[ { "model": "91M", "params": 91352064, "phase": "post", "budget_t": 18750, "checkpoint_iter": 18749, "total_training_tokens": 3839795200, "training_flops": 2104639301143756800, "tokens_per_param": 42.0, "gpu_hours": 2.5, "path": "91m/anneal/T018750/checkpoint-000018749.pt", "top1_move_match_pct": 54.093042353987165, "mean_nll_legal": 1.4127465559578585, "perplexity_legal": 4.1072206380053675, "final_val_loss": 1.419543981552124 }, { "model": "91M", "params": 91352064, "phase": "pre", "budget_t": 18750, "checkpoint_iter": 15000, "total_training_tokens": 3072000000, "training_flops": 1683801243648000000, "tokens_per_param": 33.6, "gpu_hours": 2.0, "path": "91m/trunk/checkpoint-000015000.pt", "top1_move_match_pct": 52.35954115665534, "mean_nll_legal": 1.4792722149911905, "perplexity_legal": 4.389749724033492, "final_val_loss": "" }, { "model": "91M", "params": 91352064, "phase": "post", "budget_t": 37500, "checkpoint_iter": 37499, "total_training_tokens": 7679795200, "training_flops": 4209390855703756800, "tokens_per_param": 84.1, "gpu_hours": 5.1, "path": "91m/anneal/T037500/checkpoint-000037499.pt", "top1_move_match_pct": 54.72366350733952, "mean_nll_legal": 1.3873449567792162, "perplexity_legal": 4.004204590913049, "final_val_loss": 1.3978652954101562 }, { "model": "91M", "params": 91352064, "phase": "pre", "budget_t": 37500, "checkpoint_iter": 30000, "total_training_tokens": 6144000000, "training_flops": 3367602487296000000, "tokens_per_param": 67.3, "gpu_hours": 4.1, "path": "91m/trunk/checkpoint-000030000.pt", "top1_move_match_pct": 53.01165433137756, "mean_nll_legal": 1.4534966116043373, "perplexity_legal": 4.278047062297825, "final_val_loss": "" }, { "model": "91M", "params": 91352064, "phase": "post", "budget_t": 75000, "checkpoint_iter": 74999, "total_training_tokens": 15359795200, "training_flops": 8418893964823756800, "tokens_per_param": 168.1, "gpu_hours": 10.1, "path": "91m/anneal/T075000/checkpoint-000074999.pt", "top1_move_match_pct": 55.205876597337934, "mean_nll_legal": 1.3678629669672289, "perplexity_legal": 3.9269497008374485, "final_val_loss": 1.3811485767364502 }, { "model": "91M", "params": 91352064, "phase": "pre", "budget_t": 75000, "checkpoint_iter": 60000, "total_training_tokens": 12288000000, "training_flops": 6735204974592000000, "tokens_per_param": 134.5, "gpu_hours": 8.1, "path": "91m/trunk/checkpoint-000060000.pt", "top1_move_match_pct": 53.35778899133419, "mean_nll_legal": 1.4370774849776407, "perplexity_legal": 4.208378777200237, "final_val_loss": "" }, { "model": "91M", "params": 91352064, "phase": "post", "budget_t": 150000, "checkpoint_iter": 149999, "total_training_tokens": 30719795200, "training_flops": 16837900183063756800, "tokens_per_param": 336.3, "gpu_hours": 20.3, "path": "91m/anneal/T150000/checkpoint-000149999.pt", "top1_move_match_pct": 55.55879821141135, "mean_nll_legal": 1.353085779709565, "perplexity_legal": 3.8693470803089225, "final_val_loss": 1.368669033050537 }, { "model": "91M", "params": 91352064, "phase": "pre", "budget_t": 150000, "checkpoint_iter": 120000, "total_training_tokens": 24576000000, "training_flops": 13470409949184000000, "tokens_per_param": 269.0, "gpu_hours": 16.2, "path": "91m/trunk/checkpoint-000120000.pt", "top1_move_match_pct": 53.73197639497358, "mean_nll_legal": 1.4261519260876638, "perplexity_legal": 4.162650147334285, "final_val_loss": "" }, { "model": "91M", "params": 91352064, "phase": "post", "budget_t": 300000, "checkpoint_iter": 299999, "total_training_tokens": 61439795200, "training_flops": 33675912619543756800, "tokens_per_param": 672.6, "gpu_hours": 40.5, "path": "91m/anneal/T300000/checkpoint-000299999.pt", "top1_move_match_pct": 