Add files using upload-large-folder tool

1f15334ab7e6820c9fda17c961659882ef9853cc80f7356b9a9b22f286fd7389/crowd-code-ef5ea013-ac2b-459c-8783-a7b025d58a391754900011518-2025_08_11-10.13.34.249/source.csv

@@ -0,0 +1,4 @@
+3503,453188525,"Untitled-1",0,0,"",plaintext,tab
+3504,453189769,"train_dynamics.py",0,0,"",python,tab
+3505,453190506,"Untitled-1",0,0,"",plaintext,tab
+3506,453190928,"train_dynamics.py",0,0,"",python,tab

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-066da5ef-4ba0-4726-8184-5dc4ef5e75811751465994612-2025_07_02-16.20.42.177/source.csv

@@ -0,0 +1,358 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,241,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"4:20:42 PM [info] Activating crowd-code\n4:20:42 PM [info] Recording started\n4:20:42 PM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,825,"extension-output-pdoom-org.crowd-code-#1-crowd-code",150,0,"4:20:42 PM [info] Git repository found\n4:20:42 PM [info] Git provider initialized successfully\n",Log,content
+4,1121,"extension-output-pdoom-org.crowd-code-#1-crowd-code",245,0,"4:20:43 PM [info] Initial git state: [object Object]\n",Log,content
+5,3223,"TERMINAL",0,0,"idling",,terminal_command
+6,3312,"TERMINAL",0,0,"]633;E;2025-07-02 16:20:45 idling;26cd839c-476e-4913-967a-1422bf7b3816]633;C[?1049h(B[?7hEvery 1.0s: sinfo_t_idlehkn1993.localdomain: Wed Jul  2 16:20:45 2025Partition dev_cpuonly:\t 9 nodes idle\rPartition cpuonly:     69 nodes idle\rPartition dev_accelerated:\t 0 nodes idle\rPartition accelerated:\t 0 nodes idle\rPartition dev_accelerated-h100    :\t 0 nodes idle\rPartition accelerated-h100:\t 0 nodes idle\rPartition large:\t 6 nodes idle",,terminal_output
+7,3863,"TERMINAL",0,0,"/bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt",,terminal_command
+8,3899,"TERMINAL",0,0,"]633;E;2025-07-02 16:20:45 /bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt;14446405-24c7-42a5-a6ff-f36b2c7c204e]633;C",,terminal_output
+9,3923,"TERMINAL",0,0,"]0;tum_cte0515@hkn1993:/hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash]633;D;0",,terminal_output
+10,4353,"TERMINAL",0,0,"6101",,terminal_output
+11,5398,"TERMINAL",0,0,"7\t ",,terminal_output
+12,5843,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;0",,terminal_output
+13,11230,"TERMINAL",0,0,"source .venv/bin/activate",,terminal_command
+14,15010,"TERMINAL",0,0,"queue",,terminal_command
+15,15052,"TERMINAL",0,0,"]633;E;2025-07-02 16:20:57 queue;26cd839c-476e-4913-967a-1422bf7b3816]633;C",,terminal_output
+16,15145,"TERMINAL",0,0,"[?1049h(B[?7hEvery 1.0s: squeue --mehkn1993.localdomain: Wed Jul  2 16:20:57 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3311671 accelerat train_to tum_cte0  R    1:56:41\t  1 hkn0717",,terminal_output
+17,16168,"TERMINAL",0,0,"82",,terminal_output
+18,17176,"TERMINAL",0,0,"93",,terminal_output
+19,18229,"TERMINAL",0,0,"1:004",,terminal_output
+20,19280,"TERMINAL",0,0,"15",,terminal_output
+21,20325,"TERMINAL",0,0,"26",,terminal_output
+22,21375,"TERMINAL",0,0,"37",,terminal_output
+23,22381,"TERMINAL",0,0,"48",,terminal_output
+24,23415,"TERMINAL",0,0,"59",,terminal_output
+25,24458,"TERMINAL",0,0,"650",,terminal_output
+26,25500,"TERMINAL",0,0,"71",,terminal_output
+27,26554,"TERMINAL",0,0,"82",,terminal_output
+28,27638,"TERMINAL",0,0,"93",,terminal_output
+29,28771,"TERMINAL",0,0,"104",,terminal_output
+30,29687,"TERMINAL",0,0,"15",,terminal_output
+31,30801,"TERMINAL",0,0,"26",,terminal_output
+32,31799,"TERMINAL",0,0,"37",,terminal_output
+33,32842,"TERMINAL",0,0,"48",,terminal_output
+34,33915,"TERMINAL",0,0,"59",,terminal_output
+35,34938,"TERMINAL",0,0,"67:01",,terminal_output
+36,35991,"TERMINAL",0,0,"82",,terminal_output
+37,37044,"TERMINAL",0,0,"93",,terminal_output
+38,38084,"TERMINAL",0,0,"204",,terminal_output
+39,39136,"TERMINAL",0,0,"15",,terminal_output
+40,40175,"TERMINAL",0,0,"26",,terminal_output
+41,41228,"TERMINAL",0,0,"37",,terminal_output
+42,42316,"TERMINAL",0,0,"48",,terminal_output
+43,43362,"TERMINAL",0,0,"59",,terminal_output
+44,44399,"TERMINAL",0,0,"610",,terminal_output
+45,45439,"TERMINAL",0,0,"71",,terminal_output
+46,46484,"TERMINAL",0,0,"82",,terminal_output
+47,47524,"TERMINAL",0,0,"93",,terminal_output
+48,48558,"TERMINAL",0,0,"304",,terminal_output
+49,49609,"TERMINAL",0,0,"15",,terminal_output
+50,50628,"TERMINAL",0,0,"26",,terminal_output
+51,51687,"TERMINAL",0,0,"37",,terminal_output
+52,52733,"TERMINAL",0,0,"48",,terminal_output
+53,53771,"TERMINAL",0,0,"59",,terminal_output
+54,54813,"TERMINAL",0,0,"620",,terminal_output
+55,55873,"TERMINAL",0,0,"71",,terminal_output
+56,56914,"TERMINAL",0,0,"83",,terminal_output
+57,57963,"TERMINAL",0,0,"404",,terminal_output
+58,59001,"TERMINAL",0,0,"15",,terminal_output
+59,60059,"TERMINAL",0,0,"26",,terminal_output
+60,61088,"TERMINAL",0,0,"37",,terminal_output
+61,62175,"TERMINAL",0,0,"48",,terminal_output
+62,63190,"TERMINAL",0,0,"59",,terminal_output
+63,64223,"TERMINAL",0,0,"630",,terminal_output
+64,65267,"TERMINAL",0,0,"71",,terminal_output
+65,66304,"TERMINAL",0,0,"82",,terminal_output
+66,67359,"TERMINAL",0,0,"93",,terminal_output
+67,68400,"TERMINAL",0,0,"504",,terminal_output
+68,69439,"TERMINAL",0,0,"15",,terminal_output
+69,70458,"TERMINAL",0,0,"26",,terminal_output
+70,71507,"TERMINAL",0,0,"37",,terminal_output
+71,72552,"TERMINAL",0,0,"48",,terminal_output
+72,73601,"TERMINAL",0,0,"59",,terminal_output
+73,74642,"TERMINAL",0,0,"640",,terminal_output
+74,75776,"TERMINAL",0,0,"71",,terminal_output
+75,76720,"TERMINAL",0,0,"82",,terminal_output
+76,77825,"TERMINAL",0,0,"93",,terminal_output
+77,78850,"TERMINAL",0,0,"2:004",,terminal_output
+78,79867,"TERMINAL",0,0,"15",,terminal_output
+79,80916,"TERMINAL",0,0,"27",,terminal_output
+80,82024,"TERMINAL",0,0,"48",,terminal_output
+81,83048,"TERMINAL",0,0,"59",,terminal_output
+82,84062,"TERMINAL",0,0,"650",,terminal_output
+83,85199,"TERMINAL",0,0,"71",,terminal_output
+84,86186,"TERMINAL",0,0,"82",,terminal_output
+85,87220,"TERMINAL",0,0,"93",,terminal_output
+86,88273,"TERMINAL",0,0,"104",,terminal_output
+87,89320,"TERMINAL",0,0,"15",,terminal_output
+88,90408,"TERMINAL",0,0,"26",,terminal_output
+89,91444,"TERMINAL",0,0,"37",,terminal_output
+90,92572,"TERMINAL",0,0,"48",,terminal_output
+91,93550,"TERMINAL",0,0,"59",,terminal_output
+92,94557,"TERMINAL",0,0,"68:00",,terminal_output
+93,95645,"TERMINAL",0,0,"71",,terminal_output
+94,96668,"TERMINAL",0,0,"82",,terminal_output
+95,97691,"TERMINAL",0,0,"93",,terminal_output
+96,98818,"TERMINAL",0,0,"204",,terminal_output
+97,99755,"TERMINAL",0,0,"15",,terminal_output
+98,100869,"TERMINAL",0,0,"26",,terminal_output
+99,101827,"TERMINAL",0,0,"37",,terminal_output
+100,102871,"TERMINAL",0,0,"48",,terminal_output
+101,103910,"TERMINAL",0,0,"510",,terminal_output
+102,104953,"TERMINAL",0,0,"71",,terminal_output
+103,106086,"TERMINAL",0,0,"82",,terminal_output
+104,107056,"TERMINAL",0,0,"93",,terminal_output
+105,108157,"TERMINAL",0,0,"304",,terminal_output
+106,109213,"TERMINAL",0,0,"15",,terminal_output
+107,110257,"TERMINAL",0,0,"26",,terminal_output
+108,111288,"TERMINAL",0,0,"37",,terminal_output
+109,112341,"TERMINAL",0,0,"48",,terminal_output
+110,113392,"TERMINAL",0,0,"59",,terminal_output
+111,114483,"TERMINAL",0,0,"620",,terminal_output
+112,115480,"TERMINAL",0,0,"71",,terminal_output
+113,116502,"TERMINAL",0,0,"82",,terminal_output
+114,117589,"TERMINAL",0,0,"93",,terminal_output
+115,118589,"TERMINAL",0,0,"404",,terminal_output
+116,119706,"TERMINAL",0,0,"15",,terminal_output
+117,120737,"TERMINAL",0,0,"26",,terminal_output
+118,121857,"TERMINAL",0,0,"37",,terminal_output
+119,122882,"TERMINAL",0,0,"48",,terminal_output
+120,123879,"TERMINAL",0,0,"59",,terminal_output
+121,125032,"TERMINAL",0,0,"631",,terminal_output
+122,126058,"TERMINAL",0,0,"82",,terminal_output
+123,127078,"TERMINAL",0,0,"93",,terminal_output
+124,128105,"TERMINAL",0,0,"504",,terminal_output
+125,129144,"TERMINAL",0,0,"15",,terminal_output
+126,130159,"TERMINAL",0,0,"26",,terminal_output
+127,131201,"TERMINAL",0,0,"37",,terminal_output
+128,132261,"TERMINAL",0,0,"48",,terminal_output
+129,133300,"TERMINAL",0,0,"59",,terminal_output
+130,134348,"TERMINAL",0,0,"640",,terminal_output
+131,135480,"TERMINAL",0,0,"71",,terminal_output
+132,136516,"TERMINAL",0,0,"82",,terminal_output
+133,137528,"TERMINAL",0,0,"93",,terminal_output
+134,138545,"TERMINAL",0,0,"3:004",,terminal_output
+135,139569,"TERMINAL",0,0,"15",,terminal_output
+136,140617,"TERMINAL",0,0,"26",,terminal_output
+137,141723,"TERMINAL",0,0,"37",,terminal_output
+138,142750,"TERMINAL",0,0,"48",,terminal_output
+139,143875,"TERMINAL",0,0,"59",,terminal_output
+140,144920,"TERMINAL",0,0,"650",,terminal_output
+141,145925,"TERMINAL",0,0,"71",,terminal_output
+142,146945,"TERMINAL",0,0,"82",,terminal_output
+143,147970,"TERMINAL",0,0,"94",,terminal_output
+144,149098,"TERMINAL",0,0,"115",,terminal_output
+145,150048,"TERMINAL",0,0,"26",,terminal_output
+146,151144,"TERMINAL",0,0,"37",,terminal_output
+147,152130,"TERMINAL",0,0,"48",,terminal_output
+148,153197,"TERMINAL",0,0,"59",,terminal_output
+149,154233,"TERMINAL",0,0,"69:00",,terminal_output
+150,155287,"TERMINAL",0,0,"71",,terminal_output
+151,156368,"TERMINAL",0,0,"82",,terminal_output
+152,157391,"TERMINAL",0,0,"93",,terminal_output
+153,158444,"TERMINAL",0,0,"204",,terminal_output
+154,159544,"TERMINAL",0,0,"15",,terminal_output
+155,160568,"TERMINAL",0,0,"26",,terminal_output
+156,161586,"TERMINAL",0,0,"37",,terminal_output
+157,162636,"TERMINAL",0,0,"48",,terminal_output
+158,163687,"TERMINAL",0,0,"59",,terminal_output
+159,164735,"TERMINAL",0,0,"610",,terminal_output
+160,165787,"TERMINAL",0,0,"71",,terminal_output
+161,166805,"TERMINAL",0,0,"82",,terminal_output
+162,167938,"TERMINAL",0,0,"93",,terminal_output
+163,168870,"TERMINAL",0,0,"304",,terminal_output
+164,169988,"TERMINAL",0,0,"16",,terminal_output
+165,171008,"TERMINAL",0,0,"37",,terminal_output
+166,172736,"TERMINAL",0,0,"48",,terminal_output
+167,173818,"TERMINAL",0,0,"59",,terminal_output
+168,174823,"TERMINAL",0,0,"620",,terminal_output
+169,175879,"TERMINAL",0,0,"71",,terminal_output
+170,176929,"TERMINAL",0,0,"83",,terminal_output
+171,177980,"TERMINAL",0,0,"404",,terminal_output
+172,179020,"TERMINAL",0,0,"15",,terminal_output
+173,180047,"TERMINAL",0,0,"26",,terminal_output
+174,181098,"TERMINAL",0,0,"37",,terminal_output
+175,182144,"TERMINAL",0,0,"48",,terminal_output
+176,183199,"TERMINAL",0,0,"59",,terminal_output
+177,184252,"TERMINAL",0,0,"630",,terminal_output
+178,185293,"TERMINAL",0,0,"71",,terminal_output
+179,186346,"TERMINAL",0,0,"82",,terminal_output
+180,187386,"TERMINAL",0,0,"93",,terminal_output
+181,188438,"TERMINAL",0,0,"504",,terminal_output
+182,189489,"TERMINAL",0,0,"15",,terminal_output
+183,190534,"TERMINAL",0,0,"26",,terminal_output
+184,191589,"TERMINAL",0,0,"37",,terminal_output
+185,192642,"TERMINAL",0,0,"48",,terminal_output
+186,193686,"TERMINAL",0,0,"59",,terminal_output
+187,194730,"TERMINAL",0,0,"640",,terminal_output
+188,195783,"TERMINAL",0,0,"71",,terminal_output
+189,196832,"TERMINAL",0,0,"82",,terminal_output
+190,197945,"TERMINAL",0,0,"93",,terminal_output
+191,198909,"TERMINAL",0,0,"4:005",,terminal_output
+192,199955,"TERMINAL",0,0,"26",,terminal_output
+193,201013,"TERMINAL",0,0,"37",,terminal_output
+194,202045,"TERMINAL",0,0,"48",,terminal_output
+195,203085,"TERMINAL",0,0,"59",,terminal_output
+196,204132,"TERMINAL",0,0,"650",,terminal_output
+197,205180,"TERMINAL",0,0,"71",,terminal_output
+198,206228,"TERMINAL",0,0,"82",,terminal_output
+199,207263,"TERMINAL",0,0,"93",,terminal_output
+200,208305,"TERMINAL",0,0,"104",,terminal_output
+201,209343,"TERMINAL",0,0,"15",,terminal_output
+202,210433,"TERMINAL",0,0,"26",,terminal_output
+203,211401,"TERMINAL",0,0,"37",,terminal_output
+204,212440,"TERMINAL",0,0,"48",,terminal_output
+205,213533,"TERMINAL",0,0,"59",,terminal_output
+206,214534,"TERMINAL",0,0,"62:00:00",,terminal_output
+207,215556,"TERMINAL",0,0,"71",,terminal_output
+208,216683,"TERMINAL",0,0,"82",,terminal_output
+209,217704,"TERMINAL",0,0,"93",,terminal_output
+210,218676,"TERMINAL",0,0,"204",,terminal_output
+211,219713,"TERMINAL",0,0,"15",,terminal_output
+212,220882,"TERMINAL",0,0,"26",,terminal_output
+213,221902,"TERMINAL",0,0,"37",,terminal_output
+214,222853,"TERMINAL",0,0,"48",,terminal_output
+215,223913,"TERMINAL",0,0,"510",,terminal_output
+216,224993,"TERMINAL",0,0,"71",,terminal_output
+217,225981,"TERMINAL",0,0,"82",,terminal_output
+218,227095,"TERMINAL",0,0,"93",,terminal_output
+219,228148,"TERMINAL",0,0,"304",,terminal_output
+220,229183,"TERMINAL",0,0,"15",,terminal_output
+221,230174,"TERMINAL",0,0,"26",,terminal_output
+222,231210,"TERMINAL",0,0,"37",,terminal_output
+223,232274,"TERMINAL",0,0,"48",,terminal_output
+224,233322,"TERMINAL",0,0,"59",,terminal_output
+225,234362,"TERMINAL",0,0,"620",,terminal_output
+226,235404,"TERMINAL",0,0,"71",,terminal_output
+227,236454,"TERMINAL",0,0,"82",,terminal_output
+228,237503,"TERMINAL",0,0,"93",,terminal_output
+229,238542,"TERMINAL",0,0,"404",,terminal_output
+230,239598,"TERMINAL",0,0,"15",,terminal_output
+231,240645,"TERMINAL",0,0,"26",,terminal_output
+232,241692,"TERMINAL",0,0,"37",,terminal_output
+233,242731,"TERMINAL",0,0,"48",,terminal_output
+234,243808,"TERMINAL",0,0,"59",,terminal_output
+235,244792,"TERMINAL",0,0,"630",,terminal_output
+236,245835,"TERMINAL",0,0,"71",,terminal_output
+237,246903,"TERMINAL",0,0,"83",,terminal_output
+238,248016,"TERMINAL",0,0,"504",,terminal_output
+239,249001,"TERMINAL",0,0,"15",,terminal_output
+240,250082,"TERMINAL",0,0,"26",,terminal_output
+241,251096,"TERMINAL",0,0,"37",,terminal_output
+242,252230,"TERMINAL",0,0,"48",,terminal_output
+243,253213,"TERMINAL",0,0,"59",,terminal_output
+244,254265,"TERMINAL",0,0,"640",,terminal_output
+245,255317,"TERMINAL",0,0,"71",,terminal_output
+246,256445,"TERMINAL",0,0,"82",,terminal_output
+247,257415,"TERMINAL",0,0,"93",,terminal_output
+248,258467,"TERMINAL",0,0,"5:004",,terminal_output
+249,259491,"TERMINAL",0,0,"15",,terminal_output
+250,260541,"TERMINAL",0,0,"26",,terminal_output
+251,261597,"TERMINAL",0,0,"37",,terminal_output
+252,262644,"TERMINAL",0,0,"48",,terminal_output
+253,263683,"TERMINAL",0,0,"59",,terminal_output
+254,265423,"TERMINAL",0,0,"651",,terminal_output
+255,266549,"TERMINAL",0,0,"82",,terminal_output
+256,267573,"TERMINAL",0,0,"93",,terminal_output
+257,268536,"TERMINAL",0,0,"104",,terminal_output
+258,269659,"TERMINAL",0,0,"15",,terminal_output
+259,270603,"TERMINAL",0,0,"26",,terminal_output
+260,271673,"TERMINAL",0,0,"37",,terminal_output
+261,272686,"TERMINAL",0,0,"48",,terminal_output
+262,273739,"TERMINAL",0,0,"59",,terminal_output
+263,274843,"TERMINAL",0,0,"61:00",,terminal_output
+264,275869,"TERMINAL",0,0,"71",,terminal_output
+265,276857,"TERMINAL",0,0,"82",,terminal_output
+266,277915,"TERMINAL",0,0,"93",,terminal_output
+267,278940,"TERMINAL",0,0,"205",,terminal_output
+268,280068,"TERMINAL",0,0,"26",,terminal_output
+269,281093,"TERMINAL",0,0,"37",,terminal_output
+270,282114,"TERMINAL",0,0,"48",,terminal_output
+271,283129,"TERMINAL",0,0,"59",,terminal_output
+272,284264,"TERMINAL",0,0,"610",,terminal_output
+273,285295,"TERMINAL",0,0,"71",,terminal_output
+274,286261,"TERMINAL",0,0,"82",,terminal_output
+275,287311,"TERMINAL",0,0,"93",,terminal_output
+276,288349,"TERMINAL",0,0,"304",,terminal_output
+277,289487,"TERMINAL",0,0,"15",,terminal_output
+278,290438,"TERMINAL",0,0,"26",,terminal_output
+279,291535,"TERMINAL",0,0,"37",,terminal_output
+280,292563,"TERMINAL",0,0,"48",,terminal_output
+281,293569,"TERMINAL",0,0,"59",,terminal_output
+282,294644,"TERMINAL",0,0,"620",,terminal_output
+283,295655,"TERMINAL",0,0,"71",,terminal_output
+284,296695,"TERMINAL",0,0,"82",,terminal_output
+285,297741,"TERMINAL",0,0,"93",,terminal_output
+286,298794,"TERMINAL",0,0,"404",,terminal_output
+287,299838,"TERMINAL",0,0,"15",,terminal_output
+288,300882,"TERMINAL",0,0,"26",,terminal_output
+289,301983,"TERMINAL",0,0,"38",,terminal_output
+290,302951,"TERMINAL",0,0,"59",,terminal_output
+291,303999,"TERMINAL",0,0,"630",,terminal_output
+292,305051,"TERMINAL",0,0,"71",,terminal_output
+293,306178,"TERMINAL",0,0,"82",,terminal_output
+294,307202,"TERMINAL",0,0,"93",,terminal_output
+295,308176,"TERMINAL",0,0,"504",,terminal_output
+296,309221,"TERMINAL",0,0,"15",,terminal_output
+297,310276,"TERMINAL",0,0,"26",,terminal_output
+298,311314,"TERMINAL",0,0,"37",,terminal_output
+299,312428,"TERMINAL",0,0,"48",,terminal_output
+300,313448,"TERMINAL",0,0,"59",,terminal_output
+301,314473,"TERMINAL",0,0,"640",,terminal_output
+302,315495,"TERMINAL",0,0,"71",,terminal_output
+303,316521,"TERMINAL",0,0,"82",,terminal_output
+304,317570,"TERMINAL",0,0,"93",,terminal_output
+305,318624,"TERMINAL",0,0,"6:004",,terminal_output
+306,319697,"TERMINAL",0,0,"15",,terminal_output
+307,320678,"TERMINAL",0,0,"26",,terminal_output
+308,321736,"TERMINAL",0,0,"37",,terminal_output
+309,322777,"TERMINAL",0,0,"48",,terminal_output
+310,323894,"TERMINAL",0,0,"59",,terminal_output
+311,324922,"TERMINAL",0,0,"650",,terminal_output
+312,325942,"TERMINAL",0,0,"72",,terminal_output
+313,326965,"TERMINAL",0,0,"93",,terminal_output
+314,327991,"TERMINAL",0,0,"104",,terminal_output
+315,329116,"TERMINAL",0,0,"15",,terminal_output
+316,330243,"TERMINAL",0,0,"26",,terminal_output
+317,331133,"TERMINAL",0,0,"37",,terminal_output
+318,332188,"TERMINAL",0,0,"48",,terminal_output
+319,333240,"TERMINAL",0,0,"59",,terminal_output
+320,334286,"TERMINAL",0,0,"62:00",,terminal_output
+321,334771,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;0",,terminal_output
+322,4560228,"TERMINAL",0,0,"queue",,terminal_command
+323,4560272,"TERMINAL",0,0,"]633;E;2025-07-02 17:36:42 queue;26cd839c-476e-4913-967a-1422bf7b3816]633;C",,terminal_output
+324,4560343,"TERMINAL",0,0,"[?1049h(B[?7hEvery 1.0s: squeue --mehkn1993.localdomain: Wed Jul  2 17:36:42 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3311671 accelerat train_to tum_cte0  R    3:12:26\t  1 hkn0717",,terminal_output
+325,4561383,"TERMINAL",0,0,"37",,terminal_output
+326,4561489,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;0",,terminal_output
+327,4562886,"TERMINAL",0,0,"idling",,terminal_command
+328,4562983,"TERMINAL",0,0,"]633;E;2025-07-02 17:36:44 idling;26cd839c-476e-4913-967a-1422bf7b3816]633;C[?1049h(B[?7hEvery 1.0s: sinfo_t_idlehkn1993.localdomain: Wed Jul  2 17:36:45 2025Partition dev_cpuonly:     12 nodes idle\rPartition cpuonly:\t 4 nodes idle\rPartition dev_accelerated:\t 0 nodes idle\rPartition accelerated:\t 0 nodes idle\rPartition dev_accelerated-h100    :\t 0 nodes idle\rPartition accelerated-h100:\t 0 nodes idle\rPartition large:\t 7 nodes idle",,terminal_output
+329,4564005,"TERMINAL",0,0,"6\t ",,terminal_output
+330,4565051,"TERMINAL",0,0,"7\t ",,terminal_output
+331,4566099,"TERMINAL",0,0,"8\t ",,terminal_output
+332,4566873,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;0",,terminal_output
+333,4569190,"TERMINAL",0,0,"",,terminal_focus
+334,4575053,"scripts_horeka/overfit_batch_tiny/sample.sh",0,0,"#!/usr/bin/env bash\n\n# Unload modules that may interfere\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\n\n# Activate virtual environment\nsource .venv/bin/activate\n\n# Set workspace and checkpoint directory (update slurm_job_id as needed)\nws_dir='/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared'\n# Replace the following with the actual job id/checkpoint you want to sample from\nslurm_job_id=3301029\n\n# job_name=train_dynamics_minecraft_overfit_sample_tiny\nCHECKPOINT_DIR=$ws_dir/checkpoints/${slurm_job_id}\n\n# Example: If you want to use a specific checkpoint, set it here\n# CHECKPOINT_PATH=$ws_dir/checkpoints/3299272/dynamics-tiny-overfit-big-lr-3299272_50000/\n# Or use the latest in the directory\n# CHECKPOINT_PATH=$(ls -d $CHECKPOINT_DIR/*/ | sort | tail -n 1)\nCHECKPOINT_PATH=$CHECKPOINT_DIR/genie_1751067601_200000/\n# CHECKPOINT_PATH=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/0000/genie_1751301068_2000/\n# CHECKPOINT_PATH=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data/../checkpoints/3307618/genie_1751322003_15500/\n# CHECKPOINT_PATH=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/3307619/genie_1751322003_200000/\nCHECKPOINT_PATH=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/3309699/genie_1751384516_200000/\n\n\necho ""Sampling from checkpoint: $CHECKPOINT_PATH""\n\npython sample.py \\n    --checkpoint ""$CHECKPOINT_PATH"" \\n    --tokenizer_dim=384 \\n    --latent_patch_dim=32 \\n    --num_patch_latents=1024 \\n    --patch_size=4 \\n    --tokenizer_num_blocks=8 \\n    --tokenizer_num_heads=8 \\n    --lam_dim=384 \\n    --latent_action_dim=32 \\n    --lam_patch_size=16 \\n    --lam_num_blocks=8 \\n    --lam_num_heads=8 \\n    --dyna_dim=128 \\n    --dyna_num_blocks=2 \\n    --dyna_num_heads=4 \\n    --maskgit_steps=1000 \\n    --num_latent_actions=6 \\n    --seq_len=16 \\n    --start_frame=0\n\n# python sample.py \\n    # --checkpoint ""$CHECKPOINT_PATH"" \\n    # --data_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data/coinrun_episodes\n",shellscript,tab
+335,4582430,"TERMINAL",0,0,"bash",,terminal_focus
+336,4584712,"TERMINAL",0,0,"bash",,terminal_focus
+337,4584714,"scripts_horeka/overfit_batch_tiny/sample.sh",0,0,"",shellscript,tab
+338,4585497,"scripts_horeka/modelsize_scaling/dynamics/model_sizes.md",0,0,"# Genie 1 - Model Sizes and their configs\n\n## Tokenizer model: sizes\n\ndefault: \n| Model | model_dim | num_blocks | num_heads | latent_dim | num_latents | Est. Params |\n|-------|-----------|------------|-----------|------------|-------------|-------------|\n| default | 512 | 8 | 8 | 32 | 1024 | ~38M |\n\n### scaling up \n#### (not tested yet - TODO @mihir)\n\n| Model | model_dim | num_blocks | num_heads | latent_dim | num_latents | Est. Params |\n|-------|-----------|------------|-----------|------------|-------------|-------------|\n| L1 | 768 | 12 | 12 | 64 | 2048 | ~80M |\n| L2 | 1024 | 12 | 16 | 128 | 2048 | ~140M |\n| L3 | 1152 | 16 | 16 | 128 | 4096 | ~200M |\n| L4 | 896 | 16 | 14 | 96 | 4096 | ~120M |\n| L5 | 1536 | 12 | 24 | 256 | 2048 | ~190M |\n\n\n### tiny models\n| Model | model_dim | num_blocks | num_heads | latent_dim | num_latents | Est. Params |\n|-------|-----------|------------|-----------|------------|-------------|-------------|\n| S1 | 128 | 2 | 2 | 8 | 128 | ~0.6M |\n| S2 | 192 | 2 | 3 | 16 | 128 | ~1.3M |\n| S3 | 256 | 3 | 4 | 16 | 256 | ~3.6M |\n| S4 | 320 | 4 | 5 | 24 | 256 | ~7.4M |\n| S5 | 384 | 4 | 6 | 32 | 512 | ~10M |\n\n\n## Latent Action model: sizes\ndefault: \n| Model | model_dim | num_blocks | num_heads | latent_dim | num_latents | Est. Params |\n|-------|-----------|------------|-----------|------------|-------------|-------------|\n| default | 512 | 8 | 8 | 32 | 6 | ~39M |\n\n### scaling up \n#### (not tested yet - TODO @mihir)\n\n| Name | model_dim | num_blocks | num_heads | latent_dim | num_latents | Est. Params |\n|--------------|-----------|------------|-----------|------------|-------------|-------------|\n| XL | 1024 | 12 | 16 | 64 | 12 | ~200M |\n| L | 896 | 12 | 14 | 48 | 8 | ~150M |\n| M+ | 768 | 10 | 12 | 48 | 8 | ~100M |\n| M | 640 | 10 | 10 | 32 | 8 | ~70M |\n| Base+ | 512 | 12 | 8 | 32 | 8 | ~55M |\n\n\n### tiny models\n| Name | model_dim | num_blocks | num_heads | latent_dim | num_latents | Est. Params |\n|--------------|-----------|------------|-----------|------------|-------------|-------------|\n| XS | 128 | 2 | 2 | 8 | 4 | ~0.9M |\n| S | 160 | 2 | 2 | 8 | 4 | ~1.3M |\n| S+ | 192 | 3 | 3 | 8 | 4 | ~2.4M |\n| M- | 256 | 4 | 4 | 16 | 6 | ~5.4M |\n| M | 320 | 6 | 4 | 16 | 6 | ~12M |\n\n\n## Dynamics model: sizes \n\n| Config | dyna_dim | dyna_num_blocks | dyna_num_heads | Approx. Params |\n|--------|----------|-----------------|---------------|----------------|\n| 1 | 512 | 12 | 8   | ~36M  |\n| 2 | 768 | 16 | 12  | ~110M |\n| 3 | 1024 | 16 | 16 | ~180M |\n| 4 | 1024 | 24 | 16 | ~270M |\n| 5 | 1536 | 24 | 24 | ~500M |\n\n\n### tiny models\n| Config | dyna_dim | dyna_num_blocks | dyna_num_heads | Approx. Params |\n|--------|----------|-----------------|---------------|----------------|\n| A | 128 | 2 | 4 | ~1.5M |\n| B | 256 | 2 | 4 | ~3.5M |\n| C | 256 | 4 | 4 | ~6M   |\n| D | 384 | 4 | 6 | ~12M  |\n| E | 512 | 4 | 8 | ~18M  |",markdown,tab
