61 lines
42 KiB
JSON
61 lines
42 KiB
JSON
{"repeat": 64, "batch_size": 8192, "dim_out": 4096, "dim_in": 1024, "wm": 4, "switch": false, "standard_fwd": 0.28139352798461914, "standard_gw": 0.2811811864376068, "standard_gx": 0.30258670449256897, "rowwise_fwd": 0.1994594931602478, "rowwise_bwd": 0.16159191727638245, "global_fwd": 0.19502267241477966, "global_bwd": 0.16080215573310852, "x_quantize_rowwise": 0.03306940197944641, "g_quantize_rowwise": 0.08210167288780212, "w_quantize_rowwise": 0.03385916352272034, "w_quantize_colwise_transpose": 0.08635595440864563, "w_quantize_global": 0.09237229824066162, "w_quantize_global_transpose": 0.10007619857788086, "time_standard": 0.8651614189147949, "time_rowwise": 0.8776187896728516, "time_global": 0.944625586271286}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 1024, "dim_in": 4096, "wm": 4, "switch": true, "standard_fwd": 0.262625515460968, "standard_gw": 0.2806223928928375, "standard_gx": 0.31118839979171753, "rowwise_fwd": 0.1828707754611969, "rowwise_bwd": 0.21236762404441833, "global_fwd": 0.16665831208229065, "global_bwd": 0.19929558038711548, "x_quantize_rowwise": 0.08227676153182983, "g_quantize_rowwise": 0.03310292959213257, "w_quantize_rowwise": 0.032648444175720215, "w_quantize_colwise_transpose": 0.09015202522277832, "w_quantize_global": 0.0988692045211792, "w_quantize_global_transpose": 0.10057538747787476, "time_standard": 0.8544363081455231, "time_rowwise": 0.9140409529209137, "time_global": 0.96140056848526}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 4096, "dim_in": 1024, "wm": 4, "switch": false, "standard_fwd": 0.5731917917728424, "standard_gw": 0.5709454417228699, "standard_gx": 0.5963630974292755, "rowwise_fwd": 0.37662312388420105, "rowwise_bwd": 0.281747430562973, "global_fwd": 0.36768242716789246, "global_bwd": 0.28043612837791443, "x_quantize_rowwise": 0.046547502279281616, "g_quantize_rowwise": 0.15532970428466797, "w_quantize_rowwise": 0.032436102628707886, "w_quantize_colwise_transpose": 0.08635222911834717, "w_quantize_global": 0.0947415828704834, "w_quantize_global_transpose": 0.10129809379577637, "time_standard": 1.7405003309249878, "time_rowwise": 1.5499815344810486, "time_global": 1.616980880498886}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 1024, "dim_in": 4096, "wm": 4, "switch": true, "standard_fwd": 0.5341619253158569, "standard_gw": 0.5690865218639374, "standard_gx": 0.599835067987442, "rowwise_fwd": 0.3233291208744049, "rowwise_bwd": 0.41359663009643555, "global_fwd": 0.2831108868122101, "global_bwd": 0.37280842661857605, "x_quantize_rowwise": 0.15563145279884338, "g_quantize_rowwise": 0.046741217374801636, "w_quantize_rowwise": 0.03306940197944641, "w_quantize_colwise_transpose": 0.09020790457725525, "w_quantize_global": 0.0925213098526001, "w_quantize_global_transpose": 0.09945780038833618, "time_standard": 1.7030835151672363, "time_rowwise": 1.6316622495651245, "time_global": 1.6193576157093048}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 4096, "dim_in": 1024, "wm": 4, "switch": false, "standard_fwd": 1.2199915945529938, "standard_gw": 1.1069811880588531, "standard_gx": 1.09761580824852, "rowwise_fwd": 0.738043338060379, "rowwise_bwd": 0.5549229681491852, "global_fwd": 0.7219798862934113, "global_bwd": 0.5512163043022156, "x_quantize_rowwise": 0.08748471736907959, "g_quantize_rowwise": 0.3023110330104828, "w_quantize_rowwise": 0.03182142972946167, "w_quantize_colwise_transpose": 0.08632615208625793, "w_quantize_global": 0.09445473551750183, "w_quantize_global_transpose": 0.10032951831817627, "time_standard": 3.424588590860367, "time_rowwise": 2.9078908264636993, "time_global": 2.9647573828697205}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 1024, "dim_in": 4096, "wm": 4, "switch": true, "standard_fwd": 1.1040829122066498, "standard_gw": 1.1221766471862793, "standard_gx": 1.1548101902008057, "rowwise_fwd": 0.581938773393631, "rowwise_bwd": 0.7480122148990631, "global_fwd": 0.5537159740924835, "global_bwd": 0.7232688367366791, "x_quantize_rowwise": 0.30193477869033813, "g_quantize_rowwise": 0.08745118975639343, "w_quantize_rowwise": 0.03374740481376648, "w_quantize_colwise_transpose": 0.09068101644515991, "w_quantize_global": 0.09645149111747742, "w_quantize_global_transpose": 0.10189786553382874, "time_standard": 3.3810697495937347, "time_rowwise": 2.9659420251846313, "time_global": 2.9868967831134796}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 4096, "dim_in": 1024, "wm": 4, "switch": false, "standard_fwd": 2.4533793330192566, "standard_gw": 2.1938569843769073, "standard_gx": 2.179361879825592, "rowwise_fwd": 1.4615543186664581, "rowwise_bwd": 1.0522231459617615, "global_fwd": 1.4288239181041718, "global_bwd": 1.0450035333633423, "x_quantize_rowwise": 0.1691766083240509, "g_quantize_rowwise": 0.5951300263404846, "w_quantize_rowwise": 0.03337860107421875, "w_quantize_colwise_transpose": 0.08653849363327026, "w_quantize_global": 0.0940859317779541, "w_quantize_global_transpose": 0.09976327419281006, "time_standard": 6.826598197221756, "time_rowwise": 5.5918581783771515, "time_global": 5.625840276479721}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 1024, "dim_in": 4096, "wm": 4, "switch": true, "standard_fwd": 2.1698065102100372, "standard_gw": 2.1875128149986267, "standard_gx": 2.2887587547302246, "rowwise_fwd": 1.0762326419353485, "rowwise_bwd": 1.4638006687164307, "global_fwd": 