174,
+        -0.0010528136314527396,
+        -0.00064308506056111,
+        -0.00044263735871118186,
+        -0.003074053635696678,
+        0.010917825340279057,
+        -0.003578306567450126,
+        -0.002154808376316006,
+        -0.002309330380840657,
+        -0.0007215027575303472,
+        0.00016753795150093823,
+        -0.00018784873080990766,
+        -0.0038529724185196097,
+        -0.005488420650965587,
+        0.0006026267762309557
+      ],
+      "std": [
+        0.04000467416885481,
+        0.010674814805417958,
+        0.020620868982444418,
+        0.04933512212102485,
+        0.02397405403358811,
+        0.02456253850739016,
+        0.0407766522341802,
+        0.01059710863606209,
+        0.019486522918856962,
+        0.050391572490416196,
+        0.024536011794656188,
+        0.024526979462767204,
+        0.1556866514902653,
+        0.17922227589366765,
+        0.1205800341923606,
+        0.27865972678791473,
+        0.2563142950091918,
+        0.30705078479060643,
+        0.239109891385619,
+        0.17354854197106687,
+        0.3078373844062837,
+        0.2891465136687483,
+        0.30754918704721346,
+        0.12319142801992716,
+        0.03899527201215157,
+        0.036316285109359865,
+        0.10224630432020966,
+        1.0,
+        1.0,
+        1.0,
+        1.0,
+        0.03343863902282533,
+        0.00596110384538068,
+        0.11346820734262787,
+        0.2825175110982551,
+        0.25695545500180544,
+        0.2259041594927186,
+        0.1922561099427415,
+        0.04423342112194952,
+        0.19382834536217233,
+        0.22473327752896358,
+        0.14476567032454835,
+        0.3814632413550959,
+        0.3312109934802243,
+        0.25871339676411914,
+        0.24692851740271055,
+        0.040948035930502766,
+        0.23448390819949425,
+        0.21804886668409365,
+        0.15103017257922885
+      ],
+      "q01": [
+        -0.7147952318191528,
+        -0.2836029827594757,
+        -0.6649898290634155,
+        -1.0648843050003052,
+        -1.3761959075927734,
+        -1.0490152835845947,
+        -0.7667303085327148,
+        -0.37227416038513184,
+        -0.7743780016899109,
+        -0.9116590023040771,
+        -0.7954862713813782,
+        -0.6387381553649902,
+        -1.0536649227142334,
+        -2.2260069847106934,
+        -1.9504908323287964,
+        -3.5244388580322266,
+        -1.8612771034240723,
+        -3.2712697982788086,
+        -2.776780128479004,
+        -1.951221227645874,
+        -3.6829452514648438,
+        -1.7789580821990967,
+        -3.261068105697632,
+        -0.9868412017822266,
+        -0.7457020282745361,
+        -0.4304202198982239,
+        -0.9550415277481079,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.6000000238418579,
+        -0.31246742606163025,
+        -0.2484288364648819,
+        -1.6826945543289185,
+        -1.6629345417022705,
+        -1.6377326250076294,
+        -1.6301100254058838,
+        -0.3591177463531494,
+        -1.305860161781311,
+        -1.7238388061523438,
+        -1.4127142429351807,
+        -1.6834478378295898,
+        -1.647825837135315,
+        -1.6030843257904053,
+        -1.661732792854309,
+        -0.3594921827316284,
+        -1.2992538213729858,
+        -1.7440000772476196,
+        -1.2653837203979492
+      ],
+      "q99": [
+        1.011109709739685,
+        0.8514817953109741,
+        0.46904006600379944,
+        1.3868345022201538,
+        1.3376530408859253,
+        1.370640754699707,
+        0.9320402145385742,
+        0.5613572597503662,