+339,4873342,"TERMINAL",0,0,"bash",,terminal_focus
+340,5485002,"TERMINAL",0,0,"salloc --time=00:30:00 --partition=dev_accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5 --mem=50G",,terminal_command
+341,5485037,"TERMINAL",0,0,"]633;E;2025-07-02 17:52:07 salloc --time=00:30:00 --partition=dev_accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5 --mem=50G;26cd839c-476e-4913-967a-1422bf7b3816]633;Csalloc: Pending job allocation 3312853\r\nsalloc: job 3312853 queued and waiting for resources\r\n",,terminal_output
+342,5486062,"TERMINAL",0,0,"^Csalloc: Job allocation 3312853 has been revoked.\r\nsalloc: Job aborted due to signal\r\n]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;1",,terminal_output
+343,5488299,"TERMINAL",0,0,"salloc --time=00:30:00 --partition=dev_accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5 --mem=50G^C",,terminal_command
+344,5488347,"TERMINAL",0,0,"^C[?2004l\r[?2004h[?2004l\r\r\n]633;E;;26cd839c-476e-4913-967a-1422bf7b3816]633;C]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D",,terminal_output
+345,5507657,"TERMINAL",0,0,"salloc --time=01:30:00 --partition=dev_accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5 --mem=50G",,terminal_command
+346,5507713,"TERMINAL",0,0,"]633;E;2025-07-02 17:52:29 salloc --time=01:30:00 --partition=dev_accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5 --mem=50G;26cd839c-476e-4913-967a-1422bf7b3816]633;Csalloc: error: Job submit/allocate failed: Requested time limit is invalid (missing or exceeds some limit)\r\n]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;1",,terminal_output
+347,5522205,"TERMINAL",0,0,"salloc --time=01:30:00 --partition=accelerated --nodes=1 --ntasks-per-node=4 --gres=gpu:4 --cpus-per-task=5 --mem=50G",,terminal_command
+348,5522257,"TERMINAL",0,0,"]633;E;2025-07-02 17:52:44 salloc --time=01:30:00 --partition=accelerated --nodes=1 --ntasks-per-node=4 --gres=gpu:4 --cpus-per-task=5 --mem=50G;26cd839c-476e-4913-967a-1422bf7b3816]633;Csalloc: Pending job allocation 3312854\r\nsalloc: job 3312854 queued and waiting for resources\r\n",,terminal_output
+349,5528628,"TERMINAL",0,0,"bash",,terminal_focus
+350,5529561,"TERMINAL",0,0,"queue",,terminal_command
+351,5529625,"TERMINAL",0,0,"\r\n[?2004l\r]633;E;2025-07-02 17:52:51 queue;0598f850-442d-4019-9770-f648eaf5abbd]633;C[?1049h(B[?7hEvery 1.0s: squeue --mehkn1993.localdomain: Wed Jul  2 17:52:51 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3312854 accelerat interact tum_cte0 PD\t0:00\t  1 (Priority)3311671 accelerat train_to tum_cte0  R    3:28:35\t  1 hkn0717",,terminal_output
+352,5530673,"TERMINAL",0,0,"26",,terminal_output
+353,5531169,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;0]633;P;Cwd=/home/hk-project-p0023960/tum_cte0515/Projects/jafar[?2004h",,terminal_output
+354,5532122,"TERMINAL",0,0,"idling",,terminal_command
+355,5532190,"TERMINAL",0,0,"]633;E;2025-07-02 17:52:54 idling;0598f850-442d-4019-9770-f648eaf5abbd]633;C[?1049h(B[?7hEvery 1.0s: sinfo_t_idlehkn1993.localdomain: Wed Jul  2 17:52:54 2025Partition dev_cpuonly:     11 nodes idle\rPartition cpuonly:    115 nodes idle\rPartition dev_accelerated:\t 0 nodes idle\rPartition accelerated:\t 0 nodes idle\rPartition dev_accelerated-h100    :\t 0 nodes idle\rPartition accelerated-h100:\t 0 nodes idle\rPartition large:\t 7 nodes idle",,terminal_output
+356,5533235,"TERMINAL",0,0,"5\t ",,terminal_output
+357,5533929,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jafar]633;D;0",,terminal_output
+358,5771719,"TERMINAL",0,0,"salloc",,terminal_focus

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-5e1c58f1-93d2-473f-9eaf-a2de01442cff1758800954786-2025_09_25-13.49.57.480/source.csv

@@ -0,0 +1,268 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,4,"jasmine/models/dynamics.py",0,0,"from typing import Dict\n\nimport jax\nimport jax.numpy as jnp\nimport flax.nnx as nnx\n\nfrom utils.nn import STTransformer, Transformer\n\n\nclass DynamicsMaskGIT(nnx.Module):\n    """"""\n    MaskGIT dynamics model\n\n    Dimension keys:\n        B: batch size\n        T: sequence length\n        N: number of patches per frame\n        L: latent dimension\n        V: vocabulary size (number of latents)\n    """"""\n\n    def __init__(\n        self,\n        model_dim: int,\n        ffn_dim: int,\n        num_latents: int,\n        latent_action_dim: int,\n        num_blocks: int,\n        num_heads: int,\n        dropout: float,\n        mask_limit: float,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        rngs: nnx.Rngs,\n    ):\n        self.model_dim = model_dim\n        self.ffn_dim = ffn_dim\n        self.num_latents = num_latents\n        self.latent_action_dim = latent_action_dim\n        self.num_blocks = num_blocks\n        self.num_heads = num_heads\n        self.dropout = dropout\n        self.mask_limit = mask_limit\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n\n        self.transformer = STTransformer(\n            self.model_dim,\n            self.model_dim,\n            self.ffn_dim,\n            self.num_latents,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n            self.param_dtype,\n            self.dtype,\n            use_flash_attention=self.use_flash_attention,\n            rngs=rngs,\n        )\n        self.patch_embed = nnx.Embed(self.num_latents, self.model_dim, rngs=rngs)\n        self.mask_token = nnx.Param(\n            nnx.initializers.lecun_uniform()(rngs.params(), (1, 1, 1, self.model_dim))\n        )\n        self.action_up = nnx.Linear(\n            self.latent_action_dim,\n            self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n    def __call__(\n        self,\n        batch: Dict[str, jax.Array],\n    ) -> tuple[jax.Array, jax.Array]:\n        # --- Mask videos ---\n        video_tokens_BTN = batch[""video_tokens""]\n        latent_actions_BTm11L = batch[""latent_actions""]\n        vid_embed_BTNM = self.patch_embed(video_tokens_BTN)\n\n        batch_size = vid_embed_BTNM.shape[0]\n        _rng_prob, *_rngs_mask = jax.random.split(batch[""mask_rng""], batch_size + 1)\n        mask_prob = jax.random.uniform(\n            _rng_prob, shape=(batch_size,), minval=self.mask_limit\n        )\n        per_sample_shape = vid_embed_BTNM.shape[1:-1]\n        mask = jax.vmap(\n            lambda rng, prob: jax.random.bernoulli(rng, prob, per_sample_shape),\n            in_axes=(0, 0),\n        )(jnp.asarray(_rngs_mask), mask_prob)\n        mask = mask.at[:, 0].set(False)\n        vid_embed_BTNM = jnp.where(\n            jnp.expand_dims(mask, -1), self.mask_token.value, vid_embed_BTNM\n        )\n\n        # --- Predict transition ---\n        act_embed_BTm11M = self.action_up(latent_actions_BTm11L)\n        padded_act_embed_BT1M = jnp.pad(\n            act_embed_BTm11M, ((0, 0), (1, 0), (0, 0), (0, 0))\n        )\n        padded_act_embed_BTNM = jnp.broadcast_to(\n            padded_act_embed_BT1M, vid_embed_BTNM.shape\n        )\n        vid_embed_BTNM += padded_act_embed_BTNM\n        logits_BTNV = self.transformer(vid_embed_BTNM)\n        return logits_BTNV, mask\n\n\nclass DynamicsCausal(nnx.Module):\n    """"""Causal dynamics model""""""\n\n    def __init__(\n        self,\n        model_dim: int,\n        ffn_dim: int,\n        num_latents: int,\n        latent_action_dim: int,\n        num_blocks: int,\n        num_heads: int,\n        dropout: float,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        decode: bool,\n        rngs: nnx.Rngs,\n    ):\n        self.model_dim = model_dim\n        self.ffn_dim = ffn_dim\n        self.num_latents = num_latents\n        self.latent_action_dim = latent_action_dim\n        self.num_blocks = num_blocks\n        self.num_heads = num_heads\n        self.dropout = dropout\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n        self.decode = decode\n\n        self.transformer = Transformer(\n            self.model_dim,\n            self.model_dim,\n            self.ffn_dim,\n            self.num_latents,\n            self.num_blocks,\n            self.num_heads,\n            self.dropout,\n            self.param_dtype,\n            self.dtype,\n            use_flash_attention=self.use_flash_attention,\n            decode=self.decode,\n            rngs=rngs,\n        )\n        self.patch_embed = nnx.Embed(self.num_latents, self.model_dim, rngs=rngs)\n        self.action_up = nnx.Linear(\n            self.latent_action_dim,\n            self.model_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            rngs=rngs,\n        )\n\n    def __call__(\n        self,\n        batch: Dict[str, jax.Array],\n    ) -> tuple[jax.Array, jax.Array]:\n        video_tokens_BTN = batch[""video_tokens""]\n        latent_actions_BTm11L = batch[""latent_actions""]\n        vid_embed_BTNM = self.patch_embed(video_tokens_BTN)\n        act_embed_BTm11M = self.action_up(latent_actions_BTm11L)\n        padded_act_embed_BT1M = jnp.pad(\n            act_embed_BTm11M, ((0, 0), (1, 0), (0, 0), (0, 0))\n        )\n        vid_embed_BTNp1M = jnp.concatenate(\n            [padded_act_embed_BT1M, vid_embed_BTNM], axis=2\n        )\n        logits_BTNp1V = self.transformer(vid_embed_BTNp1M)\n        logits_BTNV = logits_BTNp1V[:, :, :-1]\n        return logits_BTNV, jnp.ones_like(video_tokens_BTN)\n",python,tab
+2,732,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"1:49:57 PM [info] Activating crowd-code\n1:49:57 PM [info] Recording started\n1:49:57 PM [info] Initializing git provider using file system watchers...\n",Log,tab
+3,1087,"extension-output-pdoom-org.crowd-code-#1-crowd-code",150,0,"1:49:57 PM [info] Git repository found\n1:49:57 PM [info] Git provider initialized successfully\n1:49:57 PM [info] Initial git state: [object Object]\n",Log,content
+4,6981,"TERMINAL",0,0,"bash",,terminal_focus
+5,11991,"TERMINAL",0,0,"queue",,terminal_command
+6,12012,"TERMINAL",0,0,"]633;C[?1049h(B[?7hEvery 1.0s: squeue --mehkn1991.localdomain: Thu Sep 25 13:50:09 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)",,terminal_output
+7,12913,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1991:~/Projects/jasmine",,terminal_output
+8,21218,"TERMINAL",0,0,"salloc --time=05:00:00 --partition=accelerated --nodes=1 --gres=gpu:1  --cpus-per-task=8",,terminal_command
+9,21287,"TERMINAL",0,0,"]633;Csalloc: Pending job allocation 3520566\r\nsalloc: job 3520566 queued and waiting for resources\r\n",,terminal_output
+10,26666,"TERMINAL",0,0,"salloc: job 3520566 has been allocated resources\r\nsalloc: Granted job allocation 3520566\r\nsalloc: Waiting for resource configuration\r\n",,terminal_output
+11,53262,"TERMINAL",0,0,"salloc: Nodes hkn0801 are ready for job\r\n",,terminal_output
+12,53922,"TERMINAL",0,0,"]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h[tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+13,80884,"TERMINAL",0,0,"so",,terminal_output
+14,80963,"TERMINAL",0,0,"u",,terminal_output
+15,81017,"TERMINAL",0,0,"r",,terminal_output
+16,81231,"TERMINAL",0,0,"c",,terminal_output
+17,81353,"TERMINAL",0,0,"e",,terminal_output
+18,81418,"TERMINAL",0,0," ",,terminal_output
+19,81615,"TERMINAL",0,0,".",,terminal_output
+20,81736,"TERMINAL",0,0,"",,terminal_output
+21,82521,"TERMINAL",0,0,"v",,terminal_output
+22,82753,"TERMINAL",0,0,"e",,terminal_output
+23,82846,"TERMINAL",0,0,"n",,terminal_output
+24,82953,"TERMINAL",0,0,"v/",,terminal_output
+25,83176,"TERMINAL",0,0,"b",,terminal_output
+26,83303,"TERMINAL",0,0,"a",,terminal_output
+27,83664,"TERMINAL",0,0,"",,terminal_output
+28,83813,"TERMINAL",0,0,"i",,terminal_output
+29,83923,"TERMINAL",0,0,"n/",,terminal_output
+30,84413,"TERMINAL",0,0,"ac",,terminal_output
+31,84549,"TERMINAL",0,0,"tivate",,terminal_output
+32,85001,"TERMINAL",0,0,"\r\n[?2004l\r]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+33,85218,"TERMINAL",0,0,"\r\n[?2004l\r]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+34,86479,"TERMINAL",0,0,"q",,terminal_output
+35,86549,"TERMINAL",0,0,"u",,terminal_output
+36,86664,"TERMINAL",0,0,"eu",,terminal_output
+37,86813,"TERMINAL",0,0,"e",,terminal_output
+38,86879,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+39,86948,"TERMINAL",0,0,"[?1049h(B[?7hEvery 1.0s: squeue --mehkn0801.localdomain: Thu Sep 25 13:51:24 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3520566 accelerat interact tum_cte0  R\t1:01\t  1 hkn0801",,terminal_output
+40,87731,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+41,88231,"TERMINAL",0,0,"c",,terminal_output
+42,88284,"TERMINAL",0,0,"l",,terminal_output
+43,88350,"TERMINAL",0,0,"e",,terminal_output
+44,88515,"TERMINAL",0,0,"a",,terminal_output
+45,88684,"TERMINAL",0,0,"r",,terminal_output
+46,88766,"TERMINAL",0,0,"\r\n[?2004l\r]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+47,91357,"TERMINAL",0,0,"\r(jasmine) [tum_cte0515@hkn0801 jasmine]$ \r(jasmine) [tum_cte0515@hkn0801 jasmine]$ \r(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+48,91969,"jasmine/models/dynamics.py",0,0,"",python,tab
+49,91972,"jasmine/models/dynamics.py",2538,0,"",python,selection_mouse
+50,92574,"jasmine/models/dynamics.py",2927,0,"",python,selection_mouse
+51,93139,"jasmine/models/dynamics.py",2860,0,"",python,selection_mouse
+52,93960,"jasmine/models/dynamics.py",2926,0,"",python,selection_mouse
+53,93975,"jasmine/models/dynamics.py",2925,0,"",python,selection_command
+54,144844,"TERMINAL",0,0,"q",,terminal_output
+55,144954,"TERMINAL",0,0,"u",,terminal_output
+56,145025,"TERMINAL",0,0,"e",,terminal_output
+57,145132,"TERMINAL",0,0,"u",,terminal_output
+58,145197,"TERMINAL",0,0,"e",,terminal_output
+59,145368,"TERMINAL",0,0,"\r\n[?2004l\r[?1049h(B[?7hEvery 1.0s: squeue --mehkn0801.localdomain: Thu Sep 25 13:52:22 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3520566 accelerat interact tum_cte0  R\t1:59\t  1 hkn0801",,terminal_output
+60,146193,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+61,147631,"TERMINAL",0,0,"g",,terminal_output
+62,147695,"TERMINAL",0,0,"i",,terminal_output
+63,147769,"TERMINAL",0,0,"t",,terminal_output
+64,147833,"TERMINAL",0,0," ",,terminal_output
+65,147977,"TERMINAL",0,0,"b",,terminal_output
+66,148073,"TERMINAL",0,0,"r",,terminal_output
+67,148298,"TERMINAL",0,0,"a",,terminal_output
+68,148385,"TERMINAL",0,0,"c",,terminal_output
+69,148727,"TERMINAL",0,0,"",,terminal_output
+70,149401,"TERMINAL",0,0,"n",,terminal_output
+71,149512,"TERMINAL",0,0,"c",,terminal_output
+72,149885,"TERMINAL",0,0,"h",,terminal_output
+73,150141,"TERMINAL",0,0,"\r\n[?2004l\r[?1h=\r",,terminal_output
+74,150381,"TERMINAL",0,0,"  action-mapper\r\n  add-noise-to-combat-exposure-bias\r\n  add-wandb-name-and-tags\r\n  before-nnx\r\n  causal-mem-reduce\r\n  causal-spatiotemporal-kv-cache\r\n  causal-st-transformer\r\n  causal-transformer-dynamics-model\r\n  causal-transformer-nnx-no-kv-cache\r\n  coinrun-gt-actions\r\n  convert-to-jax-array-in-iter\r\n  correct-batched-sampling\r\n  dev\r\n  dont-let-tf-see-gpu\r\n  feat/darkness-filter\r\n  feat/explicit-image-dims\r\n  fix-action-padding-lam-future-information-access\r\n  fix-sampling\r\n  fix-transformer-forwardpass\r\n  fix/dyn-restore-after-nnx-upgrade\r\n  fix/spatiotemporal-pe-once-in-STTransformer\r\n  generate-minatar-breakout-dataset\r\n  grad-norm-log-and-clip\r\n  grain-dataloader\r\n  gt-actions\r\n  hotfix/eval-full-frame-fix\r\n  hotfix/fix-val-loss-maskgit-masking\r\n  hotfix/full-frame-eval-only-calculate-last-frame-metrics\r\n  hotfix/sampling-shapes-error\r\n  input_pipeline/add-npy2array_record\r\n  logging-variants\r\n  lr-schedules\r\n* main\r\n  maskgit-different-maskprob-per-sample\r\n  maskgit-sampling-iterative-unmasking-fix\r\n  metrics-logging-for-dynamics-model\r\n  monkey-patch\r\n  new-arch-sampling\r\n  preprocess_video\r\n  refactor-full-frame-val-loss\r\n  refactor-tmp\r\n  remove-restore-branching\r\n  revised-dataloader\r\n  runner\r\n  runner-grain\r\n  sample-ali-branch\r\n  sample-from-different-topologies\r\n  sampling-script-add-metrics\r\n  sampling-startframe-indexing-fix\r\n  speedup-tfrecord-preprocessing\r\n  train_lam_coinrun_ablation_wsd_3e-6_28747\r\n  val-loss\r\n\r[?1l>]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+75,152630,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=4\n#SBATCH --time=01:00:00\n#SBATCH --partition=accelerated\n#SBATCH --cpus-per-task=5\n#SBATCH --gres=gpu:4\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/coinrun/dynamics/sampling/maskgit/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/coinrun/dynamics/sampling/maskgit/%x_%j.log\n#SBATCH --job-name=coinrun_sample_maskgit\n\n# Unload modules that may interfere\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\n\n# Activate virtual environment\n# source .venv/bin/activate\n\narray_records_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_breakout/breakout_episodes_perfect/val\n\nCHECKPOINT_PATH=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer_smaller_lr/3519530\n\n\necho ""Sampling from checkpoint: $CHECKPOINT_PATH""\n\nsrun python jasmine/sample.py \\n    --checkpoint ""$CHECKPOINT_PATH"" \\n    --data_dir=$array_records_dir \\n    --seq_len=16 \\n    --batch_size=12 \\n    --start_frame=4 \\n    --image_height=10 \\n    --image_width=10 \\n    --dyna_type=maskgit \\n    --lam_patch_size=4 \\n    --no-print-action-indices \\n    --use_gt_actions \\n    --output_dir ""gifs/50k/gt-actions""",shellscript,tab
+76,153858,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",770,0,"",shellscript,selection_mouse
+77,154737,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",770,0,"\n",shellscript,content
+78,155739,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",770,0,"",shellscript,selection_mouse
+79,156088,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",770,0,"/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer/3519698",shellscript,content
+80,156887,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1071,0,"",shellscript,selection_mouse
+81,157929,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",915,0,"",shellscript,selection_command
+82,158165,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",914,0,"\n",shellscript,content
+83,159368,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",916,0,"",shellscript,selection_command
+84,160343,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",916,0,"#",shellscript,content
+85,160345,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",917,0,"",shellscript,selection_keyboard
+86,160830,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",917,0," ",shellscript,content
+87,160831,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",918,0,"",shellscript,selection_keyboard
+88,160967,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",917,0,"",shellscript,selection_command
+89,245053,"TERMINAL",0,0,"g",,terminal_output
+90,245177,"TERMINAL",0,0,"t",,terminal_output
+91,245888,"TERMINAL",0,0,"",,terminal_output
+92,246100,"TERMINAL",0,0,"i",,terminal_output
+93,246188,"TERMINAL",0,0,"t",,terminal_output
+94,246349,"TERMINAL",0,0," ",,terminal_output
+95,246424,"TERMINAL",0,0,"c",,terminal_output
+96,246555,"TERMINAL",0,0,"he",,terminal_output
+97,246692,"TERMINAL",0,0,"c",,terminal_output
+98,246745,"TERMINAL",0,0,"k",,terminal_output
+99,246921,"TERMINAL",0,0,"o",,terminal_output
+100,246985,"TERMINAL",0,0,"u",,terminal_output
+101,247049,"TERMINAL",0,0,"t",,terminal_output
+102,247114,"TERMINAL",0,0," ",,terminal_output
+103,247497,"TERMINAL",0,0,"add-noise-to-combat-exposure-bias",,terminal_output
+104,247717,"TERMINAL",0,0,"add-noise-to-combat-exposure-bias\r\n[?2004l\r",,terminal_output
+105,248450,"TERMINAL",0,0,"Switched to branch 'add-noise-to-combat-exposure-bias'\r\nYour branch is ahead of 'origin/add-noise-to-combat-exposure-bias' by 3 commits.\r\n  (use ""git push"" to publish your local commits)\r\n]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+106,250505,"",0,0,"Switched from branch 'main' to 'add-noise-to-combat-exposure-bias'",,git_branch_checkout
+107,275721,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",0,0,"#!/usr/bin/env bash\n\n#SBATCH --nodes=1\n#SBATCH --ntasks-per-node=4\n#SBATCH --time=01:00:00\n#SBATCH --partition=accelerated\n#SBATCH --cpus-per-task=5\n#SBATCH --gres=gpu:4\n#SBATCH --output=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/coinrun/dynamics/sampling/maskgit/%x_%j.log\n#SBATCH --error=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/coinrun/dynamics/sampling/maskgit/%x_%j.log\n#SBATCH --job-name=coinrun_sample_maskgit\n\n# Unload modules that may interfere\nmodule unload mpi/openmpi/5.0\nmodule unload devel/cuda/12.4\n\n# Activate virtual environment\n# source .venv/bin/activate\n\narray_records_dir=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/data_breakout/breakout_episodes_perfect/val\n\nCHECKPOINT_PATH=/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer/3519698\n\n# /hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer_smaller_lr/3519530\n\n\necho ""Sampling from checkpoint: $CHECKPOINT_PATH""\n\nsrun python jasmine/sample.py \\n    --checkpoint ""$CHECKPOINT_PATH"" \\n    --data_dir=$array_records_dir \\n    --seq_len=16 \\n    --batch_size=12 \\n    --start_frame=4 \\n    --image_height=10 \\n    --image_width=10 \\n    --dyna_type=maskgit \\n    --lam_patch_size=4 \\n    --no-print-action-indices \\n    --use_gt_actions \\n    --output_dir ""gifs/50k/gt-actions""",shellscript,tab
+108,275722,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1487,0,"",shellscript,selection_mouse
+109,275723,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1486,0,"",shellscript,selection_command
+110,275729,"TERMINAL",0,0,"bash",,terminal_focus
+111,275787,"TERMINAL",0,0,"git git^C[?2004l\r[?2004h[?2004l\r\r\n]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+112,279744,"TERMINAL",0,0,"srun",,terminal_focus
+113,281646,"TERMINAL",0,0,"g",,terminal_output
+114,281705,"TERMINAL",0,0,"i",,terminal_output
+115,281777,"TERMINAL",0,0,"t",,terminal_output
+116,281834,"TERMINAL",0,0," ",,terminal_output
+117,282004,"TERMINAL",0,0,"st",,terminal_output
+118,282157,"TERMINAL",0,0,"a",,terminal_output
+119,282210,"TERMINAL",0,0,"t",,terminal_output
+120,282274,"TERMINAL",0,0,"u",,terminal_output
+121,282538,"TERMINAL",0,0,"s",,terminal_output
+122,282611,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+123,282782,"TERMINAL",0,0,"On branch add-noise-to-combat-exposure-bias\r\nYour branch is ahead of 'origin/add-noise-to-combat-exposure-bias' by 3 commits.\r\n  (use ""git push"" to publish your local commits)\r\n\r\n",,terminal_output
+124,282859,"TERMINAL",0,0,"Last commands done (2 commands done):\r\n   pick ba37453 feat: generate coinrun dataset with val split\r\n   pick faadd10 feat: implemented validation loss for all three models\r\nNext commands to do (26 remaining commands):\r\n   pick 9a17dbb fix: pass val data path to dataloader\r\n   pick 6e69cdb fix typo in image logging\r\n  (use ""git rebase --edit-todo"" to view and edit)\r\nYou are currently editing a commit while rebasing branch 'gt-actions' on 'c7522f2'.\r\n  (use ""git commit --amend"" to amend the current commit)\r\n  (use ""git rebase --continue"" once you are satisfied with your changes)\r\n\r\nUntracked files:\r\n  (use ""git add <file>..."" to include in what will be committed)\r\n\tdiff.diff\r\n\tdiff2.diff\r\n\tinput_pipeline/\r\n\tkiller.sh\r\n\tkiller_partition.sh\r\n\tlog.log\r\n\toverfit_dir.zip\r\n\trequirements-franz.txt\r\n\tsamples/\r\n\tscripts_cremers/\r\n\tslurm/\r\n\ttest.py\r\n\tutils/\r\n\tuv.lock\r\n\r\nnothing added to commit but untracked files present (use ""git add"" to track)\r\n]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+125,294827,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",0,0,"",shellscript,tab
