{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/census/attention-2","entry":"attention","source":"Syntology differential census (groundwork/55, run v2_2026-09-22), per sample; not an archive number","census_date":"2026-09-22","battery_sha256":["b986f7e04d794a0d88ad4c5f32cf63ec3590b5deff0192150737bc5f1c0b4677"],"runner_sha256":["5a452d0e7c0da5b80771d1be2afe3572e253568cf5c5d59a0d08ebd663d00808"],"bucket_key":"positional (rank, kind, dtype) of each array argument; the argument name is not part of the key because the harness draws the shared array from (rank, kind) and casts it to the dtype, whatever the name","bucket_fields":["rank","kind","dtype"],"claim":"Implementations sharing this entry name were each run on one shared input fixed by the positional (rank, kind, dtype) of their array arguments (the bucket). A cluster is the set whose recorded output digest (sha256 of the output rounded to 6 decimals) is identical. Identical values to six decimals on the shared input are agreement on those inputs, not a statement about the whole domain and not a substitution claim.","n_implementations_compared":11,"n_papers":10,"n_buckets":5,"n_distinct_outputs":7,"n_class_bearing":0,"not_compared":{"not_run_on_shared_input":{"n":8,"by_error":{"RuntimeError":8}},"no_array_argument_ran_on_own_fixture_arguments_only":{"n":0,"n_papers":0,"recorded_shared_digest_equals_own_fixture_digest":0},"output_not_digested_non_numeric":{"n":0,"by_type":{}}},"withdrawn_excluded":0,"code_page":"/code/attention-2","buckets":[{"bucket":[[4,"float","float32"],[4,"float","float32"],[4,"float","float32"],[4,"float","float32"]],"n_implementations_compared":4,"n_papers":3,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"c190e723cd2e3a36","size":3,"n_papers":2,"shape":[2,3,4,4],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":0.0,"members_with_recorded_values":3,"torch_reference_conventions_with_this_digest":[],"values":[0.26010873913764954,-0.05008329823613167,-0.12917344272136688,0.1743120551109314,0.5803914070129395,-0.025736596435308456,-0.5494147539138794,0.8076320886611938,-0.009269103407859802,-0.2710424065589905,-1.7586818933486938,0.6880369782447815,-0.6355842351913452,-0.44694390892982483,-2.74556040763855,0.8345890045166016,-0.19776365160942078,0.31216517090797424,0.34990760684013367,-0.9538421034812927,-0.4699624180793762,1.160244107246399,-0.746057391166687,0.8951020240783691,0.08140887320041656,1.2267361879348755,-1.8356263637542725,-0.016265422105789185,-0.10055772960186005,0.35825133323669434,0.3305554986000061,-0.6440742611885071,-0.5083611607551575,1.2879480123519897,-0.09953615069389343,-0.7144991159439087,-0.6142417192459106,1.7646915912628174,0.10005393624305725,-1.1775208711624146,-0.24179387092590332,-1.313307523727417,-0.7121430039405823,1.0143780708312988,-0.3398871421813965,0.7060075998306274,0.03175554424524307,-0.9451687335968018,-0.014198352582752705,-0.034208763390779495,-0.057248834520578384,-0.0014193992828950286,-0.08449332416057587,-0.10709623992443085,-0.06720384955406189,-0.00022957993496675044,-1.8439106941223145,-1.7061400413513184,-1.3216067552566528,0.042170971632003784,-0.3558184802532196,-0.5317538976669312,-0.36872872710227966,-0.008068124763667583,0.8990591168403625,-0.900050163269043,1.183423399925232,1.8030997514724731,0.5815004110336304,-0.00861676037311554,3.178694248199463,0.15824158489704132,0.7853977084159851,0.0305609330534935,1.7106742858886719,0.1551772654056549,0.8184341192245483,-0.1461194008588791,1.7050528526306152,0.5068970322608948,-2.019789695739746,-0.0033594442065805197,0.2687206566333771,-0.9012560844421387,0.6827816367149353,1.102705478668213,-1.5709209442138672,0.840479850769043,-0.008