{ "val_loss": { "A": 4.011326378476172, "M": 3.8980190759760376 }, "mean_stable_rank": { "A": 21.25855786991971, "M": 51.64423119383199 }, "expert_load": { "A": [ [ 0.10730743408203125, 0.134307861328125, 0.13019943237304688, 0.11050033569335938, 0.12069320678710938, 0.1259918212890625, 0.12838363647460938, 0.14261627197265625 ], [ 0.12824630737304688, 0.12598419189453125, 0.126678466796875, 0.114471435546875, 0.13177108764648438, 0.13433456420898438, 0.11130523681640625, 0.12720870971679688 ], [ 0.12761306762695312, 0.117706298828125, 0.10703659057617188, 0.13582992553710938, 0.12671279907226562, 0.133697509765625, 0.12516021728515625, 0.12624359130859375 ], [ 0.12885665893554688, 0.11507034301757812, 0.12677001953125, 0.12511062622070312, 0.12070083618164062, 0.13104248046875, 0.13479232788085938, 0.11765670776367188 ], [ 0.11621475219726562, 0.1228179931640625, 0.13051223754882812, 0.12974166870117188, 0.12760162353515625, 0.13164138793945312, 0.1198883056640625, 0.12158203125 ], [ 0.1197052001953125, 0.11420059204101562, 0.13045883178710938, 0.12439346313476562, 0.13484954833984375, 0.13097381591796875, 0.12134933471679688, 0.1240692138671875 ], [ 0.108673095703125, 0.11298370361328125, 0.118988037109375, 0.14570236206054688, 0.13838577270507812, 0.132476806640625, 0.11540603637695312, 0.12738418579101562 ], [ 0.12687301635742188, 0.12957000732421875, 0.11715316772460938, 0.14026260375976562, 0.13788986206054688, 0.0979156494140625, 0.11894607543945312, 0.13138961791992188 ] ], "M": [ [ 0.10551834106445312, 0.119171142578125, 0.13399124145507812, 0.13648223876953125, 0.11003494262695312, 0.12681961059570312, 0.13375473022460938, 0.13422775268554688 ], [ 0.12066650390625, 0.11786651611328125, 0.11908340454101562, 0.14267730712890625, 0.12557601928710938, 0.121124267578125, 0.12000274658203125, 0.13300323486328125 ], [ 0.10382843017578125, 0.13152313232421875, 0.13177490234375, 0.1380462646484375, 0.13188552856445312, 0.10686111450195312, 0.13076019287109375, 0.1253204345703125 ], [ 0.12167739868164062, 0.11628341674804688, 0.11981964111328125, 0.12691116333007812, 0.13732147216796875, 0.12253570556640625, 0.12982177734375, 0.12562942504882812 ], [ 0.12400054931640625, 0.12731552124023438, 0.12310791015625, 0.12537384033203125, 0.13324356079101562, 0.1433258056640625, 0.11721420288085938, 0.10641860961914062 ], [ 0.12651824951171875, 0.13423919677734375, 0.12768936157226562, 0.1204681396484375, 0.13137054443359375, 0.12180709838867188, 0.1119537353515625, 0.12595367431640625 ], [ 0.11201095581054688, 0.12734603881835938, 0.12652587890625, 0.1313934326171875, 0.13904953002929688, 0.12781143188476562, 0.1175537109375, 0.11830902099609375 ], [ 0.10103988647460938, 0.12125015258789062, 0.10439300537109375, 0.13260650634765625, 0.14451217651367188, 0.14339828491210938, 0.11890792846679688, 0.13389205932617188 ] ] }, "step_time_sec": { "A": 0.2630841016769409, "M": 0.40666224956512453 }, "train_time_sec": { "A": 418.63727259635925, "M": 688.8975319862366 }, "env": { "torch": "2.14.1+cu130", "python": "3.13.12", "gpu": "NVIDIA A100 80GB PCIe" }, "config_val_loss_hint": { "A": 4.0113, "M": 3.898 } }