+126,294828,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1475,0,"",shellscript,selection_mouse
+127,295978,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1475,0,"-",shellscript,content
+128,295980,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1476,0,"",shellscript,selection_keyboard
+129,296561,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1476,0,"w",shellscript,content
+130,296563,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1477,0,"",shellscript,selection_keyboard
+131,296654,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1477,0,"-",shellscript,content
+132,296655,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1478,0,"",shellscript,selection_keyboard
+133,297206,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1478,0,"n",shellscript,content
+134,297207,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1479,0,"",shellscript,selection_keyboard
+135,297496,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1479,0,"i",shellscript,content
+136,297497,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1480,0,"",shellscript,selection_keyboard
+137,297797,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1479,1,"",shellscript,content
+138,298013,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1479,0,"p",shellscript,content
+139,298015,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1480,0,"",shellscript,selection_keyboard
+140,298115,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1480,0,"i",shellscript,content
+141,298117,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1481,0,"",shellscript,selection_keyboard
+142,298415,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1480,1,"",shellscript,content
+143,298540,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1479,1,"",shellscript,content
+144,298713,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1479,0,"o",shellscript,content
+145,298715,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1480,0,"",shellscript,selection_keyboard
+146,298844,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1480,0,"i",shellscript,content
+147,298846,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1481,0,"",shellscript,selection_keyboard
+148,298877,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1481,0,"s",shellscript,content
+149,298878,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1482,0,"",shellscript,selection_keyboard
+150,299003,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1482,0,"e",shellscript,content
+151,299004,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1483,0,"",shellscript,selection_keyboard
+152,308200,"TERMINAL",0,0,"q",,terminal_output
+153,308306,"TERMINAL",0,0,"u",,terminal_output
+154,308385,"TERMINAL",0,0,"e",,terminal_output
+155,308467,"TERMINAL",0,0,"ue",,terminal_output
+156,308698,"TERMINAL",0,0,"\r\n[?2004l\r[?1049h(B[?7hEvery 1.0s: squeue --mehkn0801.localdomain: Thu Sep 25 13:55:06 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3520566 accelerat interact tum_cte0  R\t4:43\t  1 hkn0801",,terminal_output
+157,309613,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+158,310080,"TERMINAL",0,0,"c",,terminal_output
+159,310144,"TERMINAL",0,0,"l",,terminal_output
+160,310315,"TERMINAL",0,0,"e",,terminal_output
+161,310520,"TERMINAL",0,0,"a",,terminal_output
+162,310634,"TERMINAL",0,0,"r",,terminal_output
+163,310784,"TERMINAL",0,0,"\r\n[?2004l\r]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+164,316438,"TERMINAL",0,0,"s",,terminal_output
+165,316543,"TERMINAL",0,0,"h",,terminal_output
+166,316636,"TERMINAL",0,0," ",,terminal_output
+167,317338,"TERMINAL",0,0,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",,terminal_output
+168,317731,"TERMINAL",0,0,"\rslurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch\r\n[?2004l\r",,terminal_output
+169,317863,"TERMINAL",0,0,"Sampling from checkpoint: /hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer/3519698\r\n",,terminal_output
+170,317994,"TERMINAL",0,0,"GpuFreq=control_disabled\r\n",,terminal_output
+171,333523,"TERMINAL",0,0,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jasmine/.venv/lib64/python3.12/site-packages/orbax/checkpoint/_src/serialization/type_handlers.py:1269: UserWarning: Sharding info not provided when restoring. Populating sharding info from sharding file. Please note restoration time will be slightly increased due to reading from file. Note also that this option is unsafe when restoring on a different topology than the checkpoint was saved with.\r\n  warnings.warn(\r\n",,terminal_output
+172,363785,"TERMINAL",0,0,"Per-frame SSIM:\r\n [0.30596483 0.414882   0.5163348  0.51084375 0.5155197  0.5155652\r\n 0.51382315 0.4914374  0.45811233 0.4443463  0.44947392 0.4618884 ]\r\nPer-frame PSNR:\r\n [18.745987 18.47467  18.217417 18.06475  18.032337 17.890877 17.707855\r\n 17.542704 17.384125 17.292942 17.300459 17.316257]\r\nSSIM: 0.4665159583091736\r\nPSNR: 17.83086585998535\r\n",,terminal_output
+173,364404,"TERMINAL",0,0,"W0925 13:56:01.732225 2006205 pjrt_client.cc:1469] WatchJobStateAsync failed for task goo.gle/debugproto   job_name: ""jax_worker"": UNAVAILABLE: failed to connect to all addresses; last error: UNKNOWN: ipv4:10.0.1.177:63542: Failed to connect to remote host: Connection refused\r\nAdditional GRPC error information from remote target coordination_service while calling /tensorflow.CoordinationService/WatchJobState:\r\n:UNKNOWN:Error received from peer  {grpc_message:""failed to connect to all addresses; last error: UNKNOWN: ipv4:10.0.1.177:63542: Failed to connect to remote host: Connection refused"", grpc_status:14}\r\n",,terminal_output
+174,365078,"TERMINAL",0,0,"]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+175,372087,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",0,0,"",shellscript,tab
+176,372088,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1425,0,"",shellscript,selection_mouse
+177,372588,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1487,0,"",shellscript,selection_mouse
+178,373273,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1425,0,"",shellscript,selection_mouse
+179,373893,"slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",1448,0,"",shellscript,selection_mouse
+180,442282,"jasmine/sample.py",0,0,"from dataclasses import dataclass\nimport time\nimport os\nimport optax\n\nimport dm_pix as pix\nimport einops\nimport jax\nimport jax.numpy as jnp\nimport flax.linen as nn\nimport numpy as np\nimport orbax.checkpoint as ocp\nfrom PIL import Image, ImageDraw\nimport tyro\nfrom flax import nnx\n\nfrom genie import Genie\nfrom utils.dataloader import get_dataloader\n\n\n@dataclass\nclass Args:\n    # Experiment\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_height: int = 90\n    image_width: int = 160\n    data_dir: str = ""data/coinrun_episodes""\n    checkpoint: str = """"\n    print_action_indices: bool = True\n    output_dir: str = ""gifs/""\n    # Sampling\n    batch_size: int = 1\n    maskgit_steps: int = 25\n    temperature: float = 1.0\n    sample_argmax: bool = True\n    start_frame: int = 1\n    noise_level: float = 0.0\n    noise_buckets: int = 10\n    # Tokenizer checkpoint\n    tokenizer_dim: int = 512\n    tokenizer_ffn_dim: int = 2048\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 4\n    tokenizer_num_blocks: int = 4\n    tokenizer_num_heads: int = 8\n    # LAM checkpoint\n    lam_dim: int = 512\n    lam_ffn_dim: int = 2048\n    latent_action_dim: int = 32\n    num_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 4\n    lam_num_heads: int = 8\n    use_gt_actions: bool = False\n    # Dynamics checkpoint\n    dyna_type: str = ""maskgit""\n    dyna_dim: int = 512\n    dyna_ffn_dim: int = 2048\n    dyna_num_blocks: int = 6\n    dyna_num_heads: int = 8\n    param_dtype = jnp.float32\n    dtype = jnp.bfloat16\n    use_flash_attention: bool = True\n\n\nargs = tyro.cli(Args)\n\nif __name__ == ""__main__"":\n    """"""\n    Dimension keys:\n        B: batch size\n        T: number of input (conditioning) frames\n        N: number of patches per frame\n        S: sequence length\n        H: height\n        W: width\n        E: B * (S - 1)\n    """"""\n    jax.distributed.initialize()\n\n    rng = jax.random.key(args.seed)\n\n    # --- Load Genie checkpoint ---\n    rngs = nnx.Rngs(rng)\n    genie = Genie(\n        # Tokenizer\n        in_dim=args.image_channels,\n        tokenizer_dim=args.tokenizer_dim,\n        tokenizer_ffn_dim=args.tokenizer_ffn_dim,\n        latent_patch_dim=args.latent_patch_dim,\n        num_patch_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        tokenizer_num_blocks=args.tokenizer_num_blocks,\n        tokenizer_num_heads=args.tokenizer_num_heads,\n        # LAM\n        lam_dim=args.lam_dim,\n        lam_ffn_dim=args.lam_ffn_dim,\n        latent_action_dim=args.latent_action_dim,\n        num_actions=args.num_actions,\n        lam_patch_size=args.lam_patch_size,\n        lam_num_blocks=args.lam_num_blocks,\n        lam_num_heads=args.lam_num_heads,\n        lam_co_train=False,\n        max_noise_level=0.0,\n        noise_buckets=args.noise_buckets,\n        use_gt_actions=args.use_gt_actions,\n        # Dynamics\n        dyna_type=args.dyna_type,\n        dyna_dim=args.dyna_dim,\n        dyna_ffn_dim=args.dyna_ffn_dim,\n        dyna_num_blocks=args.dyna_num_blocks,\n        dyna_num_heads=args.dyna_num_heads,\n        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n        use_flash_attention=args.use_flash_attention,\n        # FIXME (f.srambical): implement spatiotemporal KV caching and set decode=True\n        decode=False,\n        rngs=rngs,\n    )\n\n    # Need to delete lam decoder for checkpoint loading\n    if not args.use_gt_actions:\n        assert genie.lam is not None\n        del genie.lam.decoder\n\n    handler_registry = ocp.handlers.DefaultCheckpointHandlerRegistry()\n    handler_registry.add(\n        ""model_state"", ocp.args.PyTreeSave, ocp.handlers.PyTreeCheckpointHandler\n    )\n    handler_registry.add(\n        ""model_state"", ocp.args.PyTreeRestore, ocp.handlers.PyTreeCheckpointHandler\n    )\n    checkpoint_options = ocp.CheckpointManagerOptions(\n        step_format_fixed_length=6,\n    )\n    checkpoint_manager = ocp.CheckpointManager(\n        args.checkpoint,\n        options=checkpoint_options,\n        handler_registry=handler_registry,\n    )\n\n    dummy_tx = optax.adamw(\n        learning_rate=optax.linear_schedule(0.0001, 0.0001, 10000),\n        b1=0.9,\n        b2=0.9,\n        weight_decay=1e-4,\n        mu_dtype=args.dtype,\n    )\n    dummy_optimizer = nnx.ModelAndOptimizer(genie, dummy_tx)\n\n    abstract_optimizer = nnx.eval_shape(lambda: dummy_optimizer)\n    abstract_optimizer_state = nnx.state(abstract_optimizer)\n    restored = checkpoint_manager.restore(\n        checkpoint_manager.latest_step(),\n        args=ocp.args.Composite(\n            model_state=ocp.args.PyTreeRestore(abstract_optimizer_state),  # type: ignore\n        ),\n    )\n    restored_optimizer_state = restored[""model_state""]\n    nnx.update(dummy_optimizer, restored_optimizer_state)\n\n    # --- Define sampling function ---\n    def _sampling_fn(model: Genie, batch: dict) -> jax.Array:\n        """"""Runs Genie.sample with pre-defined generation hyper-parameters.""""""\n        assert args.dyna_type in [\n            ""maskgit"",\n            ""causal"",\n        ], f""Invalid dynamics type: {args.dyna_type}""\n        frames, _ = model.sample(\n            batch,\n            args.seq_len,\n            args.temperature,\n            args.sample_argmax,\n            args.maskgit_steps,\n        )\n        return frames\n\n    # --- Define autoregressive sampling loop ---\n    def _autoreg_sample(genie, rng, batch):\n        batch[""videos""] = batch[""videos""][:, : args.start_frame]\n        batch[""rng""] = rng\n        generated_vid_BSHWC = _sampling_fn(genie, batch)\n        return generated_vid_BSHWC\n\n    # --- Get video + latent actions ---\n    array_record_files = [\n        os.path.join(args.data_dir, x)\n        for x in os.listdir(args.data_dir)\n        if x.endswith("".array_record"")\n    ]\n    dataloader = get_dataloader(\n        array_record_files,\n        args.seq_len,\n        args.batch_size,\n        args.image_height,\n        args.image_width,\n        args.image_channels,\n        # We don't use workers in order to avoid grain shutdown issues (https://github.com/google/grain/issues/398)\n        num_workers=0,\n        prefetch_buffer_size=1,\n        seed=args.seed,\n    )\n    dataloader = iter(dataloader)\n    batch = next(dataloader)\n    gt_video = jnp.asarray(batch[""videos""], dtype=jnp.float32) / 255.0\n    batch[""videos""] = gt_video.astype(args.dtype)\n    # Get latent actions for all videos in the batch\n    action_batch_E = None\n    if not args.use_gt_actions:\n        action_batch_E = genie.vq_encode(batch, training=False)\n        batch[""latent_actions""] = action_batch_E\n\n    # --- Sample + evaluate video ---\n    recon_video_BSHWC = _autoreg_sample(genie, rng, batch)\n    recon_video_BSHWC = recon_video_BSHWC.astype(jnp.float32)\n\n    gt = gt_video.clip(0, 1)[:, args.start_frame :]\n    recon = recon_video_BSHWC.clip(0, 1)[:, args.start_frame :]\n\n    ssim_vmap = jax.vmap(pix.ssim, in_axes=(0, 0))\n    psnr_vmap = jax.vmap(pix.psnr, in_axes=(0, 0))\n    ssim = ssim_vmap(gt, recon)\n    psnr = psnr_vmap(gt, recon)\n    per_frame_ssim = ssim.mean(0)\n    per_frame_psnr = psnr.mean(0)\n    avg_ssim = ssim.mean()\n    avg_psnr = psnr.mean()\n\n    print(""Per-frame SSIM:\n"", per_frame_ssim)\n    print(""Per-frame PSNR:\n"", per_frame_psnr)\n\n    print(f""SSIM: {avg_ssim}"")\n    print(f""PSNR: {avg_psnr}"")\n\n    # --- Construct video ---\n    true_videos = (gt_video * 255).astype(np.uint8)\n    pred_videos = (recon_video_BSHWC * 255).astype(np.uint8)\n    video_comparison = np.zeros((2, *recon_video_BSHWC.shape), dtype=np.uint8)\n    video_comparison[0] = true_videos[:, : args.seq_len]\n    video_comparison[1] = pred_videos\n    frames = einops.rearrange(video_comparison, ""n b t h w c -> t (b h) (n w) c"")\n\n    # --- Save video ---\n    imgs = [Image.fromarray(img) for img in frames]\n    # Write actions on each frame, on each row (i.e., for each video in the batch, on the GT row)\n    B = batch[""videos""].shape[0]\n    if action_batch_E is not None:\n        action_batch_BSm11 = jnp.reshape(action_batch_E, (B, args.seq_len - 1, 1))\n    else:\n        action_batch_BSm11 = jnp.reshape(\n            batch[""actions""][:, :-1], (B, args.seq_len - 1, 1)\n        )\n    for t, img in enumerate(imgs[1:]):\n        d = ImageDraw.Draw(img)\n        for row in range(B):\n            if args.print_action_indices:\n                action = action_batch_BSm11[row, t, 0]\n                y_offset = row * batch[""videos""].shape[2] + 2\n                d.text((2, y_offset), f""{action}"", fill=255)\n\n    os.makedirs(args.output_dir, exist_ok=True)\n    imgs[0].save(\n        os.path.join(args.output_dir, f""generation_{time.time()}.gif""),\n        save_all=True,\n        append_images=imgs[1:],\n        duration=250,\n        loop=0,\n    )\n",python,tab
+181,454590,"jasmine/sample.py",5519,0,"",python,selection_mouse
+182,454728,"jasmine/sample.py",4823,0,"",python,selection_command
+183,457080,"jasmine/sample.py",5118,0,"",python,selection_mouse
+184,457562,"jasmine/genie.py",0,0,"from typing import Dict\n\nimport einops\nimport jax\nimport jax.numpy as jnp\nimport flax.nnx as nnx\nimport orbax.checkpoint as ocp\n\nfrom models.dynamics import DynamicsMaskGIT, DynamicsCausal\nfrom models.lam import LatentActionModel\nfrom models.tokenizer import TokenizerVQVAE\n\n\nclass Genie(nnx.Module):\n    """"""Genie model""""""\n\n    def __init__(\n        self,\n        in_dim: int,\n        tokenizer_dim: int,\n        tokenizer_ffn_dim: int,\n        latent_patch_dim: int,\n        num_patch_latents: int,\n        patch_size: int,\n        tokenizer_num_blocks: int,\n        tokenizer_num_heads: int,\n        lam_dim: int,\n        lam_ffn_dim: int,\n        latent_action_dim: int,\n        num_actions: int,\n        lam_patch_size: int,\n        lam_num_blocks: int,\n        lam_num_heads: int,\n        lam_co_train: bool,\n        use_gt_actions: bool,\n        dyna_type: str,\n        dyna_dim: int,\n        dyna_ffn_dim: int,\n        dyna_num_blocks: int,\n        dyna_num_heads: int,\n        max_noise_level: float,\n        noise_buckets: int,\n        param_dtype: jnp.dtype,\n        dtype: jnp.dtype,\n        use_flash_attention: bool,\n        decode: bool,\n        rngs: nnx.Rngs,\n        dropout: float = 0.0,\n        mask_limit: float = 0.0,\n    ):\n        # --- Tokenizer ---\n        self.in_dim = in_dim\n        self.tokenizer_dim = tokenizer_dim\n        self.tokenizer_ffn_dim = tokenizer_ffn_dim\n        self.latent_patch_dim = latent_patch_dim\n        self.num_patch_latents = num_patch_latents\n        self.patch_size = patch_size\n        self.tokenizer_num_blocks = tokenizer_num_blocks\n        self.tokenizer_num_heads = tokenizer_num_heads\n        # --- LAM ---\n        self.lam_dim = lam_dim\n        self.lam_ffn_dim = lam_ffn_dim\n        self.latent_action_dim = latent_action_dim\n        self.num_actions = num_actions\n        self.lam_patch_size = lam_patch_size\n        self.lam_num_blocks = lam_num_blocks\n        self.lam_num_heads = lam_num_heads\n        self.lam_co_train = lam_co_train\n        self.use_gt_actions = use_gt_actions\n        # --- Dynamics ---\n        self.dyna_type = dyna_type\n        self.dyna_dim = dyna_dim\n        self.dyna_ffn_dim = dyna_ffn_dim\n        self.dyna_num_blocks = dyna_num_blocks\n        self.dyna_num_heads = dyna_num_heads\n        self.max_noise_level = max_noise_level\n        self.noise_buckets = noise_buckets\n        self.param_dtype = param_dtype\n        self.dtype = dtype\n        self.use_flash_attention = use_flash_attention\n        self.dropout = dropout\n        self.mask_limit = mask_limit\n        self.decode = decode\n\n        self.tokenizer = TokenizerVQVAE(\n            in_dim=self.in_dim,\n            model_dim=self.tokenizer_dim,\n            ffn_dim=self.tokenizer_ffn_dim,\n            latent_dim=self.latent_patch_dim,\n            num_latents=self.num_patch_latents,\n            patch_size=self.patch_size,\n            num_blocks=self.tokenizer_num_blocks,\n            num_heads=self.tokenizer_num_heads,\n            dropout=0.0,\n            codebook_dropout=0.0,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            use_flash_attention=self.use_flash_attention,\n            rngs=rngs,\n        )\n        if self.use_gt_actions:\n            self.action_embed = nnx.Embed(\n                self.num_actions, self.latent_action_dim, rngs=rngs\n            )\n            self.lam = None\n        else:\n            self.lam = LatentActionModel(\n                in_dim=self.in_dim,\n                model_dim=self.lam_dim,\n                ffn_dim=self.lam_ffn_dim,\n                latent_dim=self.latent_patch_dim,\n                num_latents=self.num_actions,\n                patch_size=self.lam_patch_size,\n                num_blocks=self.lam_num_blocks,\n                num_heads=self.lam_num_heads,\n                dropout=0.0,\n                codebook_dropout=0.0,\n                param_dtype=self.param_dtype,\n                dtype=self.dtype,\n                use_flash_attention=self.use_flash_attention,\n                rngs=rngs,\n            )\n            self.action_embed = None\n        if self.dyna_type == ""maskgit"":\n            self.dynamics = DynamicsMaskGIT(\n                model_dim=self.dyna_dim,\n                ffn_dim=self.dyna_ffn_dim,\n                num_latents=self.num_patch_latents,\n                latent_action_dim=self.latent_action_dim,\n                num_blocks=self.dyna_num_blocks,\n                num_heads=self.dyna_num_heads,\n                dropout=self.dropout,\n                mask_limit=self.mask_limit,\n                max_noise_level=self.max_noise_level,\n                noise_buckets=self.noise_buckets,\n                param_dtype=self.param_dtype,\n                dtype=self.dtype,\n                use_flash_attention=self.use_flash_attention,\n                rngs=rngs,\n            )\n        elif self.dyna_type == ""causal"":\n            self.dynamics = DynamicsCausal(\n                model_dim=self.dyna_dim,\n                ffn_dim=self.dyna_ffn_dim,\n                num_latents=self.num_patch_latents,\n                latent_action_dim=self.latent_action_dim,\n                num_blocks=self.dyna_num_blocks,\n                num_heads=self.dyna_num_heads,\n                dropout=self.dropout,\n                max_noise_level=self.max_noise_level,\n                noise_buckets=self.noise_buckets,\n                param_dtype=self.param_dtype,\n                dtype=self.dtype,\n                use_flash_attention=self.use_flash_attention,\n                decode=decode,\n                rngs=rngs,\n            )\n        else:\n            raise ValueError(f""Invalid dynamics type: {self.dyna_type}"")\n\n    def __call__(\n        self,\n        batch: Dict[str, jax.Array],\n    ) -> Dict[str, jax.Array]:\n        videos_BTHWC = batch[""videos""]\n        tokenizer_outputs = self.tokenizer.vq_encode(videos_BTHWC, training=False)\n        token_indices_BTN = tokenizer_outputs[""indices""]\n        latent_actions_BTm11L = None\n        action_embeddings_BTm11L = None\n        if self.use_gt_actions:\n            assert self.action_embed is not None\n            action_indices_E = None\n            action_embeddings_BT1L = self.action_embed(batch[""actions""]).reshape(\n                *batch[""actions""].shape[:2], 1, self.latent_action_dim\n            )\n            action_embeddings_BTm11L = action_embeddings_BT1L[:, :-1]\n        else:\n            assert self.lam is not None\n            lam_outputs = self.lam.vq_encode(videos_BTHWC, training=False)\n            z_q_BTm11L = lam_outputs[""z_q""]\n            action_indices_E = lam_outputs[""indices""]\n            latent_actions_BTm11L = jax.lax.cond(\n                self.lam_co_train,\n                lambda: z_q_BTm11L,\n                lambda: jax.lax.stop_gradient(z_q_BTm11L),\n            )\n        outputs = dict(\n            video_tokens=jax.lax.stop_gradient(token_indices_BTN),\n            latent_actions=(\n                action_embeddings_BTm11L\n                if self.use_gt_actions\n                else latent_actions_BTm11L\n            ),\n        )\n        outputs[""mask_rng""] = batch[""rng""]\n        dyna_logits_BTNV, dyna_mask = self.dynamics(outputs)\n        outputs[""token_logits""] = dyna_logits_BTNV\n        outputs[""mask""] = dyna_mask\n        mle_indices_BTN = jnp.argmax(outputs[""token_logits""], axis=-1)\n        H, W = batch[""videos""].shape[2:4]\n        outputs[""recon""] = self.tokenizer.decode(mle_indices_BTN, (H, W))\n        if action_indices_E is not None:\n            outputs[""lam_indices""] = action_indices_E\n        return outputs\n\n    def sample(\n        self,\n        batch: Dict[str, jax.Array],\n        seq_len: int,\n        temperature: float = 1,\n        sample_argmax: bool = False,\n        maskgit_steps: int = 25,\n    ) -> tuple[jax.Array, jax.Array]:\n        if self.dyna_type == ""maskgit"":\n            return self.sample_maskgit(\n                batch, seq_len, 0.0, maskgit_steps, temperature, sample_argmax\n            )\n        elif self.dyna_type == ""causal"":\n            return self.sample_causal(batch, seq_len, temperature, sample_argmax)\n        else:\n            raise ValueError(f""Dynamics model type unknown: {self.dyna_type}"")\n\n    def sample_maskgit(\n        self,\n        batch: Dict[str, jax.Array],\n        seq_len: int,\n        noise_level: float,\n        steps: int = 25,\n        temperature: float = 1,\n        sample_argmax: bool = False,\n    ) -> tuple[jax.Array, jax.Array]:\n        """"""\n        Autoregressively samples up to `seq_len` future frames, following Figure 8 of the paper.