833835832774639,0.43210354447364807,1.8259233236312866,-1.196785807609558,0.5888994932174683,0.6878886222839355,-1.2358767986297607,0.3110019266605377],"values_recorded":96,"members":[{"code_sha256_prefix":"5bd62b8707ef0480","path":"pykt/models/simplekt.py","papers":["2302.06881"],"paper_pages":[{"arxiv_id":"2302.06881","page":"/paper/simplekt-a-simple-but-tough-to-beat-baseline"}],"arg_sig_recorded":[["q",4,"float","float32"],["k",4,"float","float32"],["v",4,"float","float32"],["mask",4,"float","float32"]],"scalar_args":{"d_k":"5","dropout":"Dropout(p=0.1, inplace=False)","zero_pad":"False"},"class_bearing":false},{"code_sha256_prefix":"72814a15da26ba0d","path":"models/pakt.py","papers":["2607.13103"],"paper_pages":[{"arxiv_id":"2607.13103","page":"/paper/arxiv-2607-13103"}],"arg_sig_recorded":[["q",4,"float","float32"],["k",4,"float","float32"],["v",4,"float","float32"],["mask",4,"float","float32"]],"scalar_args":{"d_k":"5","dropout":"Dropout(p=0.1, inplace=False)","zero_pad":"False"},"class_bearing":false},{"code_sha256_prefix":"e05cd8a1686715a5","path":"pykt/models/simplekt.py","papers":["2302.06881"],"paper_pages":[{"arxiv_id":"2302.06881","page":"/paper/simplekt-a-simple-but-tough-to-beat-baseline"}],"arg_sig_recorded":[["q",4,"float","float32"],["k",4,"float","float32"],["v",4,"float","float32"],["mask",4,"float","float32"]],"scalar_args":{"d_k":"5","dropout":"Dropout(p=0.1, inplace=False)","zero_pad":"False"},"class_bearing":false}]},{"output_sha":"ff86505cf7789df7","size":1,"n_papers":1,"shape":[2,3,4,4],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[0.0,0.0,0.0,0.0,0.5803914070129395,-0.025736596435308456,-0.5494147539138794,0.8076320886611938,-0.009269103407859802,-0.2710424065589905,-1.7586818933486938,0.6880369782447815,-0.6355842351913452,-0.44694390892982483,-2.74556040763855,0.8345890045166016,0.0,0.0,0.0,0.0,-0.4699624180793762,1.160244107246399,-0.746057391166687,0.8951020240783691,0.08140887320041656,1.2267361879348755,-1.8356263637542725,-0.016265422105789185,-0.10055772960186005,0.35825133323669434,0.3305554986000061,-0.6440742611885071,0.0,0.0,0.0,0.0,-0.6142417192459106,1.7646915912628174,0.10005393624305725,-1.1775208711624146,-0.24179387092590332,-1.313307523727417,-0.7121430039405823,1.0143780708312988,-0.3398871421813965,0.7060075998306274,0.03175554424524307,-0.9451687335968018,0.0,0.0,0.0,0.0,-0.08449332416057587,-0.10709623992443085,-0.06720384955406189,-0.00022957993496675044,-1.8439106941223145,-1.7061400413513184,-1.3216067552566528,0.042170971632003784,-0.3558184802532196,-0.5317538976669312,-0.36872872710227966,-0.008068124763667583,0.0,0.0,0.0,0.0,0.5815004110336304,-0.00861676037311554,3.178694248199463,0.15824158489704132,0.7853977084159851,0.0305609330534935,1.7106742858886719,0.1551772654056549,0.8184341192245483,-0.1461194008588791,1.7050528526306152,0.5068970322608948,0.0,0.0,0.0,0.0,0.6827816367149353,1.102705478668213,-1.5709209442138672,0.840479850769043,-0.008833835832774639,0.43210354447364807,1.8259233236312866,-1.196785807609558,0.5888994932174683,0.6878886222839355,-1.2358767986297607,0.3110019266605377],"values_recorded":96,"members":[{"code_sha256_prefix":"134e95d193454d9c","path":"model/TransKT.py","papers":["2505.13489"],"paper_pages":[{"arxiv_id":"2505.13489","page":"/paper/contrastive-cross-course-knowledge-tracing"}],"arg_sig_recorded":[["q",4,"float","float32"],["k",4,"float","float32"],["v",4,"float","float32"],["mask",4,"float","float32"]],"scalar_args":{"d_k":"5","dropout":"Dropout(p=0.1, inplace=False)","zero_pad":"True"},"class_bearing":false}]}]},{"bucket":[[4,"float","float32"]],"n_implementations_compared":3,"n_papers":3,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