1.0450668632984161, "global_bwd": 1.4308765530586243, "x_quantize_rowwise": 0.5953535437583923, "g_quantize_rowwise": 0.16899779438972473, "w_quantize_rowwise": 0.03240257501602173, "w_quantize_colwise_transpose": 0.09106099605560303, "w_quantize_global": 0.09546056389808655, "w_quantize_global_transpose": 0.09852275252342224, "time_standard": 6.6460780799388885, "time_rowwise": 5.615361034870148, "time_global": 5.621790885925293}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 4096, "dim_in": 1024, "wm": 4, "switch": false, "standard_fwd": 4.858218133449554, "standard_gw": 4.3631307780742645, "standard_gx": 4.404045641422272, "rowwise_fwd": 2.9063820838928223, "rowwise_bwd": 2.094462513923645, "global_fwd": 2.8426870703697205, "global_bwd": 2.0792782306671143, "x_quantize_rowwise": 0.33241137862205505, "g_quantize_rowwise": 1.1817105114459991, "w_quantize_rowwise": 0.03374367952346802, "w_quantize_colwise_transpose": 0.08633732795715332, "w_quantize_global": 0.09231641888618469, "w_quantize_global_transpose": 0.100012868642807, "time_standard": 13.62539455294609, "time_rowwise": 10.998178273439407, "time_global": 10.991547256708145}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 1024, "dim_in": 4096, "wm": 4, "switch": true, "standard_fwd": 4.246581345796585, "standard_gw": 4.42587211728096, "standard_gx": 4.581417888402939, "rowwise_fwd": 2.1114833652973175, "rowwise_bwd": 2.9050447046756744, "global_fwd": 2.0806826651096344, "global_bwd": 2.85966694355011, "x_quantize_rowwise": 1.1816024780273438, "g_quantize_rowwise": 0.33330172300338745, "w_quantize_rowwise": 0.033445656299591064, "w_quantize_colwise_transpose": 0.09065866470336914, "w_quantize_global": 0.09239837527275085, "w_quantize_global_transpose": 0.09984523057937622, "time_standard": 13.253871351480484, "time_rowwise": 11.081408709287643, "time_global": 11.073369532823563}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 5120, "dim_in": 1280, "wm": 4, "switch": false, "standard_fwd": 0.4859529435634613, "standard_gw": 0.46338513493537903, "standard_gx": 0.42321905493736267, "rowwise_fwd": 0.2761557698249817, "rowwise_bwd": 0.20775198936462402, "global_fwd": 0.2713911235332489, "global_bwd": 0.20639970898628235, "x_quantize_rowwise": 0.033095479011535645, "g_quantize_rowwise": 0.11894106864929199, "w_quantize_rowwise": 0.03125518560409546, "w_quantize_colwise_transpose": 0.1424551010131836, "w_quantize_global": 0.07288157939910889, "w_quantize_global_transpose": 0.08071959018707275, "time_standard": 1.372557133436203, "time_rowwise": 1.2730397284030914, "time_global": 1.2468136847019196}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 1280, "dim_in": 5120, "wm": 4, "switch": true, "standard_fwd": 0.3920421004295349, "standard_gw": 0.44424086809158325, "standard_gx": 0.4759356379508972, "rowwise_fwd": 0.23231282830238342, "rowwise_bwd": 0.28430670499801636, "global_fwd": 0.20883232355117798, "global_bwd": 0.2741999924182892, "x_quantize_rowwise": 0.12018159031867981, "g_quantize_rowwise": 0.03195926547050476, "w_quantize_rowwise": 0.026017427444458008, "w_quantize_colwise_transpose": 0.14733895659446716, "w_quantize_global": 0.07734447717666626, "w_quantize_global_transpose": 0.0788569450378418, "time_standard": 1.3122186064720154, "time_rowwise": 1.2863576412200928, "time_global": 1.235615462064743}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 5120, "dim_in": 1280, "wm": 4, "switch": false, "standard_fwd": 1.0111741721630096, "standard_gw": 0.9267590939998627, "standard_gx": 0.8254274725914001, "rowwise_fwd": 0.5434826016426086, "rowwise_bwd": 0.4077926278114319, "global_fwd": 0.5318708717823029, "global_bwd": 0.40537863969802856, "x_quantize_rowwise": 0.059738755226135254, "g_quantize_rowwise": 0.2299174666404724, "w_quantize_rowwise": 0.02545863389968872, "w_quantize_colwise_transpose": 0.14269724488258362, "w_quantize_global": 0.07300823926925659, "w_quantize_global_transpose": 0.07878988981246948, "time_standard": 2.7633607387542725, "time_rowwise": 2.335846424102783, "time_global": 2.305462956428528}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 1280, "dim_in": 5120, "wm": 4, "switch": true, "standard_fwd": 0.8095316588878632, "standard_gw": 0.8607134222984314, "standard_gx": 0.9204968810081482, "rowwise_fwd": 0.4275888204574585, "rowwise_bwd": 0.5485899746417999, "global_fwd": 0.41000545024871826, "global_bwd": 0.5317628383636475, "x_quantize_rowwise": 0.2301819622516632, "g_quantize_rowwise": 0.059254467487335205, "w_quantize_rowwise": 0.02466142177581787, "w_quantize_colwise_transpose": 0.14865398406982422, "w_quantize_global": 0.07582828402519226, "w_quantize_global_transpose": 0.08231401443481445, "time_standard": 2.5907419621944427, "time_rowwise": 2.2996440529823303, "time_global": 2.2500604391098022}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 5120, "dim_in": 1280, "wm": 4, "switch": false, "standard_fwd": 2.0658522844314575, "standard_gw": 1.718364655971527, "standard_gx": 1.6660578548908234, "rowwise_fwd": 1.066897064447403, "rowwise_bwd": 0.8070804178714752, "global_fwd": 1.0473169386386871, "global_bwd": 0.8021742105484009, "x_quantize_rowwise": 0.11274218559265137, "g_quantize_rowwise": 0.4518181085586548, "w_quantize_rowwise": 0.026501715183258057, "w_quantize_colwise_transpose": 0.14259666204452515, "w_quantize_global": 0.07484853267669678, "w_quantize_global_transpose": 0.07976219058036804, "time_standard": 