+        0.48606204986572266,
+        1.1962518692016602,
+        1.6274943351745605,
+        1.1420226097106934,
+        1.01481032371521,
+        1.457144021987915,
+        1.7250022888183594,
+        3.2938408851623535,
+        1.7842456102371216,
+        3.5122900009155273,
+        3.1115174293518066,
+        2.2157771587371826,
+        3.351306438446045,
+        1.7572286128997803,
+        3.541489362716675,
+        1.4048994779586792,
+        0.5177275538444519,
+        1.0427541732788086,
+        1.1433892250061035,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.6000000238418579,
+        1.1213501691818237,
+        0.5709735751152039,
+        1.6837127208709717,
+        1.6343809366226196,
+        1.5609186887741089,
+        1.6257437467575073,
+        0.361497163772583,
+        1.2990154027938843,
+        1.7391327619552612,
+        1.495734691619873,
+        1.6834479570388794,
+        1.656484603881836,
+        1.6322323083877563,
+        1.6257740259170532,
+        0.3592170178890228,
+        1.3917657136917114,
+        1.690704584121704,
+        1.245269775390625
+      ]
+    }
+  }
+}

pretrained_model/config.json

@@ -0,0 +1,77 @@
+{
+    "n_obs_steps": 1,
+    "device": "cuda",
+    "use_amp": false,
+    "normalization_mode": "mean_std",
+    "obs_stride_seconds": 0.0,
+    "n_state_obs_steps": 1,
+    "state_obs_stride_seconds": 0.0,
+    "n_action_steps": 50,
+    "model_type": "qwen3vlpolicy",
+    "num_steps": 10,
+    "max_state_dim": 50,
+    "max_action_dim": 50,
+    "map_to_unified_space": false,
+    "unified_space_dim": 64,
+    "add_action_space_factorization": true,
+    "image_keys": [
+        "base_0_rgb",
+        "left_wrist_0_rgb",
+        "right_wrist_0_rgb"
+    ],
+    "tokenizer_max_length": 768,
+    "proj_width": 1280,
+    "expert_block_stride": 4,
+    "base_vlm_model": "Qwen/Qwen3.5-0.8B",
+    "cognition_mode": "frozen_vlm_context_queries",
+    "action_expert_model": "Qwen/Qwen3.5-0.8B",
+    "num_cognition_tokens": 64,
+    "cognition_vlm_layer": -1,
+    "cognition_layer_mix": true,
+    "cognition_layer_mix_per_token": true,
+    "cognition_layer_mix_init": "uniform",
+    "cognition_projection": "linear",
+    "cognition_context_width": 512,
+    "cognition_context_num_heads": 8,
+    "cognition_context_ffn_multiplier": 4,
+    "cognition_frozen_vlm_reference_kernels": false,
+    "cognition_action_expert_reference_kernels": false,
+    "num_trainable_vlm_layers": null,
+    "inputs_embeds_gradient_checkpointing": false,
+    "precision": "bfloat16",
+    "suffix_attention_mask": "causal",
+    "model_mode": "flow_matching",
+    "inference_mode": "flow_matching",
+    "image_shape": [
+        448,
+        448
+    ],
+    "default_temperature": 0.7,
+    "discrete_state_input": false,
+    "continuous_state_input": false,
+    "add_state_vlm": false,
+    "state_history_dropout_prob": 0.0,
+    "add_state_proj_to_action_expert": true,
+    "apply_noise_to_state_for_flow_matching": true,
+    "state_noise_amplitude": 0.1,
+    "state_dropout_prob": 0.5,
+    "attention_implementation": "flash_attention_2",
+    "mask_padded_actions": true,
+    "hold_at_subtask_boundary": false,
+    "hold_at_subtask_boundary_loss_fraction": 1.0,
+    "mask_temporally_padded_actions": true,
+    "enable_learnable_layer_combination": true,
+    "layer_combination_init": "identity",
+    "is_knowledge_insulation": false,
+    "ce_loss_weight": 0.5,
+    "freeze_vlm": false,
+    "compile_sample_actions": false,
+    "state_special_token_id": 228,