"42c78be7a46f5dfd","size":3,"n_papers":3,"shape":[2,16],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":0.0,"members_with_recorded_values":3,"torch_reference_conventions_with_this_digest":[],"values":[0.06418933719396591,0.10517550259828568,0.18970076739788055,0.2540999948978424,0.15997618436813354,0.37546834349632263,0.019628098234534264,0.2974504232406616,0.015087969601154327,0.2558395564556122,0.4426683783531189,0.10707525163888931,0.045598022639751434,0.015389278531074524,0.5682892203330994,0.17453721165657043,0.27602770924568176,0.058726631104946136,0.20793460309505463,0.20994652807712555,0.1437697857618332,0.07855033874511719,0.7322915196418762,0.050042085349559784,0.09665584564208984,0.17736881971359253,0.24875646829605103,0.07486364245414734,0.3819640576839447,0.026341751217842102,0.11187763512134552,0.007554740644991398],"values_recorded":32,"members":[{"code_sha256_prefix":"b817aa66879a905d","path":"unlearn.py","papers":["2210.08196","2402.01401"],"paper_pages":[{"arxiv_id":"2210.08196","page":"/paper/deep-regression-unlearning"},{"arxiv_id":"2402.01401","page":"/paper/zero-shot-machine-unlearning-at-scale-via"}],"arg_sig_recorded":[["x",4,"float","float32"]],"scalar_args":{},"class_bearing":false},{"code_sha256_prefix":"8031f2b6467987ab","path":"Our_code/Pytorch/zero-shot-baseline.py","papers":["1905.09768"],"paper_pages":[{"arxiv_id":"1905.09768","page":"/paper/zero-shot-knowledge-transfer-via-adversarial"}],"arg_sig_recorded":[["x",4,"float","float32"]],"scalar_args":{},"class_bearing":false},{"code_sha256_prefix":"ab5d18fb42f8dcba","path":"Our_code/Pytorch/Basic_experiments/exp-16_2-16_1.py","papers":["1905.09768"],"paper_pages":[{"arxiv_id":"1905.09768","page":"/paper/zero-shot-knowledge-transfer-via-adversarial"}],"arg_sig_recorded":[["x",4,"float","float32"]],"scalar_args":{},"class_bearing":false}]}]},{"bucket":[[3,"float","float32"],[3,"float","float32"],[3,"float","float32"]],"n_implementations_compared":2,"n_papers":2,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"02617d7c99173618","size":1,"n_papers":1,"shape":[2,4,8],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[-1.003792405128479,1.3409873247146606,-1.1562010049819946,-0.33927449584007263,0.2120620608329773,-1.0800873041152954,-0.2803649306297302,-0.1391511857509613,-1.70661461353302,0.3993523120880127,-0.92539381980896,-0.6252562403678894,0.6017330884933472,-0.3761001229286194,-0.001095667714253068,0.2957557141780853,1.5415898561477661,-0.21601249277591705,-0.3049392104148865,0.4270317852497101,-0.690680742263794,1.125346064567566,-0.5505296587944031,1.2044429779052734,1.0218312740325928,0.6307428479194641,-0.6500711441040039,-0.12126103043556213,-0.8916053175926208,1.131993055343628,-0.7896889448165894,1.6030319929122925,-0.9819821715354919,1.6158303022384644,-0.994032621383667,-0.7865397334098816,1.038595199584961,0.6043587327003479,-0.012222290970385075,1.510730266571045,-1.0698626041412354,0.08678285032510757,-0.7584128975868225,-0.7348361611366272,1.0149993896484375,0.19363665580749512,0.10167430341243744,0.8927397727966309,-0.9612482786178589,0.6482498049736023,0.9754965901374817,-0.7310844659805298,1.3356858491897583,0.766312301158905,-0.8785896897315979,-0.5789462327957153,-1.2190170288085938,1.0972286462783813,1.526685357093811,-0.6073513031005859,1.785226583480835,1.1666909456253052,-1.2037301063537598,-0.7931501865386963],"values_recorded":64,"members":[{"code_sha256_prefix":"72ce5bea9e3abec1","path":"utils.py","papers":["1912.09363"],"paper_pages":[{"arxiv_id":"1912.09363","page":"/paper/temporal-fusion-transformers-for"}],"arg_sig_recorded":[["Q",3,"float","float32"],["K",3,"float","float32"],["