5.450274795293808, "time_rowwise": 4.326000809669495, "time_global": 4.287026822566986}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 1280, "dim_in": 5120, "wm": 4, "switch": true, "standard_fwd": 2.7549192309379578, "standard_gw": 1.6954988241195679, "standard_gx": 1.8179528415203094, "rowwise_fwd": 0.8649080991744995, "rowwise_bwd": 1.0746456682682037, "global_fwd": 0.8023083209991455, "global_bwd": 1.0471977293491364, "x_quantize_rowwise": 0.45225024223327637, "g_quantize_rowwise": 0.11286512017250061, "w_quantize_rowwise": 0.0252649188041687, "w_quantize_colwise_transpose": 0.14732033014297485, "w_quantize_global": 0.07537379860877991, "w_quantize_global_transpose": 0.0807642936706543, "time_standard": 6.268370896577835, "time_rowwise": 4.372753202915192, "time_global": 4.266258329153061}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 5120, "dim_in": 1280, "wm": 4, "switch": false, "standard_fwd": 4.098430275917053, "standard_gw": 3.3501461148262024, "standard_gx": 5.560480058193207, "rowwise_fwd": 2.112947404384613, "rowwise_bwd": 1.605246216058731, "global_fwd": 2.0697638392448425, "global_bwd": 1.5953518450260162, "x_quantize_rowwise": 0.21921470761299133, "g_quantize_rowwise": 0.8956789970397949, "w_quantize_rowwise": 0.02710893750190735, "w_quantize_colwise_transpose": 0.14268234372138977, "w_quantize_global": 0.07259473204612732, "w_quantize_global_transpose": 0.07899105548858643, "time_standard": 13.009056448936462, "time_rowwise": 8.35302472114563, "time_global": 8.281741291284561}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 1280, "dim_in": 5120, "wm": 4, "switch": true, "standard_fwd": 5.586959421634674, "standard_gw": 3.358360379934311, "standard_gx": 3.6434978246688843, "rowwise_fwd": 1.6269534826278687, "rowwise_bwd": 2.128206193447113, "global_fwd": 1.5950687229633331, "global_bwd": 2.0831897854804993, "x_quantize_rowwise": 0.8954145014286041, "g_quantize_rowwise": 0.21914392709732056, "w_quantize_rowwise": 0.026203691959381104, "w_quantize_colwise_transpose": 0.14658644795417786, "w_quantize_global": 0.07478520274162292, "w_quantize_global_transpose": 0.07964670658111572, "time_standard": 12.58881762623787, "time_rowwise": 8.400868624448776, "time_global": 8.305609226226807}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 5120, "dim_in": 1280, "wm": 4, "switch": false, "standard_fwd": 8.229725062847137, "standard_gw": 6.791356950998306, "standard_gx": 6.806455552577972, "rowwise_fwd": 4.252471029758453, "rowwise_bwd": 3.2062679529190063, "global_fwd": 4.175614565610886, "global_bwd": 3.1837262213230133, "x_quantize_rowwise": 0.4321373999118805, "g_quantize_rowwise": 1.787092536687851, "w_quantize_rowwise": 0.0270158052444458, "w_quantize_colwise_transpose": 0.1424252986907959, "w_quantize_global": 0.07348507642745972, "w_quantize_global_transpose": 0.07829815149307251, "time_standard": 21.827537566423416, "time_rowwise": 16.63876697421074, "time_global": 16.52171090245247}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 1280, "dim_in": 5120, "wm": 4, "switch": true, "standard_fwd": 11.279478669166565, "standard_gw": 6.7345499992370605, "standard_gx": 7.206875830888748, "rowwise_fwd": 3.209315240383148, "rowwise_bwd": 4.256397485733032, "global_fwd": 3.180190920829773, "global_bwd": 4.177983850240707, "x_quantize_rowwise": 1.7836056649684906, "g_quantize_rowwise": 0.4321075975894928, "w_quantize_rowwise": 0.03205239772796631, "w_quantize_colwise_transpose": 0.14675036072731018, "w_quantize_global": 0.09316205978393555, "w_quantize_global_transpose": 0.10086596012115479, "time_standard": 25.220904499292374, "time_rowwise": 16.5947787463665, "time_global": 16.502466052770615}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 5632, "dim_in": 1408, "wm": 4, "switch": false, "standard_fwd": 0.5776733160018921, "standard_gw": 0.5300231277942657, "standard_gx": 0.6005913019180298, "rowwise_fwd": 0.33330172300338745, "rowwise_bwd": 0.2957060933113098, "global_fwd": 0.32876431941986084, "global_bwd": 0.29108673334121704, "x_quantize_rowwise": 0.03466755151748657, "g_quantize_rowwise": 0.12264400720596313, "w_quantize_rowwise": 0.033874064683914185, "w_quantize_colwise_transpose": 0.1775398850440979, "w_quantize_global": 0.09503215551376343, "w_quantize_global_transpose": 0.10617449879646301, "time_standard": 1.7082877457141876, "time_rowwise": 1.5277564525604248, "time_global": 1.5083923935890198}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 1408, "dim_in": 5632, "wm": 4, "switch": true, "standard_fwd": 0.5164109170436859, "standard_gw": 0.5367249250411987, "standard_gx": 0.5876161158084869, "rowwise_fwd": 0.3132447600364685, "rowwise_bwd": 0.3396235406398773, "global_fwd": 0.2943649888038635, "global_bwd": 0.33209100365638733, "x_quantize_rowwise": 0.12357160449028015, "g_quantize_rowwise": 0.035997480154037476, "w_quantize_rowwise": 0.03213062882423401, "w_quantize_colwise_transpose": 0.17676874995231628, "w_quantize_global": 0.09861215949058533, "w_quantize_global_transpose": 0.0998862087726593, "time_standard": 1.6407519578933716, "time_rowwise": 1.5580616891384125, "time_global": 1.5212483704090118}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 5632, "dim_in": 1408, "wm": 4, "switch": false, "standard_fwd": 1.2096501886844635, "standard_gw": 1.0663382709026337, "standard_gx": 1.0961703956127167, "rowwise_fwd": 0.6396733224391937, "rowwise_bwd": 