+    "clip_state": false,
+    "freeze_vision_encoder": false,
+    "use_cache": false,
+    "add_control_mode": true,
+    "add_embodiment_name": true,
+    "action_head_batch_multiplier": 1,
+    "type": "qwen3vlpolicy"
+}

pretrained_model/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5971449a5acc18eb1ee4ec8edecf88926a56966f79e9c16516c00c6728f4ab02
+size 2716858660

training_state/composed_hydra_config.json

@@ -0,0 +1,398 @@
+{
+  "policy": {
+    "policy_config": {
+      "_target_": "lerobot.common.policies.qwen3vl_policy.configuration_qwen3vl_policy.Qwen3VLPolicyConfig",
+      "max_state_dim": 50,
+      "max_action_dim": 50,
+      "base_vlm_model": "Qwen/Qwen3.5-0.8B",
+      "cognition_mode": "frozen_vlm_context_queries",
+      "action_expert_model": "Qwen/Qwen3.5-0.8B",
+      "num_cognition_tokens": 64,
+      "cognition_vlm_layer": -1,
+      "cognition_projection": "linear",
+      "num_trainable_vlm_layers": null,
+      "proj_width": 1280,
+      "precision": "bfloat16",
+      "normalization_mode": "mean_std",
+      "model_mode": "flow_matching",
+      "image_shape": [
+        448,
+        448
+      ],
+      "image_keys": [
+        "base_0_rgb",
+        "left_wrist_0_rgb",
+        "right_wrist_0_rgb"
+      ],
+      "add_control_mode": true,
+      "add_embodiment_name": true,
+      "num_steps": 10,
+      "n_action_steps": 50,
+      "mask_padded_actions": true,
+      "hold_at_subtask_boundary": false,
+      "hold_at_subtask_boundary_loss_fraction": 1.0,
+      "mask_temporally_padded_actions": true,
+      "state_noise_amplitude": 0.1,
+      "state_dropout_prob": 0.5,
+      "apply_noise_to_state_for_flow_matching": true,
+      "action_head_batch_multiplier": 1,
+      "add_state_proj_to_action_expert": true,
+      "enable_learnable_layer_combination": true,
+      "layer_combination_init": "identity",
+      "expert_block_stride": 4,
+      "suffix_attention_mask": "causal",
+      "attention_implementation": "flash_attention_2",
+      "freeze_vlm": false,
+      "freeze_vision_encoder": false,
+      "compile_sample_actions": false,
+      "is_knowledge_insulation": false,
+      "ce_loss_weight": 0.5,
+      "cognition_layer_mix": true,
+      "cognition_layer_mix_per_token": true,
+      "cognition_layer_mix_init": "uniform",
+      "cognition_context_width": 512,
+      "cognition_context_num_heads": 8,
+      "cognition_context_ffn_multiplier": 4,
+      "cognition_frozen_vlm_reference_kernels": false,
+      "cognition_action_expert_reference_kernels": false,
+      "tokenizer_max_length": 768,
+      "n_obs_steps": 1,
+      "obs_stride_seconds": 0.0,
+      "n_state_obs_steps": 1,
+      "state_obs_stride_seconds": 0.0,
+      "add_state_vlm": false,
+      "state_history_dropout_prob": 0.0,
+      "inputs_embeds_gradient_checkpointing": false
+    },
+    "policy": {
+      "_target_": "lerobot.common.policies.qwen3vl_policy.modeling_qwen3vl_policy.Qwen3VLPolicy",
+      "config": {
+        "_target_": "lerobot.common.policies.qwen3vl_policy.configuration_qwen3vl_policy.Qwen3VLPolicyConfig",
+        "max_state_dim": 50,
+        "max_action_dim": 50,
+        "base_vlm_model": "Qwen/Qwen3.5-0.8B",
+        "cognition_mode": "frozen_vlm_context_queries",
+        "action_expert_model": "Qwen/Qwen3.5-0.8B",
+        "num_cognition_tokens": 64,
+        "cognition_vlm_layer": -1,
+        "cognition_projection": "linear",
+        "num_trainable_vlm_layers": null,
+        "proj_width": 1280,
+        "precision": "bfloat16",
+        "normalization_mode": "mean_std",
+        "model_mode": "flow_matching",