\n\n        - Input frames are tokenized once.\n        - Future frames are generated autoregressively in token space.\n        - All frames are detokenized in a single pass.\n\n        Note:\n        - For interactive or step-wise sampling, detokenization should occur after each action.\n        - To maintain consistent tensor shapes across timesteps, all current and future frames are decoded at every step.\n        - Temporal causal structure is preserved by\n            a) reapplying the mask before each decoding step.\n            b) a temporal causal mask is applied within each ST-transformer block.\n\n        Dimension keys:\n            B: batch size\n            T: number of input (conditioning) frames\n            N: number of patches per frame\n            M: model dimension\n            S: sequence length\n            H: height\n            W: width\n            E: B * (S - 1)\n            P: S * N\n        """"""\n        assert isinstance(self.dynamics, DynamicsMaskGIT)\n        # --- Encode videos and actions ---\n        videos_BTHWC = batch[""videos""]\n        tokenizer_out = self.tokenizer.vq_encode(videos_BTHWC, training=False)\n        token_idxs_BTN = tokenizer_out[""indices""]\n        B, T, N = token_idxs_BTN.shape\n        pad_shape = (B, seq_len - T, N)\n        pad = jnp.zeros(pad_shape, dtype=token_idxs_BTN.dtype)\n        token_idxs_BSN = jnp.concatenate([token_idxs_BTN, pad], axis=1)\n        init_logits_BSNV = jnp.zeros(\n            shape=(*token_idxs_BSN.shape, self.num_patch_latents)\n        )\n        noise_level = jnp.array(noise_level)\n        if self.use_gt_actions:\n            assert self.action_embed is not None\n            latent_actions_BT1L = self.action_embed(batch[""actions""]).reshape(\n                *batch[""actions""].shape[:2], 1, self.latent_action_dim\n            )\n            latent_actions_BTm11L = latent_actions_BT1L[:, :-1]\n            action_tokens_EL = latent_actions_BTm11L.reshape(-1, self.latent_action_dim)\n        else:\n            assert self.lam is not None\n            latent_actions_E = batch[""latent_actions""]\n            action_tokens_EL = self.lam.vq.get_codes(latent_actions_E)\n\n        # --- Extract submodule state ---\n        dynamics_state = nnx.state(self.dynamics)\n\n        @nnx.scan(in_axes=(nnx.Carry, 0), out_axes=nnx.Carry)\n        def maskgit_step_fn(\n            carry: tuple[jax.Array, jax.Array, jax.Array, jax.Array, jax.Array],\n            step: jax.Array,\n        ) -> tuple[jax.Array, jax.Array, jax.Array, jax.Array, jax.Array]:\n            rng, token_idxs_BSN, logits_BSNV, mask_BSN, action_tokens_EL = carry\n            S, N = token_idxs_BSN.shape[1:]\n            L = action_tokens_EL.shape[-1]\n\n            # We need to reconstruct the submodule inside scan body to prevent trace context mismatches\n            dynamics_maskgit = DynamicsMaskGIT(\n                model_dim=self.dyna_dim,\n                ffn_dim=self.dyna_ffn_dim,\n                num_latents=self.num_patch_latents,\n                latent_action_dim=self.latent_action_dim,\n                num_blocks=self.dyna_num_blocks,\n                num_heads=self.dyna_num_heads,\n                dropout=self.dropout,\n                max_noise_level=self.max_noise_level,\n                noise_buckets=self.noise_buckets,\n                mask_limit=self.mask_limit,\n                param_dtype=self.param_dtype,\n                dtype=self.dtype,\n                use_flash_attention=self.use_flash_attention,\n                rngs=nnx.Rngs(0),\n            )\n            nnx.update(dynamics_maskgit, dynamics_state)\n\n            # --- Construct + encode video ---\n            vid_embed_BSNM = dynamics_maskgit.patch_embed(token_idxs_BSN)\n            mask_token_111M = dynamics_maskgit.mask_token.value\n            mask_expanded_BSN1 = mask_BSN[..., None]\n            vid_embed_BSNM = jnp.where(\n                mask_expanded_BSN1, mask_token_111M, vid_embed_BSNM\n            )\n\n            # --- Predict transition ---\n            action_tokens_BSm1L = jnp.reshape(action_tokens_EL, (B, S - 1, L))\n            act_embed_BSm1M = dynamics_maskgit.action_up(action_tokens_BSm1L)\n            act_embed_BSM = jnp.pad(act_embed_BSm1M, ((0, 0), (1, 0), (0, 0)))\n            act_embed_BS1M = jnp.reshape(\n                act_embed_BSM, (B, S, 1, act_embed_BSM.shape[-1])\n            )\n            # TODO mihir\n\n            rng, _rng_noise = jax.random.split(rng)\n            noise_level_111 = noise_level.reshape(1, 1, 1)\n            noise_level_B11 = jnp.tile(noise_level_111, (B, 1, 1))\n            noise_bucket_idx_B11 = jnp.floor(\n                (noise_level_B11 / self.max_noise_level) * self.noise_buckets\n            ).astype(jnp.int32)\n            noise_level_embed_B11M = dynamics_maskgit.noise_level_embed(\n                noise_bucket_idx_B11\n            )\n            noise_level_embed_BS1M = jnp.tile(noise_level_embed_B11M, (1, S, 1, 1))\n            vid_embed_BSNM += jnp.expand_dims(noise_level_B11, -1)\n\n            vid_embed_BSNp2M = jnp.concatenate(\n                [act_embed_BS1M, noise_level_embed_BS1M, vid_embed_BSNM], axis=2\n            )\n            unmasked_ratio = jnp.cos(jnp.pi * (step + 1) / (steps * 2))\n            step_temp = temperature * (1.0 - unmasked_ratio)\n            final_logits_BSNp2V = (\n                dynamics_maskgit.transformer(vid_embed_BSNp2M) / step_temp\n            )\n            final_logits_BSNV = final_logits_BSNp2V[:, :, 2:]\n\n            # --- Sample new tokens for final frame ---\n            if sample_argmax:\n                sampled_token_idxs_BSN = jnp.argmax(final_logits_BSNV, axis=-1)\n            else:\n                rng, _rng = jax.random.split(rng)\n                sampled_token_idxs_BSN = jax.random.categorical(_rng, final_logits_BSNV)\n            gather_fn = jax.vmap(jax.vmap(jax.vmap(lambda x, y: x[y])))\n            final_token_probs_BSN = gather_fn(\n                jax.nn.softmax(final_logits_BSNV), sampled_token_idxs_BSN\n            )\n            final_token_probs_BSN += ~mask_BSN\n            # Update masked tokens and logits only\n            token_idxs_BSN = jnp.where(mask_BSN, sampled_token_idxs_BSN, token_idxs_BSN)\n            logits_BSNV = jnp.where(\n                jnp.expand_dims(mask_BSN, -1), final_logits_BSNV, logits_BSNV\n            )\n\n            # --- Update mask ---\n            num_unmasked_tokens = jnp.round(N * (1.0 - unmasked_ratio)).astype(int)\n            final_token_probs_flat_BP = einops.rearrange(\n                final_token_probs_BSN, ""b s n -> b (s n)""\n            )\n            idx_mask_P = (\n                jnp.arange(final_token_probs_flat_BP.shape[-1])\n                <= N - num_unmasked_tokens\n            )\n            sorted_idxs_BP = jnp.argsort(final_token_probs_flat_BP, axis=-1)\n            mask_update_fn = jax.vmap(lambda msk, ids: msk.at[ids].set(idx_mask_P))\n            mask_flat_BP = einops.rearrange(mask_BSN, ""b s n -> b (s n)"")\n            new_mask_flat_BP = mask_update_fn(mask_flat_BP, sorted_idxs_BP)\n            new_mask_BSN = einops.rearrange(new_mask_flat_BP, ""b (s n) -> b s n"", n=N)\n\n            new_carry = (\n                rng,\n                token_idxs_BSN,\n                logits_BSNV,\n                new_mask_BSN,\n                action_tokens_EL,\n            )\n            return new_carry\n\n        @nnx.scan(in_axes=(nnx.Carry, 0), out_axes=nnx.Carry)\n        def generation_step_fn(\n            carry: tuple[jax.Array, jax.Array, jax.Array], step_t: jax.Array\n        ) -> tuple[jax.Array, jax.Array, jax.Array]:\n            rng, current_token_idxs_BSN, current_logits_BSNV = carry\n            rng, step_rng = jax.random.split(rng)\n\n            # Mask current frame (i.e., t == step_t)\n            mask_S = jnp.arange(seq_len) == step_t\n            mask_BSN = jnp.broadcast_to(mask_S[None, :, None], (B, seq_len, N)).astype(\n                bool\n            )\n            masked_token_idxs_BSN = current_token_idxs_BSN * ~mask_BSN\n            masked_logits_BSNV = current_logits_BSNV * jnp.expand_dims(~mask_BSN, -1)\n\n            # --- Initialize and run MaskGIT loop ---\n            init_carry_maskgit = (\n                step_rng,\n                masked_token_idxs_BSN,\n                masked_logits_BSNV,\n                mask_BSN,\n                action_tokens_EL,\n            )\n            final_carry_maskgit = maskgit_step_fn(init_carry_maskgit, jnp.arange(steps))\n            updated_token_idxs_BSN = final_carry_maskgit[1]\n            updated_logits_BSNV = final_carry_maskgit[2]\n            new_carry = (rng, updated_token_idxs_BSN, updated_logits_BSNV)\n            return new_carry\n\n        # --- Run the autoregressive generation using jax.lax.scan ---\n        initial_carry = (batch[""rng""], token_idxs_BSN, init_logits_BSNV)\n        timesteps_to_scan = jnp.arange(T, seq_len)\n        final_carry = generation_step_fn(initial_carry, timesteps_to_scan)\n        final_token_idxs_BSN = final_carry[1]\n        final_logits_BSNV = final_carry[2]\n\n        # --- Decode all tokens at once at the end ---\n        H, W = batch[""videos""].shape[2:4]\n        final_frames_BSHWC = self.tokenizer.decode(\n            final_token_idxs_BSN,\n            video_hw=(H, W),\n        )\n        return final_frames_BSHWC, final_logits_BSNV\n\n    def sample_causal(\n        self,\n        batch: Dict[str, jax.Array],\n        seq_len: int,\n        temperature: float = 1,\n        sample_argmax: bool = False,\n    ) -> tuple[jax.Array, jax.Array]:\n        """"""\n        Autoregressively samples up to `seq_len` future frames, following Figure 8 of the paper.\n\n        - Input frames are tokenized once.\n        - Future frames are generated autoregressively in token space.\n        - All frames are detokenized in a single pass.\n\n        Note:\n        - For interactive or step-wise sampling, detokenization should occur after each action.\n        - To maintain consistent tensor shapes across timesteps, all current and future frames are decoded at every step.\n        - Temporal causal structure is preserved by\n            a) reapplying the mask before each decoding step.\n            b) a temporal causal mask is applied within each ST-transformer block.\n\n        Dimension keys:\n            B: batch size\n            T: number of input (conditioning) frames\n            N: number of patches per frame\n            M: model dimension\n            S: sequence length\n            H: height\n            W: width\n            E: B * (S - 1)\n        """"""\n        assert isinstance(self.dynamics, DynamicsCausal)\n        # --- Encode videos and actions ---\n        videos_BTHWC = batch[""videos""]\n        tokenizer_out = self.tokenizer.vq_encode(videos_BTHWC, training=False)\n        token_idxs_BTN = tokenizer_out[""indices""]\n        B, T, N = token_idxs_BTN.shape\n        pad_shape = (B, seq_len - T, N)\n        pad = jnp.zeros(pad_shape, dtype=token_idxs_BTN.dtype)\n        token_idxs_BSN = jnp.concatenate([token_idxs_BTN, pad], axis=1)\n        logits_BSNV = jnp.zeros((*token_idxs_BSN.shape, self.num_patch_latents))\n        dynamics_state = nnx.state(self.dynamics)\n\n        if self.use_gt_actions:\n            assert self.action_embed is not None\n            latent_actions_BT1L = self.action_embed(batch[""actions""]).reshape(\n                *batch[""actions""].shape[:2], 1, self.latent_action_dim\n            )\n            latent_actions_BTm11L = latent_actions_BT1L[:, :-1]\n            action_tokens_EL = latent_actions_BTm11L.reshape(-1, self.latent_action_dim)\n        else:\n            assert self.lam is not None\n            latent_actions_E = batch[""latent_actions""]\n            action_tokens_EL = self.lam.vq.get_codes(latent_actions_E)\n\n        @nnx.scan(in_axes=(nnx.Carry, 0), out_axes=nnx.Carry)\n        def causal_step_fn(\n            carry: tuple[jax.Array, jax.Array, jax.Array, jax.Array, jax.Array],\n            step_n: jax.Array,\n        ) -> tuple[jax.Array, jax.Array, jax.Array, jax.Array, jax.Array]:\n            rng, token_idxs_BSN, logits_BSNV, action_tokens_EL, step_t = carry\n            S, N = token_idxs_BSN.shape[1:]\n            L = action_tokens_EL.shape[-1]\n\n            # We need to reconstruct the submodule inside scan body to prevent trace context mismatches\n            dynamics_causal = DynamicsCausal(\n                model_dim=self.dyna_dim,\n                ffn_dim=self.dyna_ffn_dim,\n                num_latents=self.num_patch_latents,\n                latent_action_dim=self.latent_action_dim,\n                num_blocks=self.dyna_num_blocks,\n                num_heads=self.dyna_num_heads,\n                dropout=self.dropout,\n                param_dtype=self.param_dtype,\n                dtype=self.dtype,\n                use_flash_attention=self.use_flash_attention,\n                decode=self.decode,\n                rngs=nnx.Rngs(0),\n            )\n            nnx.update(dynamics_causal, dynamics_state)\n\n            # --- Construct + encode video ---\n            vid_embed_BSNM = dynamics_causal.patch_embed(token_idxs_BSN)\n\n            # --- Predict transition ---\n            action_tokens_BSm1L = jnp.reshape(action_tokens_EL, (B, S - 1, L))\n            act_embed_BSm1M = dynamics_causal.action_up(action_tokens_BSm1L)\n            act_embed_BSM = jnp.pad(act_embed_BSm1M, ((0, 0), (1, 0), (0, 0)))\n            act_embed_BS1M = jnp.reshape(\n                act_embed_BSM, (B, S, 1, act_embed_BSM.shape[-1])\n            )\n            vid_embed_BSNp1M = jnp.concatenate([act_embed_BS1M, vid_embed_BSNM], axis=2)\n            final_logits_BTNp1V = (\n                dynamics_causal.transformer(vid_embed_BSNp1M, (step_t, step_n))\n                / temperature\n            )\n            final_logits_BV = final_logits_BTNp1V[:, step_t, step_n, :]\n\n            # --- Sample new tokens for final frame ---\n            if sample_argmax:\n                sampled_token_idxs_B = jnp.argmax(final_logits_BV, axis=-1)\n            else:\n                rng, _rng = jax.random.split(rng)\n                sampled_token_idxs_B = jax.random.categorical(_rng, final_logits_BV)\n            # Update next tokens only\n            token_idxs_BSN = token_idxs_BSN.at[:, step_t, step_n].set(\n                sampled_token_idxs_B\n            )\n            logits_BSNV = logits_BSNV.at[:, step_t, step_n].set(final_logits_BV)\n\n            new_carry = (rng, token_idxs_BSN, logits_BSNV, action_tokens_EL, step_t)\n            return new_carry\n\n        @nnx.scan(in_axes=(nnx.Carry, 0), out_axes=nnx.Carry)\n        def generation_step_fn(\n            carry: tuple[jax.Array, jax.Array, jax.Array], step_t: jax.Array\n        ) -> tuple[jax.Array, jax.Array, jax.Array]:\n            rng, current_token_idxs_BSN, current_logits_BSNV = carry\n            rng, step_rng = jax.random.split(rng)\n\n            # --- Initialize and run causal loop ---\n            init_carry_causal = (\n                step_rng,\n                current_token_idxs_BSN,\n                current_logits_BSNV,\n                action_tokens_EL,\n                step_t,\n            )\n            final_carry_causal = causal_step_fn(init_carry_causal, jnp.arange(N))\n            updated_token_idxs_BSN = final_carry_causal[1]\n            updated_logits_BSNV = final_carry_causal[2]\n            new_carry = (rng, updated_token_idxs_BSN, updated_logits_BSNV)\n            return new_carry\n\n        # --- Run the autoregressive generation using jax.lax.scan ---\n        initial_carry = (batch[""rng""], token_idxs_BSN, logits_BSNV)\n        timesteps_to_scan = jnp.arange(T, seq_len)\n        final_carry = generation_step_fn(initial_carry, timesteps_to_scan)\n        final_token_idxs_BSN = final_carry[1]\n        final_logits_BSNV = final_carry[2]\n\n        # --- Decode all tokens at once at the end ---\n        H, W = batch[""videos""].shape[2:4]\n        final_frames_BSHWC = self.tokenizer.decode(\n            final_token_idxs_BSN,\n            video_hw=(H, W),\n        )\n        return final_frames_BSHWC, final_logits_BSNV\n\n    def vq_encode(self, batch: Dict[str, jax.Array], training: bool) -> jax.Array:\n        # --- Preprocess videos ---\n        assert self.lam is not None\n        video_BTHWC = batch[""videos""]\n        lam_output: Dict[str, jax.Array] = self.lam.vq_encode(\n            video_BTHWC, training=training\n        )\n        lam_indices_E = lam_output[""indices""]\n        return lam_indices_E\n\n\n# FIXME (f.srambical): add conversion script for old checkpoints\ndef restore_genie_components(\n    optimizer: nnx.ModelAndOptimizer,\n    sharding: jax.sharding.NamedSharding,\n    rng: jax.Array,\n    args,\n) -> nnx.ModelAndOptimizer:\n    """"""Restore pre-trained Genie components""""""\n    rng_tokenizer, rng_lam = jax.random.split(rng)\n    rngs_tokenizer = nnx.Rngs(rng_tokenizer)\n    rngs_lam = nnx.Rngs(rng_lam)\n\n    tx = optimizer.tx\n    model = optimizer.model\n    handler_registry = ocp.handlers.DefaultCheckpointHandlerRegistry()\n    handler_registry.add(\n        ""model_state"", ocp.args.PyTreeRestore, ocp.handlers.PyTreeCheckpointHandler\n    )\n\n    checkpoint_options = ocp.CheckpointManagerOptions(\n        step_format_fixed_length=6,\n    )\n    tokenizer_checkpoint_manager = ocp.CheckpointManager(\n        directory=args.tokenizer_checkpoint,\n        options=checkpoint_options,\n        handler_registry=handler_registry,\n    )\n    dummy_tokenizer = TokenizerVQVAE(\n        in_dim=args.image_channels,\n        model_dim=args.tokenizer_dim,\n        ffn_dim=args.tokenizer_ffn_dim,\n        latent_dim=args.latent_patch_dim,\n        num_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        num_blocks=args.tokenizer_num_blocks,\n        num_heads=args.tokenizer_num_heads,\n        dropout=args.dropout,\n        codebook_dropout=args.dropout,\n        param_dtype=args.param_dtype,\n        dtype=args.dtype,\n        use_flash_attention=args.use_flash_attention,\n        rngs=rngs_tokenizer,\n    )\n    dummy_tokenizer_optimizer = nnx.ModelAndOptimizer(dummy_tokenizer, tx)\n    dummy_tokenizer_optimizer_state = nnx.state(dummy_tokenizer_optimizer)\n    abstract_sharded_tokenizer_optimizer_state = _create_abstract_sharded_pytree(\n        dummy_tokenizer_optimizer_state, sharding\n    )\n    restored_tokenizer = tokenizer_checkpoint_manager.restore(\n        step=tokenizer_checkpoint_manager.latest_step(),\n        args=ocp.args.Composite(\n            model_state=ocp.args.PyTreeRestore(  # type: ignore\n                abstract_sharded_tokenizer_optimizer_state  # type: ignore\n            ),\n        ),\n    )[""model_state""]\n    nnx.update(dummy_tokenizer_optimizer.model, restored_tokenizer.model)\n    model.tokenizer = dummy_tokenizer_optimizer.model\n    tokenizer_checkpoint_manager.close()\n\n    if args.lam_checkpoint:\n        lam_checkpoint_manager = ocp.CheckpointManager(\n            directory=args.lam_checkpoint,\n            options=checkpoint_options,\n            handler_registry=handler_registry,\n        )\n        dummy_lam = LatentActionModel(\n            in_dim=args.image_channels,\n            model_dim=args.lam_dim,\n            ffn_dim=args.lam_ffn_dim,\n            latent_dim=args.latent_patch_dim,\n            num_latents=args.num_actions,\n            patch_size=args.lam_patch_size,\n            num_blocks=args.lam_num_blocks,\n            num_heads=args.lam_num_heads,\n            dropout=args.dropout,\n            codebook_dropout=args.dropout,\n            param_dtype=args.param_dtype,\n            dtype=args.dtype,\n            use_flash_attention=args.use_flash_attention,\n            rngs=rngs_lam,\n        )\n        dummy_lam_optimizer = nnx.ModelAndOptimizer(dummy_lam, tx)\n        dummy_lam_optimizer_state = nnx.state(dummy_lam_optimizer)\n        abstract_sharded_lam_optimizer_state = _create_abstract_sharded_pytree(\n            dummy_lam_optimizer_state, sharding\n        )\n        restored_lam_optimizer = lam_checkpoint_manager.restore(\n            step=lam_checkpoint_manager.latest_step(),\n            args=ocp.args.Composite(\n                model_state=ocp.args.PyTreeRestore(  # type: ignore\n                    abstract_sharded_lam_optimizer_state  # type: ignore\n                ),\n            ),\n        )[""model_state""]\n        nnx.update(dummy_lam_optimizer.model, restored_lam_optimizer.model)\n        model.lam = dummy_lam_optimizer.model\n        # Remove the LAM decoder to save memory and avoid unnecessary computation.\n        del model.lam.decoder\n        lam_checkpoint_manager.close()\n\n    # Reinitialize the optimizer states\n    optimizer = nnx.ModelAndOptimizer(model, tx)\n    return optimizer\n\n\ndef _create_abstract_sharded_pytree(\n    pytree_template: nnx.GraphState, sharding_spec: jax.sharding.NamedSharding\n) -> jax.Array:\n    """"""Replaces arrays in a pytree with ShapeDtypeStructs having the given sharding.""""""\n\n    def map_fn(leaf_template):\n        if hasattr(leaf_template, ""shape"") and hasattr(leaf_template, ""dtype""):\n            return jax.ShapeDtypeStruct(\n                leaf_template.shape, leaf_template.dtype, sharding=sharding_spec\n            )\n        return leaf_template\n\n    return jax.tree_util.tree_map(map_fn, pytree_template)\n",python,tab
+185,457564,"jasmine/genie.py",7548,0,"",python,selection_command
+186,460084,"jasmine/genie.py",7881,0,"",python,selection_mouse
+187,460257,"jasmine/genie.py",7881,1,"0",python,selection_mouse
+188,460258,"jasmine/genie.py",7881,2,"0.",python,selection_mouse
+189,460304,"jasmine/genie.py",7881,3,"0.0",python,selection_mouse
+190,460682,"jasmine/genie.py",7884,0,"",python,selection_mouse
+191,461284,"jasmine/genie.py",7881,0,"",python,selection_mouse
+192,461418,"jasmine/genie.py",7881,1,"0",python,selection_mouse
+193,461475,"jasmine/genie.py",7881,2,"0.",python,selection_mouse
+194,461475,"jasmine/genie.py",7881,3,"0.0",python,selection_mouse
+195,461861,"jasmine/genie.py",7884,0,"",python,selection_mouse
+196,468902,"jasmine/genie.py",7883,1,"",python,content
+197,469934,"jasmine/genie.py",7883,0,"3",python,content
+198,469935,"jasmine/genie.py",7884,0,"",python,selection_keyboard
+199,471302,"TERMINAL",0,0,"sh slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",,terminal_output
+200,472807,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+201,472919,"TERMINAL",0,0,"Sampling from checkpoint: /hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer/3519698\r\n",,terminal_output
+202,473043,"TERMINAL",0,0,"GpuFreq=control_disabled\r\n",,terminal_output