0.5173943936824799, "global_fwd": 0.6296299397945404, "global_bwd": 0.5130060017108917, "x_quantize_rowwise": 0.06211921572685242, "g_quantize_rowwise": 0.2361498773097992, "w_quantize_rowwise": 0.03260001540184021, "w_quantize_colwise_transpose": 0.17679482698440552, "w_quantize_global": 0.09361281991004944, "w_quantize_global_transpose": 0.09913742542266846, "time_standard": 3.372158855199814, "time_rowwise": 2.7310699224472046, "time_global": 2.6999935507774353}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 1408, "dim_in": 5632, "wm": 4, "switch": true, "standard_fwd": 1.1065565049648285, "standard_gw": 1.0664314031600952, "standard_gx": 1.1266544461250305, "rowwise_fwd": 0.5352050065994263, "rowwise_bwd": 0.6464086472988129, "global_fwd": 0.513765960931778, "global_bwd": 0.6284862756729126, "x_quantize_rowwise": 0.23620948195457458, "g_quantize_rowwise": 0.062271952629089355, "w_quantize_rowwise": 0.031460076570510864, "w_quantize_colwise_transpose": 0.17675384879112244, "w_quantize_global": 0.09486451745033264, "w_quantize_global_transpose": 0.09898096323013306, "time_standard": 3.2996423542499542, "time_rowwise": 2.7547404170036316, "time_global": 2.7010105550289154}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 5632, "dim_in": 1408, "wm": 4, "switch": false, "standard_fwd": 2.4367496371269226, "standard_gw": 2.0806193351745605, "standard_gx": 2.19624862074852, "rowwise_fwd": 1.2554042041301727, "rowwise_bwd": 1.0227933526039124, "global_fwd": 1.2322552502155304, "global_bwd": 1.0152235627174377, "x_quantize_rowwise": 0.11792033910751343, "g_quantize_rowwise": 0.4639364778995514, "w_quantize_rowwise": 0.03241002559661865, "w_quantize_colwise_transpose": 0.17657503485679626, "w_quantize_global": 0.09655207395553589, "w_quantize_global_transpose": 0.09958073496818542, "time_standard": 6.713617593050003, "time_rowwise": 5.149658769369125, "time_global": 5.106087774038315}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 1408, "dim_in": 5632, "wm": 4, "switch": true, "standard_fwd": 2.1935217082500458, "standard_gw": 2.0055584609508514, "standard_gx": 2.1882541477680206, "rowwise_fwd": 1.0396353900432587, "rowwise_bwd": 1.2542344629764557, "global_fwd": 1.0161921381950378, "global_bwd": 1.233428716659546, "x_quantize_rowwise": 0.4642195999622345, "g_quantize_rowwise": 0.11782720685005188, "w_quantize_rowwise": 0.033117830753326416, "w_quantize_colwise_transpose": 0.17696991562843323, "w_quantize_global": 0.09416043758392334, "w_quantize_global_transpose": 0.10101497173309326, "time_standard": 6.387334316968918, "time_rowwise": 5.091562867164612, "time_global": 5.032401531934738}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 5632, "dim_in": 1408, "wm": 4, "switch": false, "standard_fwd": 4.804681986570358, "standard_gw": 4.763372242450714, "standard_gx": 4.064023494720459, "rowwise_fwd": 2.484843134880066, "rowwise_bwd": 1.9691288471221924, "global_fwd": 2.441786229610443, "global_bwd": 1.9574686884880066, "x_quantize_rowwise": 0.2294592559337616, "g_quantize_rowwise": 0.9196549654006958, "w_quantize_rowwise": 0.0313781201839447, "w_quantize_colwise_transpose": 0.1768544316291809, "w_quantize_global": 0.09644776582717896, "w_quantize_global_transpose": 0.09847059845924377, "time_standard": 13.632077723741531, "time_rowwise": 10.574690997600555, "time_global": 10.506659746170044}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 1408, "dim_in": 5632, "wm": 4, "switch": true, "standard_fwd": 4.0907710790634155, "standard_gw": 3.9793066680431366, "standard_gx": 4.302978515625, "rowwise_fwd": 1.992940902709961, "rowwise_bwd": 2.4996213614940643, "global_fwd": 1.9551962614059448, "global_bwd": 2.457551658153534, "x_quantize_rowwise": 0.9200014173984528, "g_quantize_rowwise": 0.2293996512889862, "w_quantize_rowwise": 0.0313781201839447, "w_quantize_colwise_transpose": 0.17882883548736572, "w_quantize_global": 0.09540095925331116, "w_quantize_global_transpose": 0.09880587458610535, "time_standard": 12.373056262731552, "time_rowwise": 9.831476956605911, "time_global": 9.73566249012947}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 5632, "dim_in": 1408, "wm": 4, "switch": false, "standard_fwd": 9.655728936195374, "standard_gw": 8.261296898126602, "standard_gx": 8.064884692430496, "rowwise_fwd": 5.007706582546234, "rowwise_bwd": 3.8615092635154724, "global_fwd": 4.920527338981628, "global_bwd": 3.8330331444740295, "x_quantize_rowwise": 0.45276060700416565, "g_quantize_rowwise": 1.8306002020835876, "w_quantize_rowwise": 0.031366944313049316, "w_quantize_colwise_transpose": 0.1766495406627655, "w_quantize_global": 0.09412690997123718, "w_quantize_global_transpose": 0.09780004620552063, "time_standard": 25.981910526752472, "time_rowwise": 19.621890038251877, "time_global": 19.49014514684677}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 1408, "dim_in": 5632, "wm": 4, "switch": true, "standard_fwd": 8.033104240894318, "standard_gw": 8.2889124751091, "standard_gx": 8.622754365205765, "rowwise_fwd": 3.8747042417526245, "rowwise_bwd": 5.003921687602997, "global_fwd": 3.8315393030643463, "global_bwd": 4.9162134528160095, "x_quantize_rowwise": 1.8304847180843353, "g_quantize_rowwise": 0.4522763192653656, "w_quantize_rowwise": 0.03413110971450806, "w_quantize_colwise_transpose": 0.1771189272403717, "w_quantize_global": 0.09519979357719421, "w_quantize_global_transpose": 0.09930506348609924, "time_standard": 