V",3,"float","float32"]],"scalar_args":{},"class_bearing":false}]},{"output_sha":"c09ff4f743ddbaa7","size":1,"n_papers":1,"shape":[2,4,8],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[-0.9920259714126587,1.4934321641921997,-1.2068092823028564,-0.31088072061538696,0.20429691672325134,-1.259804606437683,-0.2971417009830475,-0.26462212204933167,-1.9056211709976196,0.2703663408756256,-0.9070405960083008,-0.6861221194267273,0.707936704158783,-0.3374894857406616,0.06391837447881699,0.3084448575973511,1.648512363433838,-0.34308287501335144,-0.24936731159687042,0.5131499171257019,-0.6771886944770813,1.1458274126052856,-0.5233363509178162,1.1631377935409546,1.0305663347244263,0.7445501685142517,-0.6860655546188354,-0.19082430005073547,-0.9688389897346497,1.205674648284912,-0.8486900329589844,1.7216312885284424,-0.9676222205162048,1.7774851322174072,-1.1198465824127197,-0.7972066402435303,1.0136923789978027,0.6203122138977051,0.02690826915204525,1.6647138595581055,-1.1055563688278198,-0.10561081767082214,-0.9081276059150696,-0.7233197093009949,0.9998772740364075,0.10791154950857162,0.21002130210399628,0.9924672842025757,-0.9973194599151611,0.7341638207435608,1.1937379837036133,-0.7096108198165894,1.431776523590088,0.8653138875961304,-0.9997137188911438,-0.7302023768424988,-1.2558900117874146,1.138271450996399,1.6224761009216309,-0.5890276432037354,1.850628137588501,1.2204967737197876,-1.2554773092269897,-0.8460952639579773],"values_recorded":64,"members":[{"code_sha256_prefix":"fe30479834a53868","path":"kandinsky2/model/text2im_model2_1.py","papers":["2310.03502"],"paper_pages":[{"arxiv_id":"2310.03502","page":"/paper/kandinsky-an-improved-text-to-image-synthesis"}],"arg_sig_recorded":[["q",3,"float","float32"],["k",3,"float","float32"],["v",3,"float","float32"]],"scalar_args":{"d_k":"4"},"class_bearing":false}]}]},{"bucket":[[4,"float","float32"],[4,"float","float32"],[4,"float","float32"],[4,"int","int64"]],"n_implementations_compared":1,"n_papers":1,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"cc2e4b612f8d03c3","size":1,"n_papers":1,"shape":[1,1,2,3,4,4],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[0.019363168627023697,-0.007465694565325975,-0.0671406239271164,-0.0007169946329668164,0.6522131562232971,0.00248402520082891,-0.34081730246543884,1.0086586475372314,-0.056063707917928696,-0.2890566289424896,-1.847430944442749,0.6892692446708679,-0.7138468623161316,-0.4617398679256439,-2.853296995162964,0.8484854102134705,-0.1735750436782837,0.22051601111888885,0.45418205857276917,-1.1850268840789795,-0.503600537776947,1.1870826482772827,-0.7472839951515198,0.981485903263092,0.22291244566440582,1.2762486934661865,-2.197171688079834,-0.23685072362422943,-0.5725869536399841,0.4005623161792755,0.40666890144348145,-0.8891136646270752,-0.5234459042549133,1.3597103357315063,-0.08336000889539719,-0.7622954845428467,-0.632891833782196,1.8340222835540771,0.11572378873825073,-1.1996283531188965,-0.23349322378635406,-1.4408787488937378,-0.7520645260810852,1.1275793313980103,-0.3039263188838959,0.6214799284934998,0.19660279154777527,-1.0916013717651367,-0.0073265135288238525,-0.020107494667172432,-0.03464241698384285,-0.0009350451291538775,-1.5342658758163452,-1.9982569217681885,-1.2833296060562134,-0.009194733574986458,-1.8680120706558228,-1.7477593421936035,-1.3257609605789185,0.04111943021416664,0.00725905317813158,-0.0015543875051662326,-0.005628818180412054,-0.0007830492104403675,0.9045149683952332,-0.9572410583496094,1.137986660003662,1.8991022109985352,0.5688660144805908,0.0003985307994298637,3.259523630142212,0.13728663325309753,0.6405546069145203,-0.03705215826630592,2.0687739849090576,0.08705367892980576,0.8147647