+        "image_shape": [
+          448,
+          448
+        ],
+        "image_keys": [
+          "base_0_rgb",
+          "left_wrist_0_rgb",
+          "right_wrist_0_rgb"
+        ],
+        "add_control_mode": true,
+        "add_embodiment_name": true,
+        "num_steps": 10,
+        "n_action_steps": 50,
+        "mask_padded_actions": true,
+        "hold_at_subtask_boundary": false,
+        "hold_at_subtask_boundary_loss_fraction": 1.0,
+        "mask_temporally_padded_actions": true,
+        "state_noise_amplitude": 0.1,
+        "state_dropout_prob": 0.5,
+        "apply_noise_to_state_for_flow_matching": true,
+        "action_head_batch_multiplier": 1,
+        "add_state_proj_to_action_expert": true,
+        "enable_learnable_layer_combination": true,
+        "layer_combination_init": "identity",
+        "expert_block_stride": 4,
+        "suffix_attention_mask": "causal",
+        "attention_implementation": "flash_attention_2",
+        "freeze_vlm": false,
+        "freeze_vision_encoder": false,
+        "compile_sample_actions": false,
+        "is_knowledge_insulation": false,
+        "ce_loss_weight": 0.5,
+        "cognition_layer_mix": true,
+        "cognition_layer_mix_per_token": true,
+        "cognition_layer_mix_init": "uniform",
+        "cognition_context_width": 512,
+        "cognition_context_num_heads": 8,
+        "cognition_context_ffn_multiplier": 4,
+        "cognition_frozen_vlm_reference_kernels": false,
+        "cognition_action_expert_reference_kernels": false,
+        "tokenizer_max_length": 768,
+        "n_obs_steps": 1,
+        "obs_stride_seconds": 0.0,
+        "n_state_obs_steps": 1,
+        "state_obs_stride_seconds": 0.0,
+        "add_state_vlm": false,
+        "state_history_dropout_prob": 0.0,
+        "inputs_embeds_gradient_checkpointing": false
+      }
+    }
+  },
+  "image_augmentations": {
+    "base_camera": {
+      "_target_": "torchvision.transforms.v2.Compose",
+      "transforms": [
+        {
+          "_target_": "torchvision.transforms.v2.Resize",
+          "size": [
+            448,
+            448
+          ],
+          "antialias": true
+        },
+        {
+          "_target_": "torchvision.transforms.v2.RandomRotation",
+          "degrees": 1.0
+        },
+        {
+          "_target_": "torchvision.transforms.v2.RandomResizedCrop",
+          "size": [
+            448,
+            448
+          ],
+          "scale": [
+            0.97,
+            1.0
+          ],
+          "ratio": [
+            0.98,
+            1.02
+          ],
+          "antialias": true
+        },
+        {
+          "_target_": "torchvision.transforms.v2.ColorJitter",
+          "brightness": 0.08,
+          "contrast": 0.08,
+          "saturation": 0.05,
+          "hue": 0.01
+        }
+      ]
+    },
+    "wrist_camera": {
+      "_target_": "torchvision.transforms.v2.Compose",
+      "transforms": [
+        {
+          "_target_": "torchvision.transforms.v2.Resize",
+          "size": [
+            448,
+            448
+          ],
+          "antialias": true
+        },
+        {
+          "_target_": "torchvision.transforms.v2.RandomRotation",
+          "degrees": 1.0
+        },
+        {
+          "_target_": "torchvision.transforms.v2.RandomResizedCrop",
+          "size": [
+            448,
+            448
+          ],
+          "scale": [
+            0.98,
+            1.0
+          ],
+          "ratio": [
+            0.99,
+            1.01
+          ],
+          "antialias": true
+        },