24.944771081209183, "time_rowwise": 19.661549478769302, "time_global": 19.51393112540245}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 6656, "dim_in": 1664, "wm": 4, "switch": false, "standard_fwd": 0.7954612374305725, "standard_gw": 0.7456131279468536, "standard_gx": 0.8799619972705841, "rowwise_fwd": 0.43267011642456055, "rowwise_bwd": 0.34622475504875183, "global_fwd": 0.42615458369255066, "global_bwd": 0.344250351190567, "x_quantize_rowwise": 0.03748014569282532, "g_quantize_rowwise": 0.13304129242897034, "w_quantize_rowwise": 0.03294646739959717, "w_quantize_colwise_transpose": 0.2407953143119812, "w_quantize_global": 0.094633549451828, "w_quantize_global_transpose": 0.10305643081665039, "time_standard": 2.4210363626480103, "time_rowwise": 1.96877121925354, "time_global": 1.8842294812202454}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 1664, "dim_in": 6656, "wm": 4, "switch": true, "standard_fwd": 0.7120333611965179, "standard_gw": 0.7622130215167999, "standard_gx": 0.8262209594249725, "rowwise_fwd": 0.3702230751514435, "rowwise_bwd": 0.4419572651386261, "global_fwd": 0.3479123115539551, "global_bwd": 0.4306286573410034, "x_quantize_rowwise": 0.13308599591255188, "g_quantize_rowwise": 0.037495046854019165, "w_quantize_rowwise": 0.03398209810256958, "w_quantize_colwise_transpose": 0.23782625794410706, "w_quantize_global": 0.09853765368461609, "w_quantize_global_transpose": 0.10247156023979187, "time_standard": 2.3004673421382904, "time_rowwise": 2.016782760620117, "time_global": 1.9123442471027374}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 6656, "dim_in": 1664, "wm": 4, "switch": false, "standard_fwd": 1.6292817890644073, "standard_gw": 1.5109702944755554, "standard_gx": 1.482747495174408, "rowwise_fwd": 0.8386112749576569, "rowwise_bwd": 0.6844550371170044, "global_fwd": 0.8220970630645752, "global_bwd": 0.6802082061767578, "x_quantize_rowwise": 0.06883963942527771, "g_quantize_rowwise": 0.25641173124313354, "w_quantize_rowwise": 0.033054500818252563, "w_quantize_colwise_transpose": 0.24027004837989807, "w_quantize_global": 0.0967271625995636, "w_quantize_global_transpose": 0.102948397397995, "time_standard": 4.622999578714371, "time_rowwise": 3.6326125264167786, "time_global": 3.5382024943828583}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 1664, "dim_in": 6656, "wm": 4, "switch": true, "standard_fwd": 1.4877021312713623, "standard_gw": 1.5015341341495514, "standard_gx": 1.529306173324585, "rowwise_fwd": 0.715944916009903, "rowwise_bwd": 0.8529908955097198, "global_fwd": 0.680088996887207, "global_bwd": 0.8224695920944214, "x_quantize_rowwise": 0.2568177878856659, "g_quantize_rowwise": 0.06864592432975769, "w_quantize_rowwise": 0.03343448042869568, "w_quantize_colwise_transpose": 0.23645907640457153, "w_quantize_global": 0.09399279952049255, "w_quantize_global_transpose": 0.10286271572113037, "time_standard": 4.518542438745499, "time_rowwise": 3.665827214717865, "time_global": 3.5264119505882263}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 6656, "dim_in": 1664, "wm": 4, "switch": false, "standard_fwd": 3.261040896177292, "standard_gw": 2.8816498816013336, "standard_gx": 2.8357282280921936, "rowwise_fwd": 1.6594752669334412, "rowwise_bwd": 1.359265297651291, "global_fwd": 1.6287527978420258, "global_bwd": 1.3503879308700562, "x_quantize_rowwise": 0.13146549463272095, "g_quantize_rowwise": 0.5035959184169769, "w_quantize_rowwise": 0.03438442945480347, "w_quantize_colwise_transpose": 0.24086236953735352, "w_quantize_global": 0.0945068895816803, "w_quantize_global_transpose": 0.10332837700843811, "time_standard": 8.978419005870819, "time_rowwise": 6.8106986582279205, "time_global": 6.693687289953232}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 1664, "dim_in": 6656, "wm": 4, "switch": true, "standard_fwd": 2.848360687494278, "standard_gw": 2.8955675661563873, "standard_gx": 3.0499882996082306, "rowwise_fwd": 1.3900883495807648, "rowwise_bwd": 1.6595833003520966, "global_fwd": 1.3514049351215363, "global_bwd": 1.629263162612915, "x_quantize_rowwise": 0.5036592483520508, "g_quantize_rowwise": 0.13118237257003784, "w_quantize_rowwise": 0.03438442945480347, "w_quantize_colwise_transpose": 0.23709610104560852, "w_quantize_global": 0.0951625406742096, "w_quantize_global_transpose": 0.10216236114501953, "time_standard": 8.793916553258896, "time_rowwise": 6.851561367511749, "time_global": 6.708402186632156}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 6656, "dim_in": 1664, "wm": 4, "switch": false, "standard_fwd": 6.4978525042533875, "standard_gw": 6.462603807449341, "standard_gx": 5.5987648665905, "rowwise_fwd": 3.2996535301208496, "rowwise_bwd": 2.6320070028305054, "global_fwd": 3.2426007091999054, "global_bwd": 2.612769603729248, "x_quantize_rowwise": 0.2561397850513458, "g_quantize_rowwise": 0.9984448552131653, "w_quantize_rowwise": 0.033076852560043335, "w_quantize_colwise_transpose": 0.24232640862464905, "w_quantize_global": 0.09618699550628662, "w_quantize_global_transpose": 0.10257214307785034, "time_standard": 18.559221178293228, "time_rowwise": 13.9242522418499, "time_global": 13.771317899227142}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 1664, "dim_in": 6656, "wm": 4, "switch": true, "standard_fwd": 5.5702440440654755, "standard_gw": 5.717620253562927, "standard_gx": 6.08203187584877, "rowwise_fwd": 2.649586647748947, "rowwise_bwd": 