+203,481577,"TERMINAL",0,0,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jasmine/.venv/lib64/python3.12/site-packages/orbax/checkpoint/_src/serialization/type_handlers.py:1269: UserWarning: Sharding info not provided when restoring. Populating sharding info from sharding file. Please note restoration time will be slightly increased due to reading from file. Note also that this option is unsafe when restoring on a different topology than the checkpoint was saved with.\r\n  warnings.warn(\r\n",,terminal_output
+204,511471,"TERMINAL",0,0,"Per-frame SSIM:\r\n [0.30596483 0.414882   0.5163348  0.51084375 0.5155197  0.5155652\r\n 0.51382315 0.4914374  0.45811233 0.4443463  0.44947392 0.4618884 ]\r\nPer-frame PSNR:\r\n [18.745987 18.47467  18.217417 18.06475  18.032337 17.890877 17.707855\r\n 17.542704 17.384125 17.292942 17.300459 17.316257]\r\nSSIM: 0.4665159583091736\r\nPSNR: 17.83086585998535\r\n",,terminal_output
+205,511990,"TERMINAL",0,0,"W0925 13:58:29.386010 2007888 pjrt_client.cc:1469] WatchJobStateAsync failed for task goo.gle/debugstr   job_name: ""jax_worker"": CANCELLED: CANCELLED\r\nAdditional GRPC error information from remote target coordination_service while calling /tensorflow.CoordinationService/WatchJobState:\r\n:UNKNOWN:Error received from peer  {grpc_message:""CANCELLED"", grpc_status:1} [type.googleapis.com/tensorflow.DerivedStatus='']\r\n",,terminal_output
+206,512521,"TERMINAL",0,0,"]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+207,541605,"jasmine/genie.py",0,0,"",python,tab
+208,541607,"jasmine/genie.py",7883,0,"",python,selection_mouse
+209,542863,"jasmine/genie.py",7883,1,"",python,content
+210,542996,"jasmine/genie.py",7883,0,"7",python,content
+211,542998,"jasmine/genie.py",7884,0,"",python,selection_keyboard
+212,545050,"TERMINAL",0,0,"sh slurm/jobs/mihir/horeka/breakout/sample_maskgit.sbatch",,terminal_output
+213,545231,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+214,545329,"TERMINAL",0,0,"Sampling from checkpoint: /hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/checkpoints/breakout/dyn-gt-actions/train_dyn_default_gt_actions_breakout_longer/3519698\r\n",,terminal_output
+215,545463,"TERMINAL",0,0,"GpuFreq=control_disabled\r\n",,terminal_output
+216,548159,"jasmine/genie.py",0,0,"",python,tab
+217,548160,"jasmine/genie.py",7841,0,"",python,selection_mouse
+218,548757,"jasmine/genie.py",8167,0,"",python,selection_command
+219,553802,"TERMINAL",0,0,"/hkfs/home/project/hk-project-p0023960/tum_cte0515/Projects/jasmine/.venv/lib64/python3.12/site-packages/orbax/checkpoint/_src/serialization/type_handlers.py:1269: UserWarning: Sharding info not provided when restoring. Populating sharding info from sharding file. Please note restoration time will be slightly increased due to reading from file. Note also that this option is unsafe when restoring on a different topology than the checkpoint was saved with.\r\n  warnings.warn(\r\n",,terminal_output
+220,555890,"jasmine/genie.py",8269,0,"",python,selection_mouse
+221,556022,"jasmine/genie.py",8264,11,"noise_level",python,selection_mouse
+222,583516,"TERMINAL",0,0,"Per-frame SSIM:\r\n [0.30596483 0.414882   0.5163348  0.51084375 0.5155197  0.5155652\r\n 0.51382315 0.4914374  0.45811233 0.4443463  0.44947392 0.4618884 ]\r\nPer-frame PSNR:\r\n [18.745987 18.47467  18.217417 18.06475  18.032337 17.890877 17.707855\r\n 17.542704 17.384125 17.292942 17.300459 17.316257]\r\nSSIM: 0.4665159583091736\r\nPSNR: 17.83086585998535\r\n",,terminal_output
+223,584098,"TERMINAL",0,0,"W0925 13:59:41.451424 2009277 pjrt_client.cc:1469] WatchJobStateAsync failed for task goo.gle/debugproto   job_name: ""jax_worker"": UNAVAILABLE: failed to connect to all addresses; last error: UNKNOWN: ipv4:10.0.1.177:63542: Failed to connect to remote host: Connection refused\r\nAdditional GRPC error information from remote target coordination_service while calling /tensorflow.CoordinationService/WatchJobState:\r\n:UNKNOWN:Error received from peer  {grpc_message:""failed to connect to all addresses; last error: UNKNOWN: ipv4:10.0.1.177:63542: Failed to connect to remote host: Connection refused"", grpc_status:14}\r\n",,terminal_output
+224,584621,"TERMINAL",0,0,"]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+225,759719,"jasmine/genie.py",13893,0,"",python,selection_mouse
+226,814183,"jasmine/genie.py",0,0,"",python,tab
+227,814184,"jasmine/genie.py",13869,0,"",python,selection_mouse
+228,823620,"jasmine/genie.py",14116,0,"",python,selection_mouse
+229,893246,"TERMINAL",0,0,"q",,terminal_output
+230,893311,"TERMINAL",0,0,"u",,terminal_output
+231,893434,"TERMINAL",0,0,"e",,terminal_output
+232,893487,"TERMINAL",0,0,"u",,terminal_output
+233,893627,"TERMINAL",0,0,"e",,terminal_output
+234,893813,"TERMINAL",0,0,"\r\n[?2004l\r[?1049h(B[?7hEvery 1.0s: squeue --mehkn0801.localdomain: Thu Sep 25 14:04:51 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3520566 accelerat interact tum_cte0  R14:28\t  1 hkn0801",,terminal_output
+235,894871,"TERMINAL",0,0,"29",,terminal_output
+236,894980,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn0801:~/Projects/jasmine[?2004h(jasmine) [tum_cte0515@hkn0801 jasmine]$ ",,terminal_output
+237,900148,"jasmine/genie.py",0,0,"",python,tab
+238,900149,"jasmine/genie.py",14080,0,"",python,selection_mouse
+239,900647,"jasmine/genie.py",14079,0,"",python,selection_command
+240,900965,"jasmine/genie.py",14115,0,"",python,selection_command
+241,901139,"jasmine/genie.py",14165,0,"",python,selection_command
+242,901286,"jasmine/genie.py",14227,0,"",python,selection_command
+243,901585,"jasmine/genie.py",14316,0,"",python,selection_command
+244,901908,"jasmine/genie.py",14373,0,"",python,selection_command
+245,902094,"jasmine/genie.py",14435,0,"",python,selection_command
+246,902232,"jasmine/genie.py",14461,0,"",python,selection_command
+247,902395,"jasmine/genie.py",14508,0,"",python,selection_command
+248,902561,"jasmine/genie.py",14559,0,"",python,selection_command
+249,902711,"jasmine/genie.py",14621,0,"",python,selection_command
+250,902867,"jasmine/genie.py",14685,0,"",python,selection_command
+251,903018,"jasmine/genie.py",14747,0,"",python,selection_command
+252,903161,"jasmine/genie.py",14777,0,"",python,selection_command
+253,903344,"jasmine/genie.py",14779,0,"",python,selection_command
+254,903485,"jasmine/genie.py",14812,0,"",python,selection_command
+255,903626,"jasmine/genie.py",14874,0,"",python,selection_command
+256,903785,"jasmine/genie.py",14954,0,"",python,selection_command
+257,904314,"jasmine/genie.py",15012,0,"",python,selection_command
+258,904340,"jasmine/genie.py",15026,0,"",python,selection_command
+259,904414,"jasmine/genie.py",15053,0,"",python,selection_command
+260,904645,"jasmine/genie.py",15115,0,"",python,selection_command
+261,904840,"jasmine/genie.py",15160,0,"",python,selection_command
+262,905342,"jasmine/genie.py",15174,0,"",python,selection_command
+263,905374,"jasmine/genie.py",15236,0,"",python,selection_command
+264,905440,"jasmine/genie.py",15313,0,"",python,selection_command
+265,905488,"jasmine/genie.py",15397,0,"",python,selection_command
+266,905520,"jasmine/genie.py",15471,0,"",python,selection_command
+267,905521,"jasmine/genie.py",15547,0,"",python,selection_command

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-9ab67804-e8b6-44ba-9ee4-ddec1e42461f1757968391960-2025_09_15-22.33.57.229/source.csv

@@ -0,0 +1,37 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,1209,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"10:33:57 PM [info] Activating crowd-code\n10:33:57 PM [info] Recording started\n10:33:57 PM [info] Initializing git provider using file system watchers...\n10:33:57 PM [info] Git repository found\n10:33:57 PM [info] Git provider initialized successfully\n10:33:57 PM [info] Initial git state: [object Object]\n",Log,tab
+3,2243,"TERMINAL",0,0,"bash",,terminal_focus
+4,10241,"TERMINAL",0,0,"git status",,terminal_command
+5,10311,"TERMINAL",0,0,"]633;COn branch val-loss\r\nYour branch is up to date with 'origin/val-loss'.\r\n\r\nChanges not staged for commit:\r\n  (use ""git add <file>..."" to update what will be committed)\r\n  (use ""git restore <file>..."" to discard changes in working directory)\r\n\tmodified:   models/dynamics.py\r\n\r\nUntracked files:\r\n  (use ""git add <file>..."" to include in what will be committed)\r\n\tdata/\r\n\tdiff.diff\r\n\tkiller.sh\r\n\tkiller_partition.sh\r\n\tlog.log\r\n\tlogs/\r\n\toverfit_dir.zip\r\n\trequirements-franz.txt\r\n\tsamples/\r\n\tscripts_cremers/\r\n\tslurm/\r\n\tutils/visualizer.py\r\n\r\nno changes added to commit (use ""git add"" and/or ""git commit -a"")\r\n]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+6,14147,"TERMINAL",0,0,"git diff",,terminal_command
+7,14189,"TERMINAL",0,0,"]633;C[?1h=\rdiff --git a/models/dynamics.py b/models/dynamics.py\r\nindex 74fde10..1ad90cf 100644\r\n--- a/models/dynamics.py\r\n+++ b/models/dynamics.py\r\n@@ -72,9 +72,14 @@ class DynamicsMaskGIT(nnx.Module):\r\n         )\r\n \r\n     def __call__(\r\n-        self, batch: Dict[str, jax.Array], training: bool = True, pred_full_frame: bool = False,\r\n+        self,\r\n+        batch: Dict[str, jax.Array],\r\n+        training: bool = True,\r\n+        pred_full_frame: bool = False,\r\n     ) -> tuple[jax.Array, jax.Array | None]:\r\n-        assert not (training and pred_full_frame), ""Cannot evaluate full frame prediction during training.""\r\n+        assert not (\r\n+            training and pred_full_frame\r\n+        ), ""Cannot evaluate full frame prediction during training.""\r\n         # --- Mask videos ---\r\n         video_tokens_BTN = batch[""video_tokens""]\r\n         latent_actions_BTm11L = batch[""latent_actions""]\r\n@@ -170,9 +175,14 @@ class DynamicsCausal(nnx.Module):\r\n         )\r\n \r\n     def __call__(\r\n-        self, batch: Dict[str, jax.Array], training: bool = True, pred_full_frame: bool = False,\r\n+        self,\r\n+        batch: Dict[str, jax.Array],\r\n+        training: bool = True,\r\n+        pred_full_frame: bool = False,\r\n     ) -> tuple[jax.Array, jax.Array | None]:\r\n:",,terminal_output
+8,15345,"TERMINAL",0,0,"",,terminal_command
+9,16216,"TERMINAL",0,0,"\r-        assert not (training and pred_full_frame), ""Cannot evaluate full frame prediction during training.""\r\n:",,terminal_output
+10,17045,"TERMINAL",0,0,"\r+        assert not (\r\n:\r+            training and pred_full_frame\r\n:\r+        ), ""Cannot evaluate full frame prediction during training.""\r\n:\r         video_tokens_BTN = batch[""video_tokens""]\r\n:\r         latent_actions_BTm11L = batch[""latent_actions""]\r\n:\r         if pred_full_frame:\r\n:\r@@ -184,16 +194,31 @@ class DynamicsCausal(nnx.Module):\r\n:\r             def _pred_full_frame(carry, step_n):\r\n:\r                 video_tokens_BTN, final_logits_BTNV = carry\r\n:\r                 # We need to reconstruct submodules inside scan body to prevent trace context mismatches\r\n:\r-                patch_embed = nnx.Embed(self.num_latents, self.model_dim, rngs=nnx.Rngs(0))\r\n:\r+                patch_embed = nnx.Embed(\r\n:\r+                    self.num_latents, self.model_dim, rngs=nnx.Rngs(0)\r\n:",,terminal_output
+11,17400,"TERMINAL",0,0,"\r+                )\r\n:\r                 nnx.update(patch_embed, patch_embed_state)\r\n:\r                 action_up = nnx.Linear(\r\n:\r-                    self.latent_action_dim, self.model_dim, param_dtype=self.param_dtype, dtype=self.dtype, rngs=nnx.Rn :\rgs(0)\r\n:\r+                    self.latent_action_dim,\r\n:\r+                    self.model_dim,\r\n:\r+                    param_dtype=self.param_dtype,\r\n:\r+                    dtype=self.dtype,\r\n:\r+                    rngs=nnx.Rngs(0),\r\n:\r                 )\r\n:\r                 nnx.update(action_up, action_up_state)\r\n:",,terminal_output
+12,17643,"TERMINAL",0,0,"\r                 transformer = Transformer(\r\n:\r-                    self.model_dim, self.model_dim, self.ffn_dim, self.num_latents, self.num_blocks, self.num_heads,\r\n:\r-                    self.dropout, self.param_dtype, self.dtype, use_flash_attention=self.use_flash_attention,\r\n:\r-                    decode=self.decode, rngs=nnx.Rngs(0)\r\n:\r+                    self.model_dim,\r\n:\r+                    self.model_dim,\r\n:\r+                    self.ffn_dim,\r\n:\r+                    self.num_latents,\r\n:",,terminal_output
+13,17762,"TERMINAL",0,0,"\r+                    self.num_blocks,\r\n:\r+                    self.num_heads,\r\n:",,terminal_output
+14,17883,"TERMINAL",0,0,"\r+                    self.dropout,\r\n:\r+                    self.param_dtype,\r\n:",,terminal_output
+15,18897,"TERMINAL",0,0,"\r+                    self.dtype,\r\n:",,terminal_output
+16,19279,"TERMINAL",0,0,"\r+                    use_flash_attention=self.use_flash_attention,\r\n:",,terminal_output
+17,19756,"TERMINAL",0,0,"\r+                    decode=self.decode,\r\n:\r+                    rngs=nnx.Rngs(0),\r\n:",,terminal_output
+18,19862,"TERMINAL",0,0,"\r                 )\r\n:\r                 nnx.update(transformer, transformer_state)\r\n:\r \r\n:",,terminal_output
+19,19966,"TERMINAL",0,0,"\r@@ -207,7 +232,9 @@ class DynamicsCausal(nnx.Module):\r\n:\r                 )\r\n:\r                 step_logits_BTNp1V = transformer(vid_embed_BTNp1M)\r\n:",,terminal_output
+20,20114,"TERMINAL",0,0,"\r                 step_logits_BV = step_logits_BTNp1V[:, -1, step_n, :]\r\n:\r-                final_logits_BTNV = final_logits_BTNV.at[:, -1, step_n].set(step_logits_BV)\r\n:\r+                final_logits_BTNV = final_logits_BTNV.at[:, -1, step_n].set(\r\n:\r+                    step_logits_BV\r\n:\r+                )\r\n:\r                 sampled_token_idxs_B = jnp.argmax(step_logits_BV, axis=-1)\r\n:",,terminal_output
+21,20303,"TERMINAL",0,0,"\r                 video_tokens_BTN = video_tokens_BTN.at[:, -1, step_n].set(\r\n:\r                     sampled_token_idxs_B\r\n:\r@@ -216,10 +243,11 @@ class DynamicsCausal(nnx.Module):\r\n:\r \r\n:\r             (_, final_logits_BTNV), _ = jax.lax.scan(\r\n:",,terminal_output
+22,20421,"TERMINAL",0,0,"\r                 _pred_full_frame,\r\n:\r-                (video_tokens_BTN, jnp.zeros((\r\n:",,terminal_output
+23,20492,"TERMINAL",0,0,"\r-                    **video_tokens_BTN.shape,\r\n:\r-                    self.num_latents))),\r\n:\r-                jnp.arange(video_tokens_BTN.shape[2])\r\n:\r+                (\r\n:",,terminal_output
+24,20800,"TERMINAL",0,0,"\r+                    video_tokens_BTN,\r\n:\r+                    jnp.zeros((*video_tokens_BTN.shape, self.num_latents)),\r\n:\r+                ),\r\n:\r+                jnp.arange(video_tokens_BTN.shape[2]),\r\n:\r             )\r\n:\r             mask_out = jnp.zeros_like(video_tokens_BTN)\r\n:\r             mask_out = mask_out.at[:, -1].set(True)\r\n:\r\r(END)\r\r(END)\r\r(END)\r\r(END)\r\r(END)",,terminal_output
+25,21746,"TERMINAL",0,0,"\r[?1l>]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+26,22679,"TERMINAL",0,0,"",,terminal_command
+27,31747,"TERMINAL",0,0,"git commit -am ""run pre-commit""",,terminal_command
+28,31829,"TERMINAL",0,0,"]633;C",,terminal_output
+29,34352,"TERMINAL",0,0,"[INFO] Installing environment for https://github.com/psf/black.\r\n[INFO] Once installed this environment will be reused.\r\n[INFO] This may take a few minutes...\r\n",,terminal_output
+30,50819,"TERMINAL",0,0,"black....................................................................",,terminal_output
+31,50994,"TERMINAL",0,0,"Passed\r\n",,terminal_output
+32,51145,"TERMINAL",0,0,"[val-loss 263a0c0] run pre-commit\r\n 1 file changed, 42 insertions(+), 14 deletions(-)\r\n]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output
+33,60713,"TERMINAL",0,0,"git push",,terminal_command
+34,60808,"TERMINAL",0,0,"]633;C",,terminal_output
+35,62167,"TERMINAL",0,0,"Enumerating objects: 7, done.\r\nCounting objects:  14% (1/7)\rCounting objects:  28% (2/7)\rCounting objects:  42% (3/7)\rCounting objects:  57% (4/7)\rCounting objects:  71% (5/7)\rCounting objects:  85% (6/7)\rCounting objects: 100% (7/7)\rCounting objects: 100% (7/7), done.\r\nDelta compression using up to 152 threads\r\nCompressing objects:  25% (1/4)\rCompressing objects:  50% (2/4)\rCompressing objects:  75% (3/4)\rCompressing objects: 100% (4/4)\rCompressing objects: 100% (4/4), done.\r\nWriting objects:  25% (1/4)\rWriting objects:  50% (2/4)\rWriting objects:  75% (3/4)\rWriting objects: 100% (4/4)\rWriting objects: 100% (4/4), 714 bytes | 714.00 KiB/s, done.\r\nTotal 4 (delta 2), reused 0 (delta 0), pack-reused 0\r\nremote: Resolving deltas:   0% (0/2)\rremote: Resolving deltas:  50% (1/2)\rremote: Resolving deltas: 100% (2/2)\rremote: Resolving deltas: 100% (2/2), completed with 2 local objects.\r\n",,terminal_output
+36,62453,"TERMINAL",0,0,"To github.com:p-doom/jasmine.git\r\n   a9f9ec1..263a0c0  val-loss -> val-loss\r\n",,terminal_output
+37,62482,"TERMINAL",0,0,"]0;tum_cte0515@hkn1993:~/Projects/jasmine",,terminal_output

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-9eb2e164-5989-4db7-8f31-6e8db1a38df41757236520211-2025_09_07-11.16.00.62/source.csv

@@ -0,0 +1,5 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,2178,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"11:16:00 AM [info] Activating crowd-code\n11:16:00 AM [info] Recording started\n11:16:00 AM [info] Initializing git provider using file system watchers...\n11:16:00 AM [info] Git repository found\n11:16:00 AM [info] Git provider initialized successfully\n11:16:01 AM [info] Initial git state: [object Object]\n",Log,tab
+3,108194,"TERMINAL",0,0,"bash",,terminal_focus
+4,215106,"TERMINAL",0,0,"bash",,terminal_focus
+5,218728,"TERMINAL",0,0,"bash",,terminal_focus

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-e9c5a28e-55ca-497d-97b5-e0c37d2af2781751878142668-2025_07_07-10.49.22.270/source.csv

@@ -0,0 +1,213 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+2,764,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"10:49:22 AM [info] Activating crowd-code\n10:49:22 AM [info] Recording started\n10:49:22 AM [info] Initializing git provider using file system watchers...\n10:49:22 AM [info] Git repository found\n10:49:22 AM [info] Git provider initialized successfully\n10:49:22 AM [info] Initial git state: [object Object]\n",Log,tab
+3,4308,"TERMINAL",0,0,"/bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt",,terminal_command
+4,4359,"TERMINAL",0,0,"]633;E;2025-07-07 10:49:26 /bin/python3 /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/printEnvVariablesToFile.py /hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash/envVars.txt;bf94d117-cd0c-449b-b408-59a02a05b60e]633;C]0;tum_cte0515@hkn1991:/hkfs/home/project/hk-project-p0023960/tum_cte0515/.cursor-server/extensions/ms-python.python-2024.12.3-linux-x64/python_files/deactivate/bash]633;D;0",,terminal_output
+5,7832,"TERMINAL",0,0,"git branch",,terminal_command
+6,7884,"TERMINAL",0,0,"]633;E;2025-07-07 10:49:30 git branch;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C[?1h=\r  add-wandb-name-and-tags\r\n  convert-to-jax-array-in-iter\r\n  dont-let-tf-see-gpu\r\n  feat/explicit-image-dims\r\n  fix-sampling\r\n  main\r\n  preprocess_video\r\n  revised-dataloader\r\n* runner\r\n  tmp\r\n",,terminal_output
+7,7960,"TERMINAL",0,0,"\r[?1l>]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+8,77190,"TERMINAL",0,0,"git status",,terminal_command
+9,77237,"TERMINAL",0,0,"]633;E;2025-07-07 10:50:39 git status;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+10,77308,"TERMINAL",0,0,"On branch runner\r\nAll conflicts fixed but you are still merging.\r\n  (use ""git commit"" to conclude merge)\r\n\r\nChanges to be committed:\r\n\tmodified:   train_dynamics.py\r\n\tmodified:   train_lam.py\r\n\tmodified:   train_tokenizer.py\r\n\r\nChanges not staged for commit:\r\n  (use ""git add <file>..."" to update what will be committed)\r\n  (use ""git restore <file>..."" to discard changes in working directory)\r\n\tmodified:   train_lam.py\r\n\tmodified:   train_tokenizer.py\r\n\r\nUntracked files:\r\n  (use ""git add <file>..."" to include in what will be committed)\r\n\tdata_tfrecord_duplicated/\r\n\tdata_tfrecords/\r\n\tlogs/\r\n\tread_tf_record.py\r\n\trequirements-franz.txt\r\n\tscripts_cremers/\r\n\tscripts_horeka/\r\n\tslurm-3309772.out\r\n\tslurm/\r\n\tutils/visualizer.py\r\n\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+11,85704,"TERMINAL",0,0,"git diff train_lam.py",,terminal_command
+12,85768,"TERMINAL",0,0,"]633;E;2025-07-07 10:50:47 git diff train_lam.py;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C[?1h=\rdiff --git a/train_lam.py b/train_lam.py\r\nindex 858990e..540a464 100644\r\n--- a/train_lam.py\r\n+++ b/train_lam.py\r\n@@ -59,7 +59,6 @@ class Args:\r\n     log_interval: int = 5\r\n     log_image_interval: int = 250\r\n     ckpt_dir: str = """"\r\n-    tmp_ckpt_dir: str = ""/tmp/checkpoints/""\r\n     log_checkpoint_interval: int = 10000\r\n \r\n \r\n\r[?1l>]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+13,97852,"TERMINAL",0,0,"git commit -am ""removed tmp""",,terminal_command
+14,97890,"TERMINAL",0,0,"]633;E;2025-07-07 10:51:00 git commit -am ""removed tmp"";f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+15,98308,"TERMINAL",0,0,"[runner 316eae6] removed tmp\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+16,109638,"TERMINAL",0,0,"git checkout revised-dataloader",,terminal_command
+17,109689,"TERMINAL",0,0,"]633;E;2025-07-07 10:51:11 git checkout revised-dataloader;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+18,110012,"TERMINAL",0,0,"Switched to branch 'revised-dataloader'\r\nYour branch is up to date with 'origin/revised-dataloader'.\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+19,110376,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"Switched from branch 'runner' to 'revised-dataloader'",Log,git_branch_checkout
+20,110484,"extension-output-pdoom-org.crowd-code-#1-crowd-code",304,0,"10:51:12 AM [info] Branch checkout detected: runner -> revised-dataloader\n10:51:12 AM [info] Recording git checkout: Switched from branch 'runner' to 'revised-dataloader'\n10:51:12 AM [info] Resetting file cache due to branch checkout\n",Log,content
+21,112104,"TERMINAL",0,0,"git status",,terminal_command
+22,112207,"TERMINAL",0,0,"]633;E;2025-07-07 10:51:14 git status;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;COn branch revised-dataloader\r\nYour branch is up to date with 'origin/revised-dataloader'.\r\n\r\nUntracked files:\r\n  (use ""git add <file>..."" to include in what will be committed)\r\n\tdata_tfrecord_duplicated/\r\n\tdata_tfrecords/\r\n\tlogs/\r\n\tread_tf_record.py\r\n\trequirements-franz.txt\r\n\tscripts_cremers/\r\n\tscripts_horeka/\r\n\tslurm-3309772.out\r\n\tslurm/\r\n\tutils/visualizer.py\r\n\r\nnothing added to commit but untracked files present (use ""git add"" to track)\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+23,128400,"TERMINAL",0,0,"git pull",,terminal_command