976875305,-0.14512526988983154,1.7427507638931274,0.5063662528991699,-2.161159038543701,-0.04399705305695534,0.28720805048942566,-0.9415680170059204,0.7107812166213989,1.1427192687988281,-1.604846477508545,0.8992951512336731,0.03170004487037659,0.4327006936073303,1.9521164894104004,-1.2489285469055176,0.6773158311843872,0.7791770696640015,-1.6069872379302979,0.45712193846702576],"values_recorded":96,"members":[{"code_sha256_prefix":"86341ad6b5c2c7a4","path":"models/oclr.py","papers":["2207.02206"],"paper_pages":[{"arxiv_id":"2207.02206","page":"/paper/segmenting-moving-objects-via-an-object"}],"arg_sig_recorded":[["q",4,"float","float32"],["k",4,"float","float32"],["v",4,"float","float32"],["mask",4,"int","int64"]],"scalar_args":{"d_k":"4","dropout":"Dropout(p=0.1, inplace=False)"},"class_bearing":false}]}]},{"bucket":[[4,"float","float32"],[4,"float","float32"],[4,"float","float32"]],"n_implementations_compared":1,"n_papers":1,"n_not_digested":0,"not_digested_by_type":{},"clusters":[{"output_sha":"ac07ca26d031ac82","size":1,"n_papers":1,"shape":[2,3,16],"dtype":"float32","type":"Tensor","finite":true,"max_difference_among_recorded_values":null,"members_with_recorded_values":1,"torch_reference_conventions_with_this_digest":[],"values":[0.5717889070510864,0.644473671913147,1.2623403072357178,0.8300621509552002,0.511626124382019,0.3214063346385956,-0.2673454284667969,0.8332299590110779,0.1596429944038391,0.16677190363407135,-1.4936769008636475,0.19890594482421875,-0.7014915347099304,-0.43326544761657715,-2.637392282485962,0.7970542907714844,-0.32508376240730286,0.3977779150009155,0.4174955487251282,-0.9632585048675537,-0.5121825933456421,1.133176565170288,-0.2890968918800354,0.9585134387016296,0.22594337165355682,0.8902930021286011,-1.8881585597991943,-0.09547661989927292,0.15498901903629303,-0.09611478447914124,0.434932678937912,-0.7482161521911621,-0.2643570005893707,0.6734958291053772,0.15141478180885315,-0.5744507908821106,-0.6379923224449158,1.8833352327346802,0.1574070006608963,-0.966434121131897,-0.09654776751995087,-1.0824912786483765,-0.9065291881561279,0.9309486150741577,0.08471505343914032,-0.1271730363368988,0.2635588049888611,-0.7812329530715942,0.9480071067810059,0.26288238167762756,1.8942724466323853,0.23927812278270721,1.5867286920547485,-0.4913731515407562,-2.3428850173950195,-0.12167798727750778,-1.3685575723648071,-1.983659267425537,-1.3460739850997925,-0.0341484509408474,-2.760305166244507,-0.7289645671844482,-0.6385515332221985,0.2086576521396637,0.8421983122825623,-0.964346706867218,0.8718371391296387,1.9152746200561523,0.4851168394088745,0.022979198023676872,3.0991275310516357,0.07439672946929932,0.2505892813205719,0.209952250123024,1.209070086479187,-0.9072842001914978,0.7614883780479431,0.18981227278709412,0.10817322880029678,0.10937247425317764,-2.115112781524658,-0.08590003848075867,0.21329039335250854,-0.7904897928237915,0.999381959438324,0.6624920964241028,-1.7657887935638428,0.6813942193984985,0.2727597653865814,0.3968070447444916,1.7269551753997803,-1.084537386894226,0.09468081593513489,0.03206448629498482,-0.8885524868965149,-0.11426423490047455],"values_recorded":96,"members":[{"code_sha256_prefix":"7f6ea85164888083","path":"hyvideo/modules/models.py","papers":["2412.03603"],"paper_pages":[{"arxiv_id":"2412.03603","page":"/paper/hunyuanvideo-a-systematic-framework-for-large"}],"arg_sig_recorded":[["q",4,"float","float32"],["k",4,"float","float32"],["v",4,"float","float32"]],"scalar_args":{"mode":"'vanilla'","drop_rate":"0.0","attn_mask":"None","causal":"False","cu_seqlens_q":"None","cu_seqlens_kv":"None","max_seqlen_q":"None","max_seqlen_kv":"None","batch_size":"2"},"class_bearing":false}]}]}]}