3.315173089504242, "global_fwd": 2.6132799685001373, "global_bwd": 3.257807344198227, "x_quantize_rowwise": 0.9980201721191406, "g_quantize_rowwise": 0.256560742855072, "w_quantize_rowwise": 0.03356859087944031, "w_quantize_colwise_transpose": 0.23729726672172546, "w_quantize_global": 0.09495764970779419, "w_quantize_global_transpose": 0.103779137134552, "time_standard": 17.369896173477173, "time_rowwise": 13.207826763391495, "time_global": 13.04202526807785}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 6656, "dim_in": 1664, "wm": 4, "switch": false, "standard_fwd": 13.058379292488098, "standard_gw": 11.480242013931274, "standard_gx": 11.092845350503922, "rowwise_fwd": 6.637874990701675, "rowwise_bwd": 5.24790957570076, "global_fwd": 6.521012634038925, "global_bwd": 5.214303731918335, "x_quantize_rowwise": 0.5057565867900848, "g_quantize_rowwise": 1.989319920539856, "w_quantize_rowwise": 0.03439188003540039, "w_quantize_colwise_transpose": 0.24280324578285217, "w_quantize_global": 0.09520724415779114, "w_quantize_global_transpose": 0.10240450501441956, "time_standard": 35.631466656923294, "time_rowwise": 26.138298213481903, "time_global": 25.908246636390686}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 1664, "dim_in": 6656, "wm": 4, "switch": true, "standard_fwd": 11.13397628068924, "standard_gw": 11.371888220310211, "standard_gx": 12.12756335735321, "rowwise_fwd": 5.2495077252388, "rowwise_bwd": 6.638709455728531, "global_fwd": 5.215313285589218, "global_bwd": 6.5222084522247314, "x_quantize_rowwise": 1.9870512187480927, "g_quantize_rowwise": 0.5058236420154572, "w_quantize_rowwise": 0.034634023904800415, "w_quantize_colwise_transpose": 0.23674964904785156, "w_quantize_global": 0.09457767009735107, "w_quantize_global_transpose": 0.10183081030845642, "time_standard": 34.63342785835266, "time_rowwise": 26.024363934993744, "time_global": 25.798693299293518}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 8192, "dim_in": 2048, "wm": 4, "switch": false, "standard_fwd": 1.2125298380851746, "standard_gw": 1.1111274361610413, "standard_gx": 1.0840706527233124, "rowwise_fwd": 0.6057210266590118, "rowwise_bwd": 0.51865354180336, "global_fwd": 0.5952082574367523, "global_bwd": 0.5167685449123383, "x_quantize_rowwise": 0.045686960220336914, "g_quantize_rowwise": 0.15827640891075134, "w_quantize_rowwise": 0.04361197352409363, "w_quantize_colwise_transpose": 0.34067779779434204, "w_quantize_global": 0.13644620776176453, "w_quantize_global_transpose": 0.14925003051757812, "time_standard": 3.407727926969528, "time_rowwise": 2.823755145072937, "time_global": 2.7127638459205627}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 2048, "dim_in": 8192, "wm": 4, "switch": true, "standard_fwd": 1.0731369256973267, "standard_gw": 1.1365897953510284, "standard_gx": 1.1498592793941498, "rowwise_fwd": 0.5573518574237823, "rowwise_bwd": 0.615488737821579, "global_fwd": 0.5220361053943634, "global_bwd": 0.5939789116382599, "x_quantize_rowwise": 0.15765801072120667, "g_quantize_rowwise": 0.04369020462036133, "w_quantize_rowwise": 0.047359615564346313, "w_quantize_colwise_transpose": 0.5526281893253326, "w_quantize_global": 0.13606995344161987, "w_quantize_global_transpose": 0.15017390251159668, "time_standard": 3.359586000442505, "time_rowwise": 3.1107664108276367, "time_global": 2.7401968836784363}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 8192, "dim_in": 2048, "wm": 4, "switch": false, "standard_fwd": 2.4274885654449463, "standard_gw": 2.1799951791763306, "standard_gx": 2.1426528692245483, "rowwise_fwd": 1.195710152387619, "rowwise_bwd": 1.027170568704605, "global_fwd": 1.1747106909751892, "global_bwd": 1.0251589119434357, "x_quantize_rowwise": 0.08098781108856201, "g_quantize_rowwise": 0.3052949905395508, "w_quantize_rowwise": 0.043764710426330566, "w_quantize_colwise_transpose": 0.33987686038017273, "w_quantize_global": 0.13646483421325684, "w_quantize_global_transpose": 0.14739856123924255, "time_standard": 6.750136613845825, "time_rowwise": 5.172800272703171, "time_global": 5.050010979175568}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 2048, "dim_in": 8192, "wm": 4, "switch": true, "standard_fwd": 2.1661892533302307, "standard_gw": 2.0948275923728943, "standard_gx": 2.306375652551651, "rowwise_fwd": 1.0587647557258606, "rowwise_bwd": 1.1999905109405518, "global_fwd": 1.0296404361724854, "global_bwd": 1.1749230325222015, "x_quantize_rowwise": 0.3054030239582062, "g_quantize_rowwise": 0.08077546954154968, "w_quantize_rowwise": 0.047225505113601685, "w_quantize_colwise_transpose": 0.600133091211319, "w_quantize_global": 0.13613328337669373, "w_quantize_global_transpose": 0.1484006643295288, "time_standard": 6.567392498254776, "time_rowwise": 5.387119948863983, "time_global": 4.97010350227356}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 8192, "dim_in": 2048, "wm": 4, "switch": false, "standard_fwd": 4.807606339454651, "standard_gw": 4.170913249254227, "standard_gx": 4.117622971534729, "rowwise_fwd": 2.370934933423996, "rowwise_bwd": 1.9481778144836426, "global_fwd": 2.3383721709251404, "global_bwd": 1.9443817436695099, "x_quantize_rowwise": 0.1547597348690033, "g_quantize_rowwise": 0.6000511348247528, "w_quantize_rowwise": 0.04361942410469055, "w_quantize_colwise_transpose": 0.3403201699256897, "w_quantize_global": 0.13600289821624756, "w_quantize_global_transpose": 0.1474134624004364, "time_standard": 