+24,128448,"TERMINAL",0,0,"]633;E;2025-07-07 10:51:30 git pull;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+25,130043,"TERMINAL",0,0,"remote: Enumerating objects: 60, done.\r\nremote: Counting objects:   1% (1/54)\rremote: Counting objects:   3% (2/54)\rremote: Counting objects:   5% (3/54)\rremote: Counting objects:   7% (4/54)\rremote: Counting objects:   9% (5/54)\rremote: Counting objects:  11% (6/54)\rremote: Counting objects:  12% (7/54)\rremote: Counting objects:  14% (8/54)\rremote: Counting objects:  16% (9/54)\rremote: Counting objects:  18% (10/54)\rremote: Counting objects:  20% (11/54)\rremote: Counting objects:  22% (12/54)\rremote: Counting objects:  24% (13/54)\rremote: Counting objects:  25% (14/54)\rremote: Counting objects:  27% (15/54)\rremote: Counting objects:  29% (16/54)\rremote: Counting objects:  31% (17/54)\rremote: Counting objects:  33% (18/54)\rremote: Counting objects:  35% (19/54)\rremote: Counting objects:  37% (20/54)\rremote: Counting objects:  38% (21/54)\rremote: Counting objects:  40% (22/54)\rremote: Counting objects:  42% (23/54)\rremote: Counting objects:  44% (24/54)\rremote: Counting objects:  46% (25/54)\rremote: Counting objects:  48% (26/54)\rremote: Counting objects:  50% (27/54)\rremote: Counting objects:  51% (28/54)\rremote: Counting objects:  53% (29/54)\rremote: Counting objects:  55% (30/54)\rremote: Counting objects:  57% (31/54)\rremote: Counting objects:  59% (32/54)\rremote: Counting objects:  61% (33/54)\rremote: Counting objects:  62% (34/54)\rremote: Counting objects:  64% (35/54)\rremote: Counting objects:  66% (36/54)\rremote: Counting objects:  68% (37/54)\rremote: Counting objects:  70% (38/54)\rremote: Counting objects:  72% (39/54)\rremote: Counting objects:  74% (40/54)\rremote: Counting objects:  75% (41/54)\rremote: Counting objects:  77% (42/54)\rremote: Counting objects:  79% (43/54)\rremote: Counting objects:  81% (44/54)\rremote: Counting objects:  83% (45/54)\rremote: Counting objects:  85% (46/54)\rremote: Counting objects:  87% (47/54)\rremote: Counting objects:  88% (48/54)\rremote: Counting objects:  90% (49/54)\rremote: Counting objects:  92% (50/54)\rremote: Counting objects:  94% (51/54)\rremote: Counting objects:  96% (52/54)\rremote: Counting objects:  98% (53/54)\rremote: Counting objects: 100% (54/54)\rremote: Counting objects: 100% (54/54), done.\r\nremote: Compressing objects:   5% (1/20)\rremote: Compressing objects:  10% (2/20)\rremote: Compressing objects:  15% (3/20)\rremote: Compressing objects:  20% (4/20)\rremote: Compressing objects:  25% (5/20)\rremote: Compressing objects:  30% (6/20)\rremote: Compressing objects:  35% (7/20)\rremote: Compressing objects:  40% (8/20)\rremote: Compressing objects:  45% (9/20)\rremote: Compressing objects:  50% (10/20)\rremote: Compressing objects:  55% (11/20)\rremote: Compressing objects:  60% (12/20)\rremote: Compressing objects:  65% (13/20)\rremote: Compressing objects:  70% (14/20)\rremote: Compressing objects:  75% (15/20)\rremote: Compressing objects:  80% (16/20)\rremote: Compressing objects:  85% (17/20)\rremote: Compressing objects:  90% (18/20)\rremote: Compressing objects:  95% (19/20)\rremote: Compressing objects: 100% (20/20)\rremote: Compressing objects: 100% (20/20), done.\r\nremote: Total 36 (delta 26), reused 22 (delta 16), pack-reused 0 (from 0)\r\n",,terminal_output
+26,130161,"TERMINAL",0,0,"Unpacking objects:   2% (1/36)\rUnpacking objects:   5% (2/36)\rUnpacking objects:   8% (3/36)\rUnpacking objects:  11% (4/36)\rUnpacking objects:  13% (5/36)\rUnpacking objects:  16% (6/36)\rUnpacking objects:  19% (7/36)\r",,terminal_output
+27,130215,"TERMINAL",0,0,"Unpacking objects:  22% (8/36)\rUnpacking objects:  25% (9/36)\r",,terminal_output
+28,130280,"TERMINAL",0,0,"Unpacking objects:  27% (10/36)\rUnpacking objects:  30% (11/36)\r",,terminal_output
+29,130404,"TERMINAL",0,0,"Unpacking objects:  33% (12/36)\rUnpacking objects:  36% (13/36)\rUnpacking objects:  38% (14/36)\rUnpacking objects:  41% (15/36)\rUnpacking objects:  44% (16/36)\rUnpacking objects:  47% (17/36)\rUnpacking objects:  50% (18/36)\rUnpacking objects:  52% (19/36)\rUnpacking objects:  55% (20/36)\rUnpacking objects:  58% (21/36)\rUnpacking objects:  61% (22/36)\rUnpacking objects:  63% (23/36)\rUnpacking objects:  66% (24/36)\r",,terminal_output
+30,130473,"TERMINAL",0,0,"Unpacking objects:  69% (25/36)\rUnpacking objects:  72% (26/36)\rUnpacking objects:  75% (27/36)\rUnpacking objects:  77% (28/36)\rUnpacking objects:  80% (29/36)\rUnpacking objects:  83% (30/36)\r",,terminal_output
+31,130587,"TERMINAL",0,0,"Unpacking objects:  86% (31/36)\rUnpacking objects:  88% (32/36)\rUnpacking objects:  91% (33/36)\rUnpacking objects:  94% (34/36)\rUnpacking objects:  97% (35/36)\rUnpacking objects: 100% (36/36)\rUnpacking objects: 100% (36/36), 6.91 KiB | 14.00 KiB/s, done.\r\n",,terminal_output
+32,130774,"TERMINAL",0,0,"From github.com:p-doom/jafar\r\n * [new branch]      correct-batched-sampling -> origin/correct-batched-sampling\r\n   4ec9ebe..9edd0c1  dynamics-lam-co-training -> origin/dynamics-lam-co-training\r\n   32b3f04..3176718  feat/restore_train_state -> origin/feat/restore_train_state\r\n   ae9451f..6e623c6  fix-sampling             -> origin/fix-sampling\r\n   c8dd7ea..9fb362e  main                     -> origin/main\r\n * [new branch]      make-warmup-default      -> origin/make-warmup-default\r\n",,terminal_output
+33,130837,"TERMINAL",0,0,"Already up to date.\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+34,133144,"TERMINAL",0,0,"git status",,terminal_command
+35,133186,"TERMINAL",0,0,"]633;E;2025-07-07 10:51:35 git status;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;COn branch revised-dataloader\r\nYour branch is up to date with 'origin/revised-dataloader'.\r\n\r\nUntracked files:\r\n  (use ""git add <file>..."" to include in what will be committed)\r\n\tdata_tfrecord_duplicated/\r\n\tdata_tfrecords/\r\n\tlogs/\r\n\tread_tf_record.py\r\n\trequirements-franz.txt\r\n\tscripts_cremers/\r\n\tscripts_horeka/\r\n\tslurm-3309772.out\r\n\tslurm/\r\n\tutils/visualizer.py\r\n\r\nnothing added to commit but untracked files present (use ""git add"" to track)\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+36,148140,"utils/dataloader.py",0,0,"import functools\nimport jax\n\nimport tensorflow as tf\n\n# reserve GPU memory for JAX only if tensorflow is built with GPU support\ntf.config.experimental.set_visible_devices([], ""GPU"")\n\n\n# --- TensorFlow function for processing: slicing, normalization ---\ndef _tf_process_episode(episode_tensor, seq_len, image_h, image_w, image_c):\n    """"""\n    Processes a raw episode tensor in TensorFlow.\n    Takes a full episode, extracts a random sequence, and normalizes it.\n    Args:\n        episode_tensor: A TensorFlow tensor representing a full video episode.\n                        Expected shape: (dynamic_length, image_h, image_w, image_c)\n                        Expected dtype: e.g., tf.uint8 (raw pixel values)\n        seq_len: The desired length of the sub-sequence to extract.\n        image_h: The height of each frame.\n        image_w: The width of each frame.\n        image_c: The number of channels in each frame.\n    Returns:\n        A TensorFlow tensor representing the processed video sequence.\n        Shape: (seq_len, image_h, image_w, image_c)\n        Dtype: tf.float32 (normalized pixel values)\n    """"""\n    current_episode_len = tf.shape(episode_tensor)[0]\n\n    max_start_idx = current_episode_len - seq_len\n\n    start_idx = tf.random.uniform(\n        shape=(), minval=0, maxval=max_start_idx + 1, dtype=tf.int32\n    )\n\n    seq = episode_tensor[start_idx : start_idx + seq_len]\n\n    seq = tf.cast(seq, tf.float32) / 255.0\n\n    # Ensure the final shape is statically known for batching.\n    # tf.reshape is robust, but tf.ensure_shape or set_shape can also be used if confident.\n    processed_sequence = tf.reshape(seq, [seq_len, image_h, image_w, image_c])\n\n    return processed_sequence\n\n\ndef _parse_tfrecord_fn(example_proto, image_h, image_w, image_c):\n    feature_description = {\n        ""height"": tf.io.FixedLenFeature([], tf.int64),\n        ""width"": tf.io.FixedLenFeature([], tf.int64),\n        ""channels"": tf.io.FixedLenFeature([], tf.int64),\n        ""sequence_length"": tf.io.FixedLenFeature([], tf.int64),\n        ""raw_video"": tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example_proto, feature_description)\n\n    video_shape = (example[""sequence_length""], image_h, image_w, image_c)\n\n    episode_tensor = tf.io.decode_raw(example[""raw_video""], out_type=tf.uint8)\n    episode_tensor = tf.reshape(episode_tensor, video_shape)\n\n    episode_tensor = tf.ensure_shape(episode_tensor, [None, image_h, image_w, image_c])\n    return episode_tensor\n\n\ndef _create_processed_dataset_from_file(file_path, image_h, image_w, image_c, seq_len, num_parallel_calls):\n    """"""Creates a fully processed dataset from a single TFRecord file.""""""\n    dataset = tf.data.TFRecordDataset([file_path])\n    \n    parse_fn = functools.partial(\n        _parse_tfrecord_fn, image_h=image_h, image_w=image_w, image_c=image_c\n    )\n    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    tf_process_fn = functools.partial(\n        _tf_process_episode,\n        seq_len=seq_len,\n        image_h=image_h,\n        image_w=image_w,\n        image_c=image_c,\n    )\n    dataset = dataset.map(tf_process_fn, num_parallel_calls=num_parallel_calls)\n    \n    return dataset\n\n\ndef get_dataloader(\n    tfrecord_paths: list[str],\n    seq_len: int,\n    global_batch_size: int,\n    image_h: int,\n    image_w: int,\n    image_c: int,\n    shuffle_buffer_size: int = 1000,\n    num_parallel_calls: int = tf.data.AUTOTUNE,\n    seed: int = 42,\n    cycle_length: int = 4,\n    block_length: int = 1,\n):\n    """"""\n    Creates a tf.data.Dataset pipeline from TFRecord files.\n    """"""\n    if not tfrecord_paths:\n        raise ValueError(""tfrecord_paths list cannot be empty."")\n\n    process_id = jax.process_index()\n    num_processes = jax.process_count()\n\n    assert (\n        global_batch_size % num_processes == 0\n    ), f""Global batch size {global_batch_size} \\n        must be divisible by the number of JAX processes {num_processes} for proper sharding.""\n    per_process_batch_size = global_batch_size // num_processes\n\n    def dataset_fn(file_path):\n        return _create_processed_dataset_from_file(\n            file_path, image_h, image_w, image_c, seq_len, num_parallel_calls\n        )\n    \n    dataset = tf.data.Dataset.from_tensor_slices(tfrecord_paths)\n    dataset = dataset.shard(num_shards=num_processes, index=process_id)\n    \n    dataset = dataset.interleave(\n        dataset_fn,\n        cycle_length=cycle_length,\n        block_length=block_length,\n        num_parallel_calls=num_parallel_calls,\n        deterministic=False\n    )\n    \n    if shuffle_buffer_size > 0:\n        dataset = dataset.shuffle(\n            buffer_size=shuffle_buffer_size, seed=seed, reshuffle_each_iteration=True\n        )\n\n    dataset = dataset.repeat(None)\n    dataset = dataset.batch(per_process_batch_size, drop_remainder=True)\n    dataset = dataset.prefetch(tf.data.AUTOTUNE)\n\n    return dataset.as_numpy_iterator()\n",python,tab
+37,153941,"utils/dataloader.py",3046,0,"",python,selection_mouse
+38,153957,"utils/dataloader.py",3045,0,"",python,selection_command
+39,167620,"TERMINAL",0,0,"git checkout runner",,terminal_command
+40,167670,"TERMINAL",0,0,"]633;E;2025-07-07 10:52:09 git checkout runner;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+41,167734,"TERMINAL",0,0,"Switched to branch 'runner'\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+42,169889,"utils/dataloader.py",0,0,"",python,tab
+43,170107,"utils/dataloader.py",253,3946,"def _tf_process_episode(episode_tensor, seq_len, image_h, image_w, image_c, seed):\n    """"""\n    Processes a raw episode tensor in TensorFlow.\n    Takes a full episode, extracts a random sequence, and normalizes it.\n    Args:\n        episode_tensor: A TensorFlow tensor representing a full video episode.\n                        Expected shape: (dynamic_length, image_h, image_w, image_c)\n                        Expected dtype: e.g., tf.uint8 (raw pixel values)\n        seq_len: The desired length of the sub-sequence to extract.\n        image_h: The height of each frame.\n        image_w: The width of each frame.\n        image_c: The number of channels in each frame.\n        seed: The seed for the random number generator.\n    Returns:\n        A TensorFlow tensor representing the processed video sequence.\n        Shape: (seq_len, image_h, image_w, image_c)\n        Dtype: tf.float32 (normalized pixel values)\n    """"""\n    current_episode_len = tf.shape(episode_tensor)[0]\n\n    max_start_idx = current_episode_len - seq_len\n\n    start_idx = tf.random.uniform(\n        shape=(), minval=0, maxval=max_start_idx + 1, dtype=tf.int32, seed=seed\n    )\n\n    seq = episode_tensor[start_idx : start_idx + seq_len]\n\n    seq = tf.cast(seq, tf.float32) / 255.0\n\n    # Ensure the final shape is statically known for batching.\n    # tf.reshape is robust, but tf.ensure_shape or set_shape can also be used if confident.\n    processed_sequence = tf.reshape(seq, [seq_len, image_h, image_w, image_c])\n\n    return processed_sequence\n\n\ndef _parse_tfrecord_fn(example_proto, image_h, image_w, image_c):\n    feature_description = {\n        ""height"": tf.io.FixedLenFeature([], tf.int64),\n        ""width"": tf.io.FixedLenFeature([], tf.int64),\n        ""channels"": tf.io.FixedLenFeature([], tf.int64),\n        ""sequence_length"": tf.io.FixedLenFeature([], tf.int64),\n        ""raw_video"": tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example_proto, feature_description)\n\n    video_shape = (example[""sequence_length""], image_h, image_w, image_c)\n\n    episode_tensor = tf.io.decode_raw(example[""raw_video""], out_type=tf.uint8)\n    episode_tensor = tf.reshape(episode_tensor, video_shape)\n\n    episode_tensor = tf.ensure_shape(episode_tensor, [None, image_h, image_w, image_c])\n    return episode_tensor\n\n\ndef _create_processed_dataset_from_file(file_path, image_h, image_w, image_c, seq_len, num_parallel_calls, seed):\n    """"""Creates a fully processed dataset from a single TFRecord file.""""""\n    dataset = tf.data.TFRecordDataset([file_path])\n    \n    parse_fn = functools.partial(\n        _parse_tfrecord_fn, image_h=image_h, image_w=image_w, image_c=image_c\n    )\n    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)\n\n    tf_process_fn = functools.partial(\n        _tf_process_episode,\n        seq_len=seq_len,\n        image_h=image_h,\n        image_w=image_w,\n        image_c=image_c,\n        seed=seed,\n    )\n    dataset = dataset.map(tf_process_fn, num_parallel_calls=num_parallel_calls)\n    \n    return dataset\n\n\ndef get_dataloader(\n    tfrecord_paths: list[str],\n    seq_len: int,\n    global_batch_size: int,\n    image_h: int,\n    image_w: int,\n    image_c: int,\n    shuffle_buffer_size: int = 10,\n    num_parallel_calls: int = tf.data.AUTOTUNE,\n    seed: int = 42,\n    cycle_length: int = 4,\n    block_length: int = 1,\n):\n    """"""\n    Creates a tf.data.Dataset pipeline from TFRecord files.\n    """"""\n    if not tfrecord_paths:\n        raise ValueError(""tfrecord_paths list cannot be empty."")\n\n    process_id = jax.process_index()\n    num_processes = jax.process_count()\n\n    assert (\n        global_batch_size % num_processes == 0\n    ), f""Global batch size {global_batch_size} \\n        must be divisible by the number of JAX processes {num_processes} for proper sharding.""\n    per_process_batch_size = global_batch_size // num_processes\n\n    def dataset_fn(file_path):\n        return _create_processed_dataset_from_file(\n            file_path, image_h, image_w, image_c, seq_len, num_parallel_calls, seed\n",python,content
+44,170389,"utils/dataloader.py",0,0,"Switched from branch 'revised-dataloader' to 'runner'",python,git_branch_checkout
+45,172501,"utils/dataloader.py",3153,0,"",python,selection_mouse
+46,172512,"utils/dataloader.py",3152,0,"",python,selection_command
+47,172733,"utils/dataloader.py",3152,1,"n",python,selection_mouse
+48,172734,"utils/dataloader.py",3152,6,"n\n    ",python,selection_mouse
+49,172735,"utils/dataloader.py",3135,17,"sor)[0] >= seq_le",python,selection_mouse
+50,172735,"utils/dataloader.py",3079,73,"des(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_le",python,selection_mouse
+51,172751,"utils/dataloader.py",3153,0,"",python,selection_command
+52,172824,"utils/dataloader.py",3027,126,"isodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len",python,selection_mouse
+53,172825,"utils/dataloader.py",3021,132,"out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len",python,selection_mouse
+54,172826,"utils/dataloader.py",3018,135,"er out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len",python,selection_mouse
+55,172842,"utils/dataloader.py",3014,139,"Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len",python,selection_mouse
+56,172874,"utils/dataloader.py",3007,146,"\n    # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len",python,selection_mouse
+57,173126,"utils/dataloader.py",2932,221,"    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len",python,selection_mouse
+58,177555,"utils/dataloader.py",3210,0,"",python,selection_mouse
+59,177618,"utils/dataloader.py",3209,0,"",python,selection_command
+60,177705,"utils/dataloader.py",3209,1,")",python,selection_mouse
+61,177723,"utils/dataloader.py",3210,0,"",python,selection_command
+62,177807,"utils/dataloader.py",3207,3,"es)",python,selection_mouse
+63,177808,"utils/dataloader.py",3158,52,"\n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+64,177815,"utils/dataloader.py",3131,79,"_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+65,177818,"utils/dataloader.py",3127,83,"sode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+66,177843,"utils/dataloader.py",3124,86,"episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+67,177843,"utils/dataloader.py",3123,87,"(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+68,177910,"utils/dataloader.py",3121,89,"pe(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+69,177912,"utils/dataloader.py",3119,91,"hape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+70,177920,"utils/dataloader.py",3116,94,"f.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+71,177943,"utils/dataloader.py",3114,96," tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+72,178059,"utils/dataloader.py",3112,98,"rn tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+73,178060,"utils/dataloader.py",3111,99,"urn tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+74,178060,"utils/dataloader.py",3109,101,"eturn tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+75,178062,"utils/dataloader.py",3108,102,"return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+76,178062,"utils/dataloader.py",3060,150," filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+77,178083,"utils/dataloader.py",3059,151,"f filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+78,178145,"utils/dataloader.py",3058,152,"ef filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+79,178216,"utils/dataloader.py",3057,153,"def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+80,178360,"utils/dataloader.py",3056,154," def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+81,178361,"utils/dataloader.py",3055,155,"  def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+82,178504,"utils/dataloader.py",3010,200,"  # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+83,178505,"utils/dataloader.py",3055,155,"  def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+84,178505,"utils/dataloader.py",3009,201,"   # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+85,178827,"utils/dataloader.py",3008,202,"    # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,selection_mouse
+86,189453,"TERMINAL",0,0,"git checkout revised-dataloader",,terminal_command
+87,189506,"TERMINAL",0,0,"]633;E;2025-07-07 10:52:31 git checkout revised-dataloader;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;CSwitched to branch 'revised-dataloader'\r\nYour branch is up to date with 'origin/revised-dataloader'.\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+88,190380,"",0,0,"Switched from branch 'runner' to 'revised-dataloader'",,git_branch_checkout
+89,190846,"utils/dataloader.py",0,0,"import functools\nimport jax\n\nimport tensorflow as tf\n\n# reserve GPU memory for JAX only if tensorflow is built with GPU support\ntf.config.experimental.set_visible_devices([], ""GPU"")\n\n\n# --- TensorFlow function for processing: slicing, normalization ---\ndef _tf_process_episode(episode_tensor, seq_len, image_h, image_w, image_c, seed):\n    """"""\n    Processes a raw episode tensor in TensorFlow.\n    Takes a full episode, extracts a random sequence, and normalizes it.\n    Args:\n        episode_tensor: A TensorFlow tensor representing a full video episode.\n                        Expected shape: (dynamic_length, image_h, image_w, image_c)\n                        Expected dtype: e.g., tf.uint8 (raw pixel values)\n        seq_len: The desired length of the sub-sequence to extract.\n        image_h: The height of each frame.\n        image_w: The width of each frame.\n        image_c: The number of channels in each frame.\n        seed: The seed for the random number generator.\n    Returns:\n        A TensorFlow tensor representing the processed video sequence.\n        Shape: (seq_len, image_h, image_w, image_c)\n        Dtype: tf.float32 (normalized pixel values)\n    """"""\n    current_episode_len = tf.shape(episode_tensor)[0]\n\n    max_start_idx = current_episode_len - seq_len\n\n    start_idx = tf.random.uniform(\n        shape=(), minval=0, maxval=max_start_idx + 1, dtype=tf.int32, seed=seed\n    )\n\n    seq = episode_tensor[start_idx : start_idx + seq_len]\n\n    seq = tf.cast(seq, tf.float32) / 255.0\n\n    # Ensure the final shape is statically known for batching.\n    # tf.reshape is robust, but tf.ensure_shape or set_shape can also be used if confident.\n    processed_sequence = tf.reshape(seq, [seq_len, image_h, image_w, image_c])\n\n    return processed_sequence\n\n\ndef _parse_tfrecord_fn(example_proto, image_h, image_w, image_c):\n    feature_description = {\n        ""height"": tf.io.FixedLenFeature([], tf.int64),\n        ""width"": tf.io.FixedLenFeature([], tf.int64),\n        ""channels"": tf.io.FixedLenFeature([], tf.int64),\n        ""sequence_length"": tf.io.FixedLenFeature([], tf.int64),\n        ""raw_video"": tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example_proto, feature_description)\n\n    video_shape = (example[""sequence_length""], image_h, image_w, image_c)\n\n    episode_tensor = tf.io.decode_raw(example[""raw_video""], out_type=tf.uint8)\n    episode_tensor = tf.reshape(episode_tensor, video_shape)\n\n    episode_tensor = tf.ensure_shape(episode_tensor, [None, image_h, image_w, image_c])\n    return episode_tensor\n\n\ndef _create_processed_dataset_from_file(file_path, image_h, image_w, image_c, seq_len, num_parallel_calls, seed):\n    """"""Creates a fully processed dataset from a single TFRecord file.""""""\n    dataset = tf.data.TFRecordDataset([file_path])\n    \n    parse_fn = functools.partial(\n        _parse_tfrecord_fn, image_h=image_h, image_w=image_w, image_c=image_c\n    )\n    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)\n\n    tf_process_fn = functools.partial(\n        _tf_process_episode,\n        seq_len=seq_len,\n        image_h=image_h,\n        image_w=image_w,\n        image_c=image_c,\n        seed=seed,\n    )\n    dataset = dataset.map(tf_process_fn, num_parallel_calls=num_parallel_calls)\n    \n    return dataset\n\n\ndef get_dataloader(\n    tfrecord_paths: list[str],\n    seq_len: int,\n    global_batch_size: int,\n    image_h: int,\n    image_w: int,\n    image_c: int,\n    shuffle_buffer_size: int = 10,\n    num_parallel_calls: int = tf.data.AUTOTUNE,\n    seed: int = 42,\n    cycle_length: int = 4,\n    block_length: int = 1,\n):\n    """"""\n    Creates a tf.data.Dataset pipeline from TFRecord files.\n    """"""\n    if not tfrecord_paths:\n        raise ValueError(""tfrecord_paths list cannot be empty."")\n\n    process_id = jax.process_index()\n    num_processes = jax.process_count()\n\n    assert (\n        global_batch_size % num_processes == 0\n    ), f""Global batch size {global_batch_size} \\n        must be divisible by the number of JAX processes {num_processes} for proper sharding.""\n    per_process_batch_size = global_batch_size // num_processes\n\n    def dataset_fn(file_path):\n        return _create_processed_dataset_from_file(\n            file_path, image_h, image_w, image_c, seq_len, num_parallel_calls, seed\n        )\n    \n    dataset = tf.data.Dataset.from_tensor_slices(tfrecord_paths)\n    dataset = dataset.shard(num_shards=num_processes, index=process_id)\n    \n    dataset = dataset.interleave(\n        dataset_fn,\n        cycle_length=cycle_length,\n        block_length=block_length,\n        num_parallel_calls=num_parallel_calls,\n        deterministic=False\n    )\n    \n    if shuffle_buffer_size > 0:\n        dataset = dataset.shuffle(\n            buffer_size=shuffle_buffer_size, seed=seed, reshuffle_each_iteration=True\n        )\n\n    dataset = dataset.repeat(None)\n    dataset = dataset.batch(per_process_batch_size, drop_remainder=True)\n    dataset = dataset.prefetch(tf.data.AUTOTUNE)\n\n    return dataset.as_numpy_iterator()\n",python,tab