+        {
+          "_target_": "torchvision.transforms.v2.ColorJitter",
+          "brightness": 0.06,
+          "contrast": 0.06,
+          "saturation": 0.04,
+          "hue": 0.01
+        }
+      ]
+    }
+  },
+  "robotics_dataset": {
+    "_target_": "lerobot.common.datasets.data_config.MixtureDataConfigFactory",
+    "name": "scripted_simxr_20260904_s0s1_step1_state_action_subtasks_shared",
+    "repo_id": "scripted_simxr_20260904",
+    "data_configs": [
+      {
+        "_target_": "lerobot.common.datasets.data_config.LeRobotHumanoidS0S1DataConfig",
+        "name": "scripted_vla_s0s1_step1_state_action_subtasks_shared",
+        "asset_id": "scripted_simxr_20260904_s0s1_step1_state_action_shared",
+        "repo_id": "scripted_vla",
+        "root_dir": "/mnt/datasets/iefremov/scripted/vla",
+        "assets": {
+          "assets_dir": "/mnt/virtual_ai0001071-01239_SR006-nfs2/nerus/assets_rl",
+          "asset_id": "scripted_simxr_20260904_s0s1_step1_state_action_shared"
+        },
+        "action_sequence_keys": [
+          "action"
+        ],
+        "action_horizon": 50,
+        "action_offset": 0,
+        "action_sample_step": 1,
+        "control_mode": "joint",
+        "smooth_actions": true,
+        "has_velocities": true,
+        "use_stationary_masks": true,
+        "map_to_unified_space": false,
+        "map_to_closed_kinematic": true,
+        "input_kinematic_space": "open",
+        "model_kinematic_space": "closed",
+        "map_old_wrists_to_new": false,
+        "map_inspire_to_leadshipe": false,
+        "return_subtasks": true,
+        "return_subtasks_mode": "strict",
+        "prompt_from_subtask": true,
+        "exclude_negative_subtasks": false,
+        "base_config": {
+          "_target_": "lerobot.common.datasets.data_config.DataConfig",
+          "local_files_only": true,
+          "prompt_from_task": false
+        }
+      },
+      {
+        "_target_": "lerobot.common.datasets.data_config.LeRobotHumanoidS0S1DataConfig",
+        "name": "simxr_merged_all_20260904_s0s1_step1_state_action_subtasks_shared",
+        "asset_id": "scripted_simxr_20260904_s0s1_step1_state_action_shared",
+        "repo_id": "simxr_merged_all_20260904",
+        "root_dir": "/mnt/virtual_ai0001071-01239_SR006-nfs2/iefremov/data/simxr_merged_all_20260904",
+        "validation_episodes": "/mnt/virtual_ai0001071-01239_SR006-nfs2/iefremov/data/simxr_merged_all_20260904/meta/validation_episodes.json",
+        "assets": {
+          "assets_dir": "/mnt/virtual_ai0001071-01239_SR006-nfs2/nerus/assets_rl",
+          "asset_id": "scripted_simxr_20260904_s0s1_step1_state_action_shared"
+        },
+        "action_sequence_keys": [
+          "action"
+        ],
+        "action_horizon": 50,
+        "action_offset": 0,
+        "action_sample_step": 1,
+        "control_mode": "joint",
+        "smooth_actions": true,
+        "has_velocities": true,
+        "use_stationary_masks": true,
+        "map_to_unified_space": false,
+        "map_to_closed_kinematic": true,
+        "input_kinematic_space": "open",
+        "model_kinematic_space": "closed",
+        "map_old_wrists_to_new": false,
+        "map_inspire_to_leadshipe": false,
+        "return_subtasks": true,
+        "return_subtasks_mode": "strict",
+        "prompt_from_subtask": true,
+        "exclude_negative_subtasks": false,
+        "base_config": {
+          "_target_": "lerobot.common.datasets.data_config.DataConfig",
+          "local_files_only": true,
+          "prompt_from_task": false
+        }
+      }
+    ],