13.096142560243607, "time_rowwise": 9.628776460886002, "time_global": 9.491894394159317}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 2048, "dim_in": 8192, "wm": 4, "switch": true, "standard_fwd": 4.1619837284088135, "standard_gw": 4.181284457445145, "standard_gx": 4.635505378246307, "rowwise_fwd": 1.9684135913848877, "rowwise_bwd": 2.3750364780426025, "global_fwd": 1.9445866346359253, "global_bwd": 2.3551955819129944, "x_quantize_rowwise": 0.6004162132740021, "g_quantize_rowwise": 0.15468522906303406, "w_quantize_rowwise": 0.04730746150016785, "w_quantize_colwise_transpose": 0.5999617278575897, "w_quantize_global": 0.1364201307296753, "w_quantize_global_transpose": 0.14847144484519958, "time_standard": 12.978773564100266, "time_rowwise": 9.927105158567429, "time_global": 9.521059691905975}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 8192, "dim_in": 2048, "wm": 4, "switch": false, "standard_fwd": 9.52371209859848, "standard_gw": 8.354485034942627, "standard_gx": 8.69860127568245, "rowwise_fwd": 4.717472940683365, "rowwise_bwd": 3.8843750953674316, "global_fwd": 4.645414650440216, "global_bwd": 3.8761012256145477, "x_quantize_rowwise": 0.3024861216545105, "g_quantize_rowwise": 1.1897757649421692, "w_quantize_rowwise": 0.04366785287857056, "w_quantize_colwise_transpose": 0.33988431096076965, "w_quantize_global": 0.1359507441520691, "w_quantize_global_transpose": 0.14724582433700562, "time_standard": 26.576798409223557, "time_rowwise": 18.832147121429443, "time_global": 18.651459366083145}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 2048, "dim_in": 8192, "wm": 4, "switch": true, "standard_fwd": 8.307881653308868, "standard_gw": 8.214320987462997, "standard_gx": 9.21182706952095, "rowwise_fwd": 3.8919784128665924, "rowwise_bwd": 4.72346693277359, "global_fwd": 3.8761794567108154, "global_bwd": 4.673641175031662, "x_quantize_rowwise": 1.1893920600414276, "g_quantize_rowwise": 0.3024972975254059, "w_quantize_rowwise": 0.04708021879196167, "w_quantize_colwise_transpose": 0.6039328873157501, "w_quantize_global": 0.13624504208564758, "w_quantize_global_transpose": 0.14867261052131653, "time_standard": 25.734029710292816, "time_rowwise": 18.972668796777725, "time_global": 18.540948629379272}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 8192, "dim_in": 2048, "wm": 4, "switch": false, "standard_fwd": 19.30372044444084, "standard_gw": 16.480475664138794, "standard_gx": 17.61433482170105, "rowwise_fwd": 9.49602946639061, "rowwise_bwd": 7.768530398607254, "global_fwd": 9.3533955514431, "global_bwd": 7.749464362859726, "x_quantize_rowwise": 0.5977451801300049, "g_quantize_rowwise": 2.3684948682785034, "w_quantize_rowwise": 0.04375725984573364, "w_quantize_colwise_transpose": 0.34042075276374817, "w_quantize_global": 0.13628974556922913, "w_quantize_global_transpose": 0.14671683311462402, "time_standard": 53.398530930280685, "time_rowwise": 37.09545359015465, "time_global": 36.83258220553398}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 2048, "dim_in": 8192, "wm": 4, "switch": true, "standard_fwd": 18.041003495454788, "standard_gw": 17.770148813724518, "standard_gx": 17.70009845495224, "rowwise_fwd": 7.756810635328293, "rowwise_bwd": 9.502101689577103, "global_fwd": 7.7384114265441895, "global_bwd": 9.36170294880867, "x_quantize_rowwise": 2.3686252534389496, "g_quantize_rowwise": 0.5980581045150757, "w_quantize_rowwise": 0.04723668098449707, "w_quantize_colwise_transpose": 0.6035342812538147, "w_quantize_global": 0.13603642582893372, "w_quantize_global_transpose": 0.1485198736190796, "time_standard": 53.511250764131546, "time_rowwise": 38.64651545882225, "time_global": 38.121502846479416}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 16384, "dim_in": 4096, "wm": 4, "switch": false, "standard_fwd": 4.598241299390793, "standard_gw": 4.294309765100479, "standard_gx": 4.261095076799393, "rowwise_fwd": 2.0976848900318146, "rowwise_bwd": 1.9718967378139496, "global_fwd": 2.0763762295246124, "global_bwd": 1.9703581929206848, "x_quantize_rowwise": 0.08216872811317444, "g_quantize_rowwise": 0.4405900835990906, "w_quantize_rowwise": 0.1553371548652649, "w_quantize_colwise_transpose": 1.6110725700855255, "w_quantize_global": 0.481240451335907, "w_quantize_global_transpose": 0.5061514675617218, "time_standard": 13.153646141290665, "time_rowwise": 10.653059929609299, "time_global": 9.85119491815567}
|
|
{"repeat": 64, "batch_size": 8192, "dim_out": 4096, "dim_in": 16384, "wm": 4, "switch": true, "standard_fwd": 4.35885414481163, "standard_gw": 4.29583340883255, "standard_gx": 4.5370906591415405, "rowwise_fwd": 2.0015686750411987, "rowwise_bwd": 2.097565680742264, "global_fwd": 1.969795674085617, "global_bwd": 2.075403928756714, "x_quantize_rowwise": 0.43984130024909973, "g_quantize_rowwise": 0.08216127753257751, "w_quantize_rowwise": 0.22544339299201965, "w_quantize_colwise_transpose": 2.4342015385627747, "w_quantize_global": 0.48087164759635925, "w_quantize_global_transpose": 0.5099289119243622, "time_standard": 13.19177821278572, "time_rowwise": 11.576615273952484, "time_global": 9.85383614897728}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 16384, "dim_in": 4096, "wm": 4, "switch": false, "standard_fwd": 9.09888744354248, "standard_gw": 8.230950683355331, "standard_gx": 8.465446531772614, "rowwise_fwd": 4.182614386081696, "rowwise_bwd": 3.747660666704178, "global_fwd": 4.138719290494919, "global_bwd": 3.74777615070343, "x_quantize_rowwise": 0.15515834093093872, "g_quantize_rowwise": 0.8699297904968262, "w_quantize_rowwise": 0.15544891357421875, "w_quantize_colwise_transpose": 1.6132444143295288, "w_quantize_global": 0.48100948333740234, "w_quantize_global_transpose": 0.5051903426647186, "time_standard": 25.795284658670425, "time_rowwise": 18.955007195472717, "time_global": 18.128734081983566}