+90,191033,"utils/dataloader.py",253,4252,"def _tf_process_episode(episode_tensor, seq_len, image_h, image_w, image_c):\n    """"""\n    Processes a raw episode tensor in TensorFlow.\n    Takes a full episode, extracts a random sequence, and normalizes it.\n    Args:\n        episode_tensor: A TensorFlow tensor representing a full video episode.\n                        Expected shape: (dynamic_length, image_h, image_w, image_c)\n                        Expected dtype: e.g., tf.uint8 (raw pixel values)\n        seq_len: The desired length of the sub-sequence to extract.\n        image_h: The height of each frame.\n        image_w: The width of each frame.\n        image_c: The number of channels in each frame.\n    Returns:\n        A TensorFlow tensor representing the processed video sequence.\n        Shape: (seq_len, image_h, image_w, image_c)\n        Dtype: tf.float32 (normalized pixel values)\n    """"""\n    current_episode_len = tf.shape(episode_tensor)[0]\n\n    max_start_idx = current_episode_len - seq_len\n\n    start_idx = tf.random.uniform(\n        shape=(), minval=0, maxval=max_start_idx + 1, dtype=tf.int32\n    )\n\n    seq = episode_tensor[start_idx : start_idx + seq_len]\n\n    seq = tf.cast(seq, tf.float32) / 255.0\n\n    # Ensure the final shape is statically known for batching.\n    # tf.reshape is robust, but tf.ensure_shape or set_shape can also be used if confident.\n    processed_sequence = tf.reshape(seq, [seq_len, image_h, image_w, image_c])\n\n    return processed_sequence\n\n\ndef _parse_tfrecord_fn(example_proto, image_h, image_w, image_c):\n    feature_description = {\n        ""height"": tf.io.FixedLenFeature([], tf.int64),\n        ""width"": tf.io.FixedLenFeature([], tf.int64),\n        ""channels"": tf.io.FixedLenFeature([], tf.int64),\n        ""sequence_length"": tf.io.FixedLenFeature([], tf.int64),\n        ""raw_video"": tf.io.FixedLenFeature([], tf.string),\n    }\n    example = tf.io.parse_single_example(example_proto, feature_description)\n\n    video_shape = (example[""sequence_length""], image_h, image_w, image_c)\n\n    episode_tensor = tf.io.decode_raw(example[""raw_video""], out_type=tf.uint8)\n    episode_tensor = tf.reshape(episode_tensor, video_shape)\n\n    episode_tensor = tf.ensure_shape(episode_tensor, [None, image_h, image_w, image_c])\n    return episode_tensor\n\n\ndef _create_processed_dataset_from_file(file_path, image_h, image_w, image_c, seq_len, num_parallel_calls):\n    """"""Creates a fully processed dataset from a single TFRecord file.""""""\n    dataset = tf.data.TFRecordDataset([file_path])\n    \n    parse_fn = functools.partial(\n        _parse_tfrecord_fn, image_h=image_h, image_w=image_w, image_c=image_c\n    )\n    dataset = dataset.map(parse_fn, num_parallel_calls=num_parallel_calls)\n\n    tf_process_fn = functools.partial(\n        _tf_process_episode,\n        seq_len=seq_len,\n        image_h=image_h,\n        image_w=image_w,\n        image_c=image_c,\n    )\n    dataset = dataset.map(tf_process_fn, num_parallel_calls=num_parallel_calls)\n    \n    return dataset\n\n\ndef get_dataloader(\n    tfrecord_paths: list[str],\n    seq_len: int,\n    global_batch_size: int,\n    image_h: int,\n    image_w: int,\n    image_c: int,\n    shuffle_buffer_size: int = 1000,\n    num_parallel_calls: int = tf.data.AUTOTUNE,\n    seed: int = 42,\n    cycle_length: int = 4,\n    block_length: int = 1,\n):\n    """"""\n    Creates a tf.data.Dataset pipeline from TFRecord files.\n    """"""\n    if not tfrecord_paths:\n        raise ValueError(""tfrecord_paths list cannot be empty."")\n\n    process_id = jax.process_index()\n    num_processes = jax.process_count()\n\n    assert (\n        global_batch_size % num_processes == 0\n    ), f""Global batch size {global_batch_size} \\n        must be divisible by the number of JAX processes {num_processes} for proper sharding.""\n    per_process_batch_size = global_batch_size // num_processes\n\n    def dataset_fn(file_path):\n        return _create_processed_dataset_from_file(\n            file_path, image_h, image_w, image_c, seq_len, num_parallel_calls\n",python,content
+91,192106,"utils/dataloader.py",2928,0,"",python,selection_mouse
+92,193410,"utils/dataloader.py",2928,0,"\n",python,content
+93,193787,"utils/dataloader.py",2929,0,"    # Filter out episodes that are too short\n    def filter_short_episodes(episode_tensor):\n        return tf.shape(episode_tensor)[0] >= seq_len\n    \n    dataset = dataset.filter(filter_short_episodes)",python,content
+94,194713,"utils/dataloader.py",3131,0,"\n    ",python,content
+95,195064,"utils/dataloader.py",3132,4,"",python,content
+96,195607,"utils/dataloader.py",3080,0,"",python,selection_command
+97,195836,"utils/dataloader.py",3075,0,"",python,selection_command
+98,195995,"utils/dataloader.py",3021,0,"",python,selection_command
+99,196145,"utils/dataloader.py",2974,0,"",python,selection_command
+100,196311,"utils/dataloader.py",2929,0,"",python,selection_command
+101,202514,"TERMINAL",0,0,"git status",,terminal_command
+102,202573,"TERMINAL",0,0,"]633;E;2025-07-07 10:52:44 git status;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;COn branch revised-dataloader\r\nYour branch is up to date with 'origin/revised-dataloader'.\r\n\r\nChanges not staged for commit:\r\n  (use ""git add <file>..."" to update what will be committed)\r\n  (use ""git restore <file>..."" to discard changes in working directory)\r\n\tmodified:   utils/dataloader.py\r\n\r\nUntracked files:\r\n  (use ""git add <file>..."" to include in what will be committed)\r\n\tdata_tfrecord_duplicated/\r\n\tdata_tfrecords/\r\n\tlogs/\r\n\tread_tf_record.py\r\n\trequirements-franz.txt\r\n\tscripts_cremers/\r\n\tscripts_horeka/\r\n\tslurm-3309772.out\r\n\tslurm/\r\n\tutils/visualizer.py\r\n\r\nno changes added to commit (use ""git add"" and/or ""git commit -a"")\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+103,206815,"TERMINAL",0,0,"git add utils/dataloader.py",,terminal_command
+104,206843,"TERMINAL",0,0,"]633;E;2025-07-07 10:52:49 git add utils/dataloader.py ;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+105,217694,"TERMINAL",0,0,"git commit -m ""added filter for too short episodes in dataloader""",,terminal_command
+106,217741,"TERMINAL",0,0,"]633;E;2025-07-07 10:52:59 git commit -m ""added filter for too short episodes in dataloader"";f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+107,217921,"TERMINAL",0,0,"[revised-dataloader 1e306ff] added filter for too short episodes in dataloader\r\n 1 file changed, 6 insertions(+)\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+108,219165,"TERMINAL",0,0,"git push",,terminal_command
+109,219215,"TERMINAL",0,0,"]633;E;2025-07-07 10:53:01 git push;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+110,220680,"TERMINAL",0,0,"Enumerating objects: 7, done.\r\nCounting objects:  14% (1/7)\rCounting objects:  28% (2/7)\rCounting objects:  42% (3/7)\rCounting objects:  57% (4/7)\rCounting objects:  71% (5/7)\rCounting objects:  85% (6/7)\rCounting objects: 100% (7/7)\rCounting objects: 100% (7/7), done.\r\nDelta compression using up to 152 threads\r\nCompressing objects:  25% (1/4)\rCompressing objects:  50% (2/4)\rCompressing objects:  75% (3/4)\rCompressing objects: 100% (4/4)\rCompressing objects: 100% (4/4), done.\r\nWriting objects:  25% (1/4)\rWriting objects:  50% (2/4)\rWriting objects:  75% (3/4)\rWriting objects: 100% (4/4)\rWriting objects: 100% (4/4), 489 bytes | 244.00 KiB/s, done.\r\nTotal 4 (delta 3), reused 0 (delta 0), pack-reused 0\r\nremote: Resolving deltas:   0% (0/3)\rremote: Resolving deltas:  33% (1/3)\rremote: Resolving deltas:  66% (2/3)\rremote: Resolving deltas: 100% (3/3)\rremote: Resolving deltas: 100% (3/3), completed with 3 local objects.\r\n",,terminal_output
+111,220984,"TERMINAL",0,0,"To github.com:p-doom/jafar.git\r\n   1eac634..1e306ff  revised-dataloader -> revised-dataloader\r\n]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+112,704052,"TERMINAL",0,0,"queue",,terminal_command
+113,704136,"TERMINAL",0,0,"]633;E;2025-07-07 11:01:06 queue;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C[?1049h(B[?7hEvery 1.0s: squeue --mehkn1991.localdomain: Mon Jul  7 11:01:06 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3320180 accelerat train_la tum_cte0 CG\t0:00\t  1 hkn0405",,terminal_output
+114,705241,"TERMINAL",0,0,"7\t ",,terminal_output
+115,706271,"TERMINAL",0,0,"8\t ",,terminal_output
+116,707322,"TERMINAL",0,0,"9\t ",,terminal_output
+117,708343,"TERMINAL",0,0,"10\t ",,terminal_output
+118,708459,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1991:~/Projects/jafar]633;D;0",,terminal_output
+119,722511,"TERMINAL",0,0,"cd $ws_dir",,terminal_command
+120,724566,"TERMINAL",0,0,"cd  ..",,terminal_command
+121,726017,"TERMINAL",0,0,"cd logs/",,terminal_command
+122,726381,"TERMINAL",0,0,"ls",,terminal_command
+123,726418,"TERMINAL",0,0,"]633;E;2025-07-07 11:01:28 ls;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C3306965  logs_alfred  logs_franz  logs_mihir\r\n]0;tum_cte0515@hkn1991:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs]633;D;0",,terminal_output
+124,728472,"TERMINAL",0,0,"cd logs_mihir/",,terminal_command
+125,728753,"TERMINAL",0,0,"ls",,terminal_command
+126,728804,"TERMINAL",0,0,"]633;E;2025-07-07 11:01:30 ls;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C",,terminal_output
+127,729035,"TERMINAL",0,0,"train_lam_action_space_scaling_10_3320179.log  train_lam_action_space_scaling_6_3318549.log    train_lam_model_size_scaling_38M_3317231.log            train_tokenizer_batch_size_scaling_8_node_3320176.log  train_tokenizer_model_size_scaling_200M_3313563.log  train_tokenizer_model_size_scaling_37M_3316022.log  train_tokenizer_model_size_scaling_80M_3313564.log\r\ntrain_lam_action_space_scaling_10_3321529.log  train_lam_action_space_scaling_6_3320178.log    train_tokenizer_batch_size_scaling_16_node_3321526.log  train_tokenizer_batch_size_scaling_8_node_3321525.log  train_tokenizer_model_size_scaling_200M_3316020.log  train_tokenizer_model_size_scaling_37M_3317232.log  train_tokenizer_model_size_scaling_80M_3316026.log\r\ntrain_lam_action_space_scaling_12_3318546.log  train_lam_action_space_scaling_6_3321528.log    train_tokenizer_batch_size_scaling_1_node_3318551.log   train_tokenizer_minecraft_overfit_sample_3309656.log   train_tokenizer_model_size_scaling_227M_3317234.log  train_tokenizer_model_size_scaling_37M_3317239.log\r\ntrain_lam_action_space_scaling_12_3320177.log  train_lam_action_space_scaling_8_3318550.log    train_tokenizer_batch_size_scaling_2_node_3318552.log   train_tokenizer_model_size_scaling_127M_3317233.log    train_tokenizer_model_size_scaling_227M_3318555.log  train_tokenizer_model_size_scaling_37M_3318556.log\r\ntrain_lam_action_space_scaling_12_3321527.log  train_lam_minecraft_overfit_sample_3309655.log  train_tokenizer_batch_size_scaling_4_node_3318553.log   train_tokenizer_model_size_scaling_127M_3318554.log    train_tokenizer_model_size_scaling_227M_3320173.log  train_tokenizer_model_size_scaling_74M_3318557.log\r\ntrain_lam_action_space_scaling_20_3318547.log  train_lam_model_size_scaling_38M_3317098.log    train_tokenizer_batch_size_scaling_4_node_3320175.log   train_tokenizer_model_size_scaling_140M_3313562.log    train_tokenizer_model_size_scaling_227M_3321523.log  train_tokenizer_model_size_scaling_74M_3320174.log\r\ntrain_lam_action_space_scaling_50_3320180.log  train_lam_model_size_scaling_38M_3317115.log    train_tokenizer_batch_size_scaling_4_node_3321524.log   train_tokenizer_model_size_scaling_140M_3316019.log    train_tokenizer_model_size_scaling_37M_3313565.log   train_tokenizer_model_size_scaling_74M_3321522.log\r\n]0;tum_cte0515@hkn1991:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir]633;D;0",,terminal_output
+128,756754,"TERMINAL",0,0,"echo $(pwd)/train_tokenizer_batch_size_scaling_16_node_3321526.log",,terminal_command
+129,756798,"TERMINAL",0,0,"]633;E;2025-07-07 11:01:58 echo $(pwd)/train_tokenizer_batch_size_scaling_16_node_3321526.log;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir/train_tokenizer_batch_size_scaling_16_node_3321526.log\r\n]0;tum_cte0515@hkn1991:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir]633;D;0",,terminal_output
+130,1152416,"utils/dataloader.py",0,0,"",python,tab
+131,1177005,"TERMINAL",0,0,"git checkout runner",,terminal_command
+132,1177092,"TERMINAL",0,0,"]633;E;2025-07-07 11:08:59 git checkout runner;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;Cfatal: not a git repository (or any parent up to mount point /hkfs)\r\nStopping at filesystem boundary (GIT_DISCOVERY_ACROSS_FILESYSTEM not set).\r\n]0;tum_cte0515@hkn1991:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir]633;D;128",,terminal_output
+133,1190795,"utils/dataloader.py",0,0,"",python,tab
+134,1384534,"utils/dataloader.py",4648,0,"",python,selection_mouse
+135,1384551,"utils/dataloader.py",4647,0,"",python,selection_command
+136,1388072,"TERMINAL",0,0,"queue",,terminal_command
+137,1388130,"TERMINAL",0,0,"]633;E;2025-07-07 11:12:30 queue;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C[?1049h(B[?7hEvery 1.0s: squeue --mehkn1991.localdomain: Mon Jul  7 11:12:30 2025JOBID PARTITION     NAME     USER ST\tTIME  NODES NODELIST(REASON)3320180 accelerat train_la tum_cte0 CG\t0:00\t  1 hkn0405",,terminal_output
+138,1389126,"TERMINAL",0,0,"1\t ",,terminal_output
+139,1390163,"TERMINAL",0,0,"[?1049l\r[?1l>]0;tum_cte0515@hkn1991:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir]633;D;0",,terminal_output
+140,1393075,"TERMINAL",0,0,"idling",,terminal_command
+141,1393136,"TERMINAL",0,0,"]633;E;2025-07-07 11:12:35 idling;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C[?1049h(B[?7hEvery 1.0s: sinfo_t_idlehkn1991.localdomain: Mon Jul  7 11:12:35 2025Partition dev_cpuonly:     11 nodes idle\rPartition cpuonly:    325 nodes idle\rPartition dev_accelerated:\t 3 nodes idle\rPartition accelerated:     39 nodes idle\rPartition dev_accelerated-h100    :\t 0 nodes idle\rPartition accelerated-h100:\t 1 nodes idle\rPartition large:\t 7 nodes idle",,terminal_output
+142,1394285,"TERMINAL",0,0,"6\t [?1049l\r[?1l>]0;tum_cte0515@hkn1991:/hkfs/work/workspace/scratch/tum_ind3695-jafa_ws_shared/logs/logs_mihir]633;D;0",,terminal_output
+143,1397165,"TERMINAL",0,0,"cd",,terminal_command
+144,1397180,"TERMINAL",0,0,"]633;E;2025-07-07 11:12:39 cd;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;C]0;tum_cte0515@hkn1991:~]633;D;0]633;P;Cwd=/home/hk-project-p0023960/tum_cte0515",,terminal_output
+145,1404501,"TERMINAL",0,0,"cd Projects/jafar",,terminal_command
+146,1445110,"TERMINAL",0,0,"salloc --time=10:00:00 --partition=accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5",,terminal_command
+147,1445175,"TERMINAL",0,0,"]633;E;2025-07-07 11:13:27 salloc --time=10:00:00 --partition=accelerated --nodes=1 --ntasks-per-node=1 --gres=gpu:1 --cpus-per-task=5 ;f27e7af0-1f60-464b-afac-68c0fe98f46d]633;Csalloc: Granted job allocation 3326035\r\n",,terminal_output
+148,1445309,"TERMINAL",0,0,"salloc: Waiting for resource configuration\r\n",,terminal_output
+149,1472365,"TERMINAL",0,0,"salloc: Nodes hkn0734 are ready for job\r\n",,terminal_output
+150,1473217,"TERMINAL",0,0,"]0;tum_cte0515@hkn0734:~/Projects/jafar[?2004h[tum_cte0515@hkn0734 jafar]$ ",,terminal_output
+151,1699781,"TERMINAL",0,0,"s",,terminal_output
+152,1699852,"TERMINAL",0,0,"[?25lo[?25h",,terminal_output
+153,1699960,"TERMINAL",0,0,"[?25lu[?25h",,terminal_output
+154,1700026,"TERMINAL",0,0,"[?25lr[?25h",,terminal_output
+155,1700174,"TERMINAL",0,0,"[?25lc[?25h",,terminal_output
+156,1700311,"TERMINAL",0,0,"[?25le[?25h",,terminal_output
+157,1700451,"TERMINAL",0,0,"[?25l [?25h",,terminal_output
+158,1700517,"TERMINAL",0,0,"[?25l.[?25h",,terminal_output
+159,1700959,"TERMINAL",0,0,"[?25lv[?25h[?25le[?25h",,terminal_output
+160,1701278,"TERMINAL",0,0,"nv/",,terminal_output
+161,1701883,"TERMINAL",0,0,"[?25lb[?25h",,terminal_output
+162,1702072,"TERMINAL",0,0,"in/",,terminal_output
+163,1702317,"TERMINAL",0,0,"[?25la[?25h",,terminal_output
+164,1702396,"TERMINAL",0,0,"[?25lc[?25h",,terminal_output
+165,1702648,"TERMINAL",0,0,"tivate",,terminal_output
+166,1702996,"TERMINAL",0,0,"\r\n[?2004l\r]0;tum_cte0515@hkn0734:~/Projects/jafar[?2004h(jafar) [tum_cte0515@hkn0734 jafar]$ ",,terminal_output
+167,1705224,"TERMINAL",0,0,"[?25lgi[?25h[?25li[?25h",,terminal_output
+168,1705388,"TERMINAL",0,0,"[?25lt[?25h",,terminal_output
+169,1705441,"TERMINAL",0,0,"[?25l [?25h",,terminal_output
+170,1705641,"TERMINAL",0,0,"[?25le[?25h",,terminal_output
+171,1707051,"TERMINAL",0,0,"[?25lc[?25h",,terminal_output
+172,1707116,"TERMINAL",0,0,"[?25lh[?25h",,terminal_output
+173,1707261,"TERMINAL",0,0,"[?25le[?25h",,terminal_output
+174,1707413,"TERMINAL",0,0,"[?25lc[?25h[?25lk[?25h",,terminal_output
+175,1707656,"TERMINAL",0,0,"[?25lo[?25h",,terminal_output
+176,1707736,"TERMINAL",0,0,"[?25lu[?25h",,terminal_output
+177,1707856,"TERMINAL",0,0,"[?25lt[?25h",,terminal_output
+178,1707922,"TERMINAL",0,0,"[?25l [?25h",,terminal_output
+179,1708772,"TERMINAL",0,0,"[?25lf[?25h",,terminal_output
+180,1708895,"TERMINAL",0,0,"[?25li[?25h",,terminal_output
+181,1709131,"TERMINAL",0,0,"[?25lx[?25h",,terminal_output
+182,1709618,"TERMINAL",0,0,"[?25l-[?25h",,terminal_output
+183,1709774,"TERMINAL",0,0,"[?25ls[?25h",,terminal_output
+184,1709934,"TERMINAL",0,0,"[?25la[?25h",,terminal_output
+185,1710041,"TERMINAL",0,0,"[?25lm[?25h",,terminal_output
+186,1710274,"TERMINAL",0,0,"[?25lp[?25h[?25ll[?25h",,terminal_output
+187,1710539,"TERMINAL",0,0,"[?25li[?25h[?25ln[?25h",,terminal_output
+188,1710608,"TERMINAL",0,0,"[?25lg[?25h",,terminal_output
+189,1710830,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+190,1711585,"TERMINAL",0,0,"Switched to branch 'fix-sampling'\r\nYour branch is behind 'origin/fix-sampling' by 2 commits, and can be fast-forwarded.\r\n  (use ""git pull"" to update your local branch)\r\n]0;tum_cte0515@hkn0734:~/Projects/jafar[?2004h(jafar) [tum_cte0515@hkn0734 jafar]$ ",,terminal_output
+191,1712699,"TERMINAL",0,0,"g",,terminal_output
+192,1712812,"TERMINAL",0,0,"[?25li[?25h",,terminal_output
+193,1712904,"TERMINAL",0,0,"[?25lt[?25h[?25l [?25h",,terminal_output
+194,1713002,"TERMINAL",0,0,"[?25lp[?25h",,terminal_output
+195,1713211,"TERMINAL",0,0,"[?25lu[?25h",,terminal_output
+196,1713391,"TERMINAL",0,0,"[?25ll[?25h",,terminal_output
+197,1713523,"TERMINAL",0,0,"[?25ll[?25h",,terminal_output
+198,1713632,"TERMINAL",0,0,"\r\n[?2004l\r",,terminal_output
+199,1715291,"TERMINAL",0,0,"Updating ae9451f..6e623c6\r\nFast-forward\r\n genie.py | 16 ++++++++--------\r\n 1 file changed, 8 insertions(+), 8 deletions(-)\r\n]0;tum_cte0515@hkn0734:~/Projects/jafar[?2004h(jafar) [tum_cte0515@hkn0734 jafar]$ ",,terminal_output
+200,1715555,"",0,0,"Switched from branch 'revised-dataloader' to 'fix-sampling'",,git_branch_checkout
+201,1722688,"genie.py",0,0,"from typing import Dict, Any\n\nimport optax\nimport jax\nimport jax.numpy as jnp\nimport flax.linen as nn\nfrom jax import NamedSharding\nfrom flax.training.train_state import TrainState\nfrom flax.training import orbax_utils\nfrom orbax.checkpoint import PyTreeCheckpointer\n\nfrom models.dynamics import DynamicsMaskGIT\nfrom models.lam import LatentActionModel\nfrom models.tokenizer import TokenizerVQVAE\n\n\nclass Genie(nn.Module):\n    """"""Genie model""""""\n\n    # --- Tokenizer ---\n    in_dim: int\n    tokenizer_dim: int\n    latent_patch_dim: int\n    num_patch_latents: int\n    patch_size: int\n    tokenizer_num_blocks: int\n    tokenizer_num_heads: int\n    # --- LAM ---\n    lam_dim: int\n    latent_action_dim: int\n    num_latent_actions: int\n    lam_patch_size: int\n    lam_num_blocks: int\n    lam_num_heads: int\n    # --- Dynamics ---\n    dyna_dim: int\n    dyna_num_blocks: int\n    dyna_num_heads: int\n    dropout: float = 0.0\n    mask_limit: float = 0.0\n\n    def setup(self):\n        self.tokenizer = TokenizerVQVAE(\n            in_dim=self.in_dim,\n            model_dim=self.tokenizer_dim,\n            latent_dim=self.latent_patch_dim,\n            num_latents=self.num_patch_latents,\n            patch_size=self.patch_size,\n            num_blocks=self.tokenizer_num_blocks,\n            num_heads=self.tokenizer_num_heads,\n            dropout=0.0,\n            codebook_dropout=0.0,\n        )\n        self.lam = LatentActionModel(\n            in_dim=self.in_dim,\n            model_dim=self.lam_dim,\n            latent_dim=self.latent_patch_dim,\n            num_latents=self.num_latent_actions,\n            patch_size=self.lam_patch_size,\n            num_blocks=self.lam_num_blocks,\n            num_heads=self.lam_num_heads,\n            dropout=0.0,\n            codebook_dropout=0.0,\n        )\n        self.dynamics = DynamicsMaskGIT(\n            model_dim=self.dyna_dim,\n            num_latents=self.num_patch_latents,\n            num_blocks=self.dyna_num_blocks,\n            num_heads=self.dyna_num_heads,\n            dropout=self.dropout,\n            mask_limit=self.mask_limit,\n        )\n\n    def __call__(self, batch: Dict[str, Any], training: bool = True) -> Dict[str, Any]:\n        tokenizer_outputs = self.tokenizer.vq_encode(batch[""videos""], training=False)\n        lam_outputs = self.lam.vq_encode(batch[""videos""], training=False)\n        outputs = dict(\n            video_tokens=jax.lax.stop_gradient(tokenizer_outputs[""indices""]),\n            latent_actions=jax.lax.stop_gradient(lam_outputs[""z_q""]),\n        )\n        outputs[""mask_rng""] = batch[""mask_rng""]\n        dyna_outputs = self.dynamics(outputs, training)\n        outputs.update(dyna_outputs)\n        mle_indices = jnp.argmax(outputs[""token_logits""], axis=-1)\n        outputs[""recon""] = self.tokenizer.decode(\n            mle_indices, batch[""videos""].shape[2:4]\n        )\n        return outputs\n\n    @nn.compact\n    def sample(\n        self,\n        batch: Dict[str, Any],\n        seq_len: int,\n        steps: int = 25,\n        temperature: float = 1,\n        sample_argmax: bool = False,\n    ) -> Any:\n        """"""\n        Autoregressively samples up to `seq_len` future frames, following Figure 8 of the paper.\n\n        - Input frames are tokenized once.\n        - Future frames are generated autoregressively in token space.\n        - All frames are detokenized in a single pass.\n\n        Note:\n        - For interactive or step-wise sampling, detokenization should occur after each action.\n        - To maintain consistent tensor shapes across timesteps, all current and future frames are decoded at every step.