+    "weights": [
+      10745464,
+      4423926
+    ]
+  },
+  "optimizer": {
+    "_target_": "lerobot.common.optim.optimizers.AdamWConfig",
+    "lr": 5e-05,
+    "betas": [
+      0.9,
+      0.95
+    ],
+    "eps": 1e-08,
+    "weight_decay": 1e-08,
+    "grad_clip_norm": 1.0
+  },
+  "scheduler": {
+    "_target_": "lerobot.common.optim.schedulers.WarmupStableDecaySchedulerConfig",
+    "num_warmup_steps": 992,
+    "stable_phase_steps": 11851,
+    "num_decay_steps": 1971,
+    "peak_lr": 5e-05,
+    "decay_lr": 0.0
+  },
+  "training": {
+    "batch_size": 8,
+    "steps": 14814,
+    "num_warmup_steps": 992,
+    "num_decay_steps": null,
+    "log_freq": 10,
+    "save_freq": 14814,
+    "num_workers": 16,
+    "num_workers_validation": 8,
+    "shuffle": true,
+    "seed": 42,
+    "resume": false,
+    "prefetch_factor": 4,
+    "output_dir": "/mnt/virtual_ai0001071-01239_SR006-nfs2/nerus/training_runs/postnikov_qwen35_08b_scripted_simxr_20260904_step1_verylight_state_action_subtasks_sharedstats_nogc_1epoch_fastkernels_32gpu_2026-09-05_00-02",
+    "grad_accum_steps": 4,
+    "find_unused_parameters": true,
+    "use_fsdp": false,
+    "use_fsdp2": true,
+    "fsdp_sharding": "full",
+    "gradient_checkpointing_enabled": false,
+    "activation_offload_enabled": false,
+    "activation_gpu_limit_gb": 0.0,
+    "compile": false,
+    "enable_validation": true,
+    "use_validation_list": true,
+    "val_split": 0.03,
+    "val_freq": 14814,
+    "max_val_steps": 100,
+    "recompute_norm_stats": false,
+    "visualization": {
+      "plots_per_validation": 5,
+      "subplot_height": 3,
+      "subplot_width": 8
+    },
+    "fsdp_keep_trainable_vlm_unsharded": false,
+    "validate_at_start": false,
+    "validation_batch_size": 1,
+    "kernel_prewarm": true,
+    "save_checkpoint": true
+  },
+  "logging": {
+    "clearml": {
+      "task": {
+        "project_name": "Manipulation/2D Manipulation/lerobot",
+        "task_name": "postnikov_qwen35_08b_scripted_simxr_20260904_step1_verylight_state_action_subtasks_sharedstats_nogc_1epoch_fastkernels_32gpu_2026-09-05_00-02",
+        "task_type": "training"
+      },
+      "logging": {
+        "level": "INFO"
+      },
+      "upload_method": "rclone",
+      "s3_output_uri": "s3://obs.ru-moscow-1.hc.sbercloud.ru:443/amr-usr-manip/checkpoints",
+      "keep_latest_local": 1
+    }
+  },
+  "research_ledger_mode": "ineligible",
+  "run_name": "postnikov_qwen35_08b_scripted_simxr_20260904_step1_verylight_state_action_subtasks_sharedstats_nogc_1epoch_fastkernels_32gpu_2026-09-05_00-02",
+  "tags": [
+    "postnikov-aij",
+    "qwen3.5",
+    "qwen35-0.8b",
+    "scripted-simxr-mix",
+    "shared-norm-stats",
+    "state-action-target",
+    "next-state-target",
+    "subtask-prompt",
+    "no-gradient-checkpointing",
+    "s0s1",
+    "step1",
+    "very-light-aug",
+    "no-history",
+    "from-scratch",
+    "one-epoch",
+    "fast-kernels"
+  ],
+  "assets_dir": "/mnt/virtual_ai0001071-01239_SR006-nfs2/nerus/assets_rl",
+  "pretrained_path": null,
+  "model_id": null
+}

training_state/rng_state.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6379712072629d3e40d7bd0eaf2ed08b05b5e3fb132fd1d4560fc19f38f21ffe
+size 15708

training_state/scheduler_state.json

@@ -0,0 +1,14 @@
+{
+    "base_lrs": [
+        5e-05
+    ],
+    "last_epoch": 474048,
+    "_step_count": 474049,
+    "_get_lr_called_within_step": false,
+    "_last_lr": [
+        0.0
+    ],
+    "lr_lambdas": [
+        null
+    ]
+}

training_state/training_step.json

@@ -0,0 +1,3 @@
+{
+    "step": 14814
+}