|
|
{"repeat": 64, "batch_size": 16384, "dim_out": 4096, "dim_in": 16384, "wm": 4, "switch": true, "standard_fwd": 8.378107100725174, "standard_gw": 8.923027664422989, "standard_gx": 9.049762040376663, "rowwise_fwd": 3.765825182199478, "rowwise_bwd": 4.183519631624222, "global_fwd": 3.744799643754959, "global_bwd": 4.1590481996536255, "x_quantize_rowwise": 0.8693933486938477, "g_quantize_rowwise": 0.1553073525428772, "w_quantize_rowwise": 0.2258792519569397, "w_quantize_colwise_transpose": 2.4386271834373474, "w_quantize_global": 0.4811100661754608, "w_quantize_global_transpose": 0.5102269351482391, "time_standard": 26.350896805524826, "time_rowwise": 20.5615796148777, "time_global": 18.842913210392}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 16384, "dim_in": 4096, "wm": 4, "switch": false, "standard_fwd": 18.266115337610245, "standard_gw": 17.671160399913788, "standard_gx": 17.10302010178566, "rowwise_fwd": 8.347474038600922, "rowwise_bwd": 7.514089345932007, "global_fwd": 8.263226598501205, "global_bwd": 7.487393915653229, "x_quantize_rowwise": 0.3021806478500366, "g_quantize_rowwise": 1.7319358885288239, "w_quantize_rowwise": 0.15519559383392334, "w_quantize_colwise_transpose": 1.6133114695549011, "w_quantize_global": 0.48247724771499634, "w_quantize_global_transpose": 0.506427139043808, "time_standard": 53.04029583930969, "time_rowwise": 37.3353473842144, "time_global": 36.44480183720589}
|
|
{"repeat": 64, "batch_size": 32768, "dim_out": 4096, "dim_in": 16384, "wm": 4, "switch": true, "standard_fwd": 17.73649826645851, "standard_gw": 16.359902918338776, "standard_gx": 18.0993489921093, "rowwise_fwd": 7.493957877159119, "rowwise_bwd": 8.352488279342651, "global_fwd": 7.486194372177124, "global_bwd": 8.28903540968895, "x_quantize_rowwise": 1.7313472926616669, "g_quantize_rowwise": 0.30205026268959045, "w_quantize_rowwise": 0.2255477011203766, "w_quantize_colwise_transpose": 2.4363920092582703, "w_quantize_global": 0.4815347492694855, "w_quantize_global_transpose": 0.5103759467601776, "time_standard": 52.195750176906586, "time_rowwise": 36.90168634057045, "time_global": 35.16044095158577}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 16384, "dim_in": 4096, "wm": 4, "switch": false, "standard_fwd": 36.309611052274704, "standard_gw": 32.85098075866699, "standard_gx": 34.34552624821663, "rowwise_fwd": 16.74525812268257, "rowwise_bwd": 15.026237815618515, "global_fwd": 16.574162989854813, "global_bwd": 14.977734535932541, "x_quantize_rowwise": 0.5954466760158539, "g_quantize_rowwise": 3.4569576382637024, "w_quantize_rowwise": 0.15521422028541565, "w_quantize_colwise_transpose": 1.6133897006511688, "w_quantize_global": 0.4822872579097748, "w_quantize_global_transpose": 0.5065612494945526, "time_standard": 103.50611805915833, "time_rowwise": 70.44348493218422, "time_global": 69.44413110613823}
|
|
{"repeat": 64, "batch_size": 65536, "dim_out": 4096, "dim_in": 16384, "wm": 4, "switch": true, "standard_fwd": 35.40017828345299, "standard_gw": 33.037226647138596, "standard_gx": 36.30436211824417, "rowwise_fwd": 15.043705701828003, "rowwise_bwd": 16.756191849708557, "global_fwd": 15.011314302682877, "global_bwd": 16.580048948526382, "x_quantize_rowwise": 3.4548528492450714, "g_quantize_rowwise": 0.5951337516307831, "w_quantize_rowwise": 0.22584572434425354, "w_quantize_colwise_transpose": 2.4329908192157745, "w_quantize_global": 0.4813261330127716, "w_quantize_global_transpose": 0.5101598799228668, "time_standard": 104.74176704883575, "time_rowwise": 71.54594734311104, "time_global": 69.67006251215935}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 16384, "dim_in": 4096, "wm": 4, "switch": false, "standard_fwd": 73.40333238244057, "standard_gw": 73.76311346888542, "standard_gx": 70.41774317622185, "rowwise_fwd": 33.37597846984863, "rowwise_bwd": 30.345775187015533, "global_fwd": 33.00366923213005, "global_bwd": 30.218638479709625, "x_quantize_rowwise": 1.1825822293758392, "g_quantize_rowwise": 6.902601569890976, "w_quantize_rowwise": 0.15529245138168335, "w_quantize_colwise_transpose": 1.6109198331832886, "w_quantize_global": 0.48149004578590393, "w_quantize_global_transpose": 0.5066059529781342, "time_standard": 217.58418902754784, "time_rowwise": 147.33626320958138, "time_global": 146.05870097875595}
|
|
{"repeat": 64, "batch_size": 131072, "dim_out": 4096, "dim_in": 16384, "wm": 4, "switch": true, "standard_fwd": 71.5160183608532, "standard_gw": 73.76786693930626, "standard_gx": 72.98104092478752, "rowwise_fwd": 30.291248112916946, "rowwise_bwd": 33.36654230952263, "global_fwd": 30.181586742401123, "global_bwd": 33.082425594329834, "x_quantize_rowwise": 6.902430206537247, "g_quantize_rowwise": 1.1815279722213745, "w_quantize_rowwise": 0.2262219786643982, "w_quantize_colwise_transpose": 2.4421699345111847, "w_quantize_global": 0.4816502332687378, "w_quantize_global_transpose": 0.5105249583721161, "time_standard": 218.26492622494698, "time_rowwise": 148.17800745368004, "time_global": 146.1080126464367}
|