\n        - Temporal causal structure is preserved by \n            a) reapplying the mask before each decoding step.\n            b) a temporal causal mask is applied within each ST-transformer block.\n\n        Dimension keys:\n            B: batch size  \n            T: number of input (conditioning) frames  \n            N: patches per frame  \n            S: sequence length  \n            A: action space  \n            D: model latent dimension\n        """"""\n        # --- Encode videos and actions ---\n        tokenizer_out = self.tokenizer.vq_encode(batch[""videos""], training=False)\n        token_idxs = tokenizer_out[""indices""] # (B, T, N)\n        B, T, N = token_idxs.shape\n        pad_shape = (B, seq_len - T, N)\n        pad = jnp.zeros(pad_shape, dtype=token_idxs.dtype)\n        token_idxs = jnp.concatenate([token_idxs, pad], axis=1) # (B, S, N)\n        action_tokens = self.lam.vq.get_codes(batch[""latent_actions""])\n\n        for step_t in range(T, seq_len):\n            print(f""Sampling Frame {step_t}..."")\n            # mask current and future frames (i.e., t >= step_t)\n            mask = jnp.arange(seq_len) >= step_t # (S,)\n            mask = jnp.broadcast_to(mask[None, :, None], (B, seq_len, N)) # (B, S, N)\n            mask = mask.astype(bool)\n            token_idxs *= ~mask\n\n            # --- Initialize MaskGIT ---\n            init_carry = (\n                batch[""rng""],\n                token_idxs,\n                mask,\n                action_tokens,\n            )\n\n            MaskGITLoop = nn.scan(\n                MaskGITStep,\n                variable_broadcast=""params"",\n                split_rngs={""params"": False},\n                in_axes=0,\n                out_axes=0,\n                length=steps,\n            )\n\n            # --- Run MaskGIT loop ---\n            loop_fn = MaskGITLoop(\n                dynamics=self.dynamics,\n                tokenizer=self.tokenizer,\n                temperature=temperature,\n                sample_argmax=sample_argmax,\n                steps=steps,\n            )\n            final_carry, _ = loop_fn(init_carry, jnp.arange(steps))\n            token_idxs = final_carry[1]\n\n        final_frames = self.tokenizer.decode(\n            token_idxs,\n            video_hw=batch[""videos""].shape[2:4],\n        )\n        return final_frames \n\n    def vq_encode(self, batch, training) -> Dict[str, Any]:\n        # --- Preprocess videos ---\n        lam_output = self.lam.vq_encode(batch[""videos""], training=training)\n        return lam_output[""indices""]\n\n\nclass MaskGITStep(nn.Module):\n    dynamics: nn.Module\n    tokenizer: nn.Module\n    temperature: float\n    sample_argmax: bool\n    steps: int\n\n    @nn.compact\n    def __call__(self, carry, x):\n        rng, token_idxs, mask, action_tokens = carry\n        step = x\n        N = token_idxs.shape[2]\n\n        # --- Construct + encode video ---\n        vid_embed = self.dynamics.patch_embed(token_idxs) # (B, S, N, D)\n        mask_token = self.dynamics.mask_token  # (1, 1, 1, D,)\n        mask_expanded = mask[..., None] # (B, S, N, 1) \n        vid_embed = jnp.where(mask_expanded, mask_token, vid_embed)\n\n        # --- Predict transition ---\n        act_embed = self.dynamics.action_up(action_tokens)\n        vid_embed += jnp.pad(act_embed, ((0, 0), (1, 0), (0, 0), (0, 0)))\n        unmasked_ratio = jnp.cos(jnp.pi * (step + 1) / (self.steps * 2))\n        step_temp = self.temperature * (1.0 - unmasked_ratio)\n        final_logits = self.dynamics.dynamics(vid_embed) / step_temp\n\n        # --- Sample new tokens for final frame ---\n        if self.sample_argmax:\n            sampled_token_idxs = jnp.argmax(final_logits, axis=-1)\n        else:\n            rng, _rng = jax.random.split(rng)\n            sampled_token_idxs = jnp.where(\n                step == self.steps - 1,\n                jnp.argmax(final_logits, axis=-1),\n                jax.random.categorical(_rng, final_logits),\n            )\n        gather_fn = jax.vmap(jax.vmap(jax.vmap(lambda x, y: x[y])))\n        final_token_probs = gather_fn(jax.nn.softmax(final_logits), sampled_token_idxs)\n        final_token_probs += ~mask\n        # Update masked tokens only\n        token_idxs = jnp.where(mask, sampled_token_idxs, token_idxs)\n\n        # --- Update mask ---\n        num_unmasked_tokens = jnp.round(N * (1.0 - unmasked_ratio)).astype(int)\n        idx_mask = jnp.arange(final_token_probs.shape[-1]) > num_unmasked_tokens\n        sorted_idxs = jnp.argsort(final_token_probs, axis=-1, descending=True)\n        mask_update_fn = jax.vmap(lambda msk, ids: msk.at[ids].set(idx_mask))\n        new_mask = mask_update_fn(mask, sorted_idxs)\n\n        new_carry = (rng, token_idxs, new_mask, action_tokens)\n        return new_carry, None\n\ndef restore_genie_components(\n    train_state: TrainState,\n    sharding: NamedSharding,\n    inputs: Dict[str, jax.Array],\n    rng: jax.Array,\n    args,\n):\n    """"""Restore pre-trained Genie components""""""\n    rng, _rng = jax.random.split(rng)\n\n    dummy_tokenizer = TokenizerVQVAE(\n        in_dim=args.image_channels,\n        model_dim=args.tokenizer_dim,\n        latent_dim=args.latent_patch_dim,\n        num_latents=args.num_patch_latents,\n        patch_size=args.patch_size,\n        num_blocks=args.tokenizer_num_blocks,\n        num_heads=args.tokenizer_num_heads,\n        dropout=args.dropout,\n        codebook_dropout=args.dropout,\n    )\n    dummy_lam = LatentActionModel(\n        in_dim=args.image_channels,\n        model_dim=args.lam_dim,\n        latent_dim=args.latent_patch_dim,\n        num_latents=args.num_latent_actions,\n        patch_size=args.lam_patch_size,\n        num_blocks=args.lam_num_blocks,\n        num_heads=args.lam_num_heads,\n        dropout=args.dropout,\n        codebook_dropout=args.dropout,\n    )\n    tokenizer_init_params = dummy_tokenizer.init(_rng, inputs)\n    lam_init_params = dummy_lam.init(_rng, inputs)\n\n    # dummy values since we only use tx to initialize the dummy train states\n    dummy_tx = optax.adamw(\n        learning_rate=optax.constant_schedule(args.max_lr),\n        b1=0.9,\n        b2=0.9,\n        weight_decay=1e-4,\n    )\n\n    dummy_tokenizer_train_state = TrainState.create(\n        apply_fn=dummy_tokenizer.apply, params=tokenizer_init_params, tx=dummy_tx\n    )\n    dummy_lam_train_state = TrainState.create(\n        apply_fn=dummy_lam.apply, params=lam_init_params, tx=dummy_tx\n    )\n\n    def create_abstract_sharded_pytree(pytree_template, sharding_spec):\n        """"""Replaces arrays in a pytree with ShapeDtypeStructs having the given sharding.""""""\n\n        def map_fn(leaf_template):\n            if hasattr(leaf_template, ""shape"") and hasattr(leaf_template, ""dtype""):\n                return jax.ShapeDtypeStruct(\n                    leaf_template.shape, leaf_template.dtype, sharding=sharding_spec\n                )\n            return leaf_template\n\n        return jax.tree_util.tree_map(map_fn, pytree_template)\n\n    abstract_sharded_tokenizer_state = create_abstract_sharded_pytree(\n        dummy_tokenizer_train_state, sharding\n    )\n    abstract_sharded_lam_state = create_abstract_sharded_pytree(\n        dummy_lam_train_state, sharding\n    )\n\n    tokenizer_restore_target = {""model"": abstract_sharded_tokenizer_state}\n    lam_restore_target = {""model"": abstract_sharded_lam_state}\n\n    tokenizer_restore_args = orbax_utils.restore_args_from_target(\n        tokenizer_restore_target\n    )\n    lam_restore_args = orbax_utils.restore_args_from_target(lam_restore_target)\n\n    restored_tokenizer_params = (\n        PyTreeCheckpointer()\n        .restore(\n            args.tokenizer_checkpoint,\n            item=tokenizer_restore_target,\n            restore_args=tokenizer_restore_args,\n        )[""model""]\n        .params[""params""]\n    )\n    restored_lam_params = (\n        PyTreeCheckpointer()\n        .restore(\n            args.lam_checkpoint, item=lam_restore_target, restore_args=lam_restore_args\n        )[""model""]\n        .params[""params""]\n    )\n    # Genie does not initialize all LAM modules, thus we omit those extra modules during restoration\n    # (f.srambical) FIXME: Currently, this is a small HBM memory crunch since the LAM's decoder is loaded into HBM and immediately dicarded.\n    # A workaround would be to restore to host memory first, and only move the weights to HBM after pruning the decoder\n    restored_lam_params = {\n        k: v\n        for k, v in restored_lam_params.items()\n        if k in train_state.params[""params""][""lam""]\n    }\n\n    train_state.params[""params""][""tokenizer""].update(restored_tokenizer_params)\n    train_state.params[""params""][""lam""].update(restored_lam_params)\n\n    return train_state\n",python,tab
+202,1731948,"sample.py",0,0,"from dataclasses import dataclass\nimport time\nimport os\n\nimport dm_pix as pix\nimport einops\nimport jax\nimport jax.numpy as jnp\nimport flax.linen as nn\nimport numpy as np\nfrom orbax.checkpoint import PyTreeCheckpointer\nfrom PIL import Image, ImageDraw\nimport tyro\n\nfrom genie import Genie\nfrom utils.dataloader import get_dataloader\n\n\n@dataclass\nclass Args:\n    # Experiment\n    seed: int = 0\n    seq_len: int = 16\n    image_channels: int = 3\n    image_height: int = 90\n    image_width: int = 160\n    data_dir: str = ""data/coinrun_episodes""\n    checkpoint: str = """"\n    # Sampling\n    batch_size: int = 1\n    maskgit_steps: int = 25\n    temperature: float = 1.0\n    sample_argmax: bool = True\n    start_frame: int = 0\n    # Tokenizer checkpoint\n    tokenizer_dim: int = 512\n    latent_patch_dim: int = 32\n    num_patch_latents: int = 1024\n    patch_size: int = 4\n    tokenizer_num_blocks: int = 8\n    tokenizer_num_heads: int = 8\n    # LAM checkpoint\n    lam_dim: int = 512\n    latent_action_dim: int = 32\n    num_latent_actions: int = 6\n    lam_patch_size: int = 16\n    lam_num_blocks: int = 8\n    lam_num_heads: int = 8\n    # Dynamics checkpoint\n    dyna_dim: int = 512\n    dyna_num_blocks: int = 12\n    dyna_num_heads: int = 8\n\n\nargs = tyro.cli(Args)\nrng = jax.random.PRNGKey(args.seed)\n\n# --- Load Genie checkpoint ---\ngenie = Genie(\n    # Tokenizer\n    in_dim=args.image_channels,\n    tokenizer_dim=args.tokenizer_dim,\n    latent_patch_dim=args.latent_patch_dim,\n    num_patch_latents=args.num_patch_latents,\n    patch_size=args.patch_size,\n    tokenizer_num_blocks=args.tokenizer_num_blocks,\n    tokenizer_num_heads=args.tokenizer_num_heads,\n    # LAM\n    lam_dim=args.lam_dim,\n    latent_action_dim=args.latent_action_dim,\n    num_latent_actions=args.num_latent_actions,\n    lam_patch_size=args.lam_patch_size,\n    lam_num_blocks=args.lam_num_blocks,\n    lam_num_heads=args.lam_num_heads,\n    # Dynamics\n    dyna_dim=args.dyna_dim,\n    dyna_num_blocks=args.dyna_num_blocks,\n    dyna_num_heads=args.dyna_num_heads,\n)\nrng, _rng = jax.random.split(rng)\nimage_shape = (args.image_height, args.image_width, args.image_channels)\ndummy_inputs = dict(\n    videos=jnp.zeros((args.batch_size, args.seq_len, *image_shape), dtype=jnp.float32),\n    mask_rng=_rng,\n)\nrng, _rng = jax.random.split(rng)\nparams = genie.init(_rng, dummy_inputs)\nckpt = PyTreeCheckpointer().restore(args.checkpoint)[""model""][""params""][""params""]\nparams[""params""].update(ckpt)\n\n\ndef _sampling_wrapper(module, batch):\n    return module.sample(batch, args.seq_len, args.maskgit_steps, args.temperature, args.sample_argmax)\n\n# --- Define autoregressive sampling loop ---\ndef _autoreg_sample(rng, video_batch, action_batch):\n    vid = video_batch[:, : args.start_frame + 1]\n    sampling_fn = jax.jit(nn.apply(_sampling_wrapper, genie)) \n    rng, _rng = jax.random.split(rng)\n    batch = dict(videos=vid, latent_actions=action_batch, rng=_rng)\n    generated_vid = sampling_fn(\n        params,\n        batch\n    )\n    return generated_vid\n\n# --- Get video + latent actions ---\ntfrecord_files = [\n    os.path.join(args.data_dir, x)\n    for x in os.listdir(args.data_dir)\n    if x.endswith("".tfrecord"")\n]\ndataloader = get_dataloader(\n    tfrecord_files,\n    args.seq_len,\n    args.batch_size,\n    args.image_height,\n    args.image_width,\n    args.image_channels,\n    seed=args.seed,\n)\nvideo_batch = next(iter(dataloader))\n# Get latent actions from first video only; clip them down to the specified seq_len\nfirst_video = video_batch[:1, :args.seq_len]\nbatch = dict(videos=first_video)\naction_batch = genie.apply(params, batch, False, method=Genie.vq_encode)\naction_batch = action_batch.reshape(1, args.seq_len - 1, 1)\n# Use actions from first video for all videos\naction_batch = jnp.repeat(action_batch, video_batch.shape[0], axis=0)\n\n# --- Sample + evaluate video ---\nvid = _autoreg_sample(rng, video_batch, action_batch)\ngt = video_batch[:, : vid.shape[1]].clip(0, 1).reshape(-1, *video_batch.shape[2:])\nrecon = vid.clip(0, 1).reshape(-1, *vid.shape[2:])\nssim = pix.ssim(gt[:, args.start_frame + 1 :], recon[:, args.start_frame + 1 :]).mean()\nprint(f""SSIM: {ssim}"")\n\n# --- Construct video ---\nfirst_true = (video_batch[0:1] * 255).astype(np.uint8)\nfirst_pred = (vid[0:1] * 255).astype(np.uint8)\nfirst_video_comparison = np.zeros((2, *vid.shape[1:5]), dtype=np.uint8)\nfirst_video_comparison[0] = first_true[:, : vid.shape[1]]\nfirst_video_comparison[1] = first_pred\n# For other videos, only show generated video\nother_preds = (vid[1:] * 255).astype(np.uint8)\nall_frames = np.concatenate([first_video_comparison, other_preds], axis=0)\nflat_vid = einops.rearrange(all_frames, ""n t h w c -> t h (n w) c"")\n\n# --- Save video ---\nimgs = [Image.fromarray(img) for img in flat_vid]\n# Write actions on each frame\nfor img, action in zip(imgs[1:], action_batch[0, :, 0]):\n    d = ImageDraw.Draw(img)\n    d.text((2, 2), f""{action}"", fill=255)\nimgs[0].save(\n    f""generation_{time.time()}.gif"",\n    save_all=True,\n    append_images=imgs[1:],\n    duration=250,\n    loop=0,\n)\n",python,tab
+203,1733276,"sample.py",3057,0,"",python,selection_mouse
+204,1733276,"sample.py",3056,0,"",python,selection_command
+205,1734341,"genie.py",0,0,"",python,tab
+206,1758226,"genie.py",4621,0,"",python,selection_mouse
+207,1761801,"genie.py",4527,0,"",python,selection_mouse
+208,1773944,"genie.py",4950,0,"",python,selection_mouse
+209,1774091,"genie.py",4946,5,"batch",python,selection_mouse
+210,1782472,"genie.py",4954,0,"",python,selection_mouse
+211,1782652,"genie.py",4953,3,"rng",python,selection_mouse
+212,10441352,"genie.py",0,0,"Switched from branch 'fix-sampling' to 'correct-batched-sampling'",python,git_branch_checkout
+213,15726982,"genie.py",0,0,"Switched from branch 'correct-batched-sampling' to 'main'",python,git_branch_checkout

927a8af5474e5654810c00ce2e09fd2de87d3e5722f33fa1090d867db114e403/crowd-code-f1a23455-555b-44b7-b7f2-5fb8550d75021753722279799-2025_07_28-19.04.51.762/source.csv

@@ -0,0 +1,3 @@
+Sequence,Time,File,RangeOffset,RangeLength,Text,Language,Type
+1,5,"utils/nn.py",0,0,"import math\nfrom typing import Tuple\n\nfrom flax import linen as nn\nimport jax\nimport jax.numpy as jnp\nimport einops\n\n\nclass PositionalEncoding(nn.Module):\n    """"""https://uvadlc-notebooks.readthedocs.io/en/latest/tutorial_notebooks/JAX/tutorial6/Transformers_and_MHAttention.html""""""\n\n    d_model: int  # Hidden dimensionality of the input.\n    max_len: int = 5000  # Maximum length of a sequence to expect.\n\n    def setup(self):\n        # Create matrix of [SeqLen, HiddenDim] representing the positional encoding for max_len inputs\n        self.pe = jnp.zeros((self.max_len, self.d_model))\n        position = jnp.arange(0, self.max_len, dtype=jnp.float32)[:, None]\n        div_term = jnp.exp(\n            jnp.arange(0, self.d_model, 2) * (-math.log(10000.0) / self.d_model)\n        )\n        self.pe = self.pe.at[:, 0::2].set(jnp.sin(position * div_term))\n        self.pe = self.pe.at[:, 1::2].set(jnp.cos(position * div_term))\n\n    def __call__(self, x):\n        x = x + self.pe[: x.shape[2]]\n        return x\n\n\nclass STBlock(nn.Module):\n    dim: int\n    ffn_dim: int\n    num_heads: int\n    dropout: float\n    param_dtype: jnp.dtype\n    dtype: jnp.dtype\n    use_flash_attention: bool\n\n    @nn.remat\n    @nn.compact\n    def __call__(self, x: jax.Array) -> jax.Array:\n        # --- Spatial attention ---\n        z = PositionalEncoding(self.dim)(x)\n        z = nn.LayerNorm(\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n        )(z)\n        z = nn.MultiHeadAttention(\n            num_heads=self.num_heads,\n            qkv_features=self.dim,\n            dropout_rate=self.dropout,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            attention_fn=_create_flash_attention_fn(\n                self.use_flash_attention, is_causal=False\n            ),\n        )(z)\n        x = x + z\n\n        # --- Temporal attention ---\n        x = x.swapaxes(1, 2)\n        z = PositionalEncoding(self.dim)(x)\n        z = nn.LayerNorm(\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n        )(z)\n        causal_mask = jnp.tri(z.shape[-2])\n        z = nn.MultiHeadAttention(\n            num_heads=self.num_heads,\n            qkv_features=self.dim,\n            dropout_rate=self.dropout,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n            attention_fn=_create_flash_attention_fn(\n                self.use_flash_attention, is_causal=True\n            ),\n            # FIXME (f.srambical): check whether we should still pass the mask if we set is_causal=True\n        )(z, mask=causal_mask)\n        x = x + z\n        x = x.swapaxes(1, 2)\n\n        # --- Feedforward ---\n        z = nn.LayerNorm(\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n        )(x)\n        z = nn.Dense(\n            self.ffn_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n        )(z)\n        z = nn.gelu(z)\n        z = nn.Dense(\n            self.dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n        )(z)\n        x = x + z\n\n        return x\n\n\nclass STTransformer(nn.Module):\n    model_dim: int\n    ffn_dim: int\n    out_dim: int\n    num_blocks: int\n    num_heads: int\n    dropout: float\n    param_dtype: jnp.dtype\n    dtype: jnp.dtype\n    use_flash_attention: bool\n\n    @nn.compact\n    def __call__(self, x: jax.Array) -> jax.Array:\n        x = nn.Sequential(\n            [\n                nn.LayerNorm(\n                    param_dtype=self.param_dtype,\n                    dtype=self.dtype,\n                ),\n                nn.Dense(\n                    self.model_dim,\n                    param_dtype=self.param_dtype,\n                    dtype=self.dtype,\n                ),\n                nn.LayerNorm(\n                    param_dtype=self.param_dtype,\n                    dtype=self.dtype,\n                ),\n            ]\n        )(x)\n        for _ in range(self.num_blocks):\n            x = STBlock(\n                dim=self.model_dim,\n                ffn_dim=self.ffn_dim,\n                num_heads=self.num_heads,\n                dropout=self.dropout,\n                param_dtype=self.param_dtype,\n                dtype=self.dtype,\n                use_flash_attention=self.use_flash_attention,\n            )(x)\n        x = nn.Dense(\n            self.out_dim,\n            param_dtype=self.param_dtype,\n            dtype=self.dtype,\n        )(x)\n        return x  # (B, T, E)\n\n\ndef normalize(x):\n    return x / (jnp.linalg.norm(x, ord=2, axis=-1, keepdims=True) + 1e-8)\n\n\nclass VectorQuantizer(nn.Module):\n    latent_dim: int\n    num_latents: int\n    dropout: float\n\n    def setup(self):\n        self.codebook = normalize(\n            self.param(\n                ""codebook"",\n                nn.initializers.lecun_uniform(),\n                (self.num_latents, self.latent_dim),\n            )\n        )\n        self.drop = nn.Dropout(self.dropout, deterministic=False)\n\n    def __call__(\n        self, x: jax.Array, training: bool\n    ) -> Tuple[jax.Array, jax.Array, jax.Array, jax.Array]:\n        # --- Compute distances ---\n        x = normalize(x)\n        codebook = normalize(self.codebook)\n        distance = -jnp.matmul(x, codebook.T)\n        if training:\n            dropout_key = self.make_rng(""dropout"")\n            distance = self.drop(distance, rng=dropout_key)\n\n        # --- Get indices and embeddings ---\n        indices = jnp.argmin(distance, axis=-1)\n        z = self.codebook[indices]\n\n        # --- Straight through estimator ---\n        z_q = x + jax.lax.stop_gradient(z - x)\n        return z_q, z, x, indices\n\n    def get_codes(self, indices: jax.Array):\n        return self.codebook[indices]\n\n\ndef _create_flash_attention_fn(use_flash_attention: bool, is_causal: bool):\n    """"""\n    Create an attention function that uses flash attention if enabled.\n\n    Flax MultiHeadAttention provides tensors with shape (batch..., length, num_heads, head_dim)\n    jax.nn.dot_product_attention expects (batch, length, num_heads, head_dim).\n\n    We need to reshape to ensure compatibility. cuDNN's flash attention additionally\n    requires a sequence length that is a multiple of 4. We pad the sequence length to the nearest\n    multiple of 4 and mask accordingly.\n    """"""\n\n    def attention_fn(query, key, value, bias=None, mask=None, **kwargs):\n        implementation = ""cudnn"" if use_flash_attention else None\n\n        def _rearrange(x):\n            return einops.rearrange(x, ""... l h d -> (...) l h d"")\n\n        def _pad(x):\n            return jnp.pad(x, ((0, 0), (0, pad_size), (0, 0), (0, 0)))\n\n        def _fuse_masks(mask: jax.Array, attention_mask: jax.Array) -> jax.Array:\n            mask_bool = mask.astype(jnp.bool_)\n            expanded_mask = jnp.pad(\n                mask_bool, ((0, pad_size), (0, pad_size)), constant_values=False\n            )\n            return jnp.logical_and(attention_mask, expanded_mask)\n\n        original_shape = query.shape\n        original_seq_len = query.shape[-3]\n\n        # Pad to nearest multiple of 4\n        target_seq_len = ((original_seq_len + 3) // 4) * 4\n        pad_size = target_seq_len - original_seq_len\n\n        query_4d = _pad(_rearrange(query))\n        key_4d = _pad(_rearrange(key))\n        value_4d = _pad(_rearrange(value))\n\n        attention_mask = jnp.ones((target_seq_len, target_seq_len), dtype=jnp.bool_)\n        attention_mask = attention_mask.at[original_seq_len:, :].set(False)\n        attention_mask = attention_mask.at[:, original_seq_len:].set(False)\n\n        mask_4d = (\n            _fuse_masks(mask, attention_mask) if mask is not None else attention_mask\n        )\n        mask_4d = mask_4d[jnp.newaxis, jnp.newaxis, :, :]  # (1, 1, seq_len, seq_len)\n\n        bias_4d = _pad(_rearrange(bias)) if bias is not None else None\n\n        output_4d = jax.nn.dot_product_attention(\n            query=query_4d,\n            key=key_4d,\n            value=value_4d,\n            bias=bias_4d,\n            mask=mask_4d,\n            implementation=implementation,\n            is_causal=is_causal,\n            **kwargs\n        )\n        return output_4d[..., :original_seq_len, :, :].reshape(original_shape)\n\n    return attention_fn\n",python,tab
+2,511,"extension-output-pdoom-org.crowd-code-#1-crowd-code",0,0,"7:04:51 PM [info] Activating crowd-code\n7:04:51 PM [info] Recording started\n7:04:51 PM [info] Initializing git provider using file system watchers...\n7:04:52 PM [info] Git repository found\n7:04:52 PM [info] Git provider initialized successfully\n7:04:52 PM [info] Initial git state: [object Object]\n",Log,tab