55.84554702284602, "mean_nll_legal": 1.3425605147111692, "perplexity_legal": 3.8288347524109945, "final_val_loss": 1.3608551025390625 }, { "model": "91M", "params": 91352064, "phase": "pre", "budget_t": 300000, "checkpoint_iter": 240000, "total_training_tokens": 49152000000, "training_flops": 26940819898368000000, "tokens_per_param": 538.1, "gpu_hours": 32.4, "path": "91m/trunk/checkpoint-000240000.pt", "top1_move_match_pct": 53.863190841231656, "mean_nll_legal": 1.4182724359958372, "perplexity_legal": 4.129979469915587, "final_val_loss": "" }, { "model": "91M", "params": 91352064, "phase": "pre", "budget_t": 600000, "checkpoint_iter": 480000, "total_training_tokens": 98304000000, "training_flops": 53881639796736000000, "tokens_per_param": 1076.1, "gpu_hours": 64.9, "path": "91m/trunk/checkpoint-000480000.pt", "top1_move_match_pct": 53.96488203708165, "mean_nll_legal": 1.41351416384541, "perplexity_legal": 4.110374583304795, "final_val_loss": "" }, { "model": "91M", "params": 91352064, "phase": "trunk-final", "budget_t": 600000, "checkpoint_iter": 599999, "total_training_tokens": 122879795200, "training_flops": 67351937492503756800, "tokens_per_param": 1345.1, "gpu_hours": 81.1, "path": "91m/trunk/checkpoint-000599999.pt", "top1_move_match_pct": 56.05379339832973, "mean_nll_legal": 1.3347227693112564, "perplexity_legal": 3.798942616940196, "final_val_loss": 1.3550746440887451 }, { "model": "350M", "params": 316887040, "phase": "post", "budget_t": 18750, "checkpoint_iter": 18749, "total_training_tokens": 3839795200, "training_flops": 7300688010805248000, "tokens_per_param": 12.1, "gpu_hours": 7.8, "path": "350m/anneal/T018750/checkpoint-000018749.pt", "top1_move_match_pct": 54.998082685462, "mean_nll_legal": 1.3743365070090774, "perplexity_legal": 3.952453427601623, "final_val_loss": 1.3888936042785645 }, { "model": "350M", "params": 316887040, "phase": "pre", "budget_t": 18750, "checkpoint_iter": 15000, "total_training_tokens": 3072000000, "training_flops": 5840861921280000000, "tokens_per_param": 9.7, "gpu_hours": 6.3, "path": "350m/trunk/checkpoint-000015000.pt", "top1_move_match_pct": 53.4109534652491, "mean_nll_legal": 1.436669928291017, "perplexity_legal": 4.206663973753286, "final_val_loss": "" }, { "model": "350M", "params": 316887040, "phase": "post", "budget_t": 37500, "checkpoint_iter": 37499, "total_training_tokens": 7679795200, "training_flops": 14601765412405248000, "tokens_per_param": 24.2, "gpu_hours": 15.7, "path": "350m/anneal/T037500/checkpoint-000037499.pt", "top1_move_match_pct": 55.644200717381054, "mean_nll_legal": 1.3499370063335419, "perplexity_legal": 3.8571825449730675, "final_val_loss": 1.3687193393707275 }, { "model": "350M", "params": 316887040, "phase": "pre", "budget_t": 37500, "checkpoint_iter": 30000, "total_training_tokens": 6144000000, "training_flops": 11681723842560000000, "tokens_per_param": 19.4, "gpu_hours": 12.5, "path": "350m/trunk/checkpoint-000030000.pt", "top1_move_match_pct": 54.00820542752721, "mean_nll_legal": 1.4130794183765105, "perplexity_legal": 4.108588004960783, "final_val_loss": "" }, { "model": "350M", "params": 316887040, "phase": "post", "budget_t": 75000, "checkpoint_iter": 74999, "total_training_tokens": 15359795200, "training_flops": 29203920215605248000, "tokens_per_param": 48.5, "gpu_hours": 31.4, "path": "350m/anneal/T075000/checkpoint-000074999.pt", "top1_move_match_pct": 56.12188917130159, "mean_nll_legal": 1.3307202424063012, "perplexity_legal": 3.7837676362985357, "final_val_loss": 1.3526790142059326 }, { "model": "350M", "params": 316887040, "phase": "pre", "budget_t": 75000, "checkpoint_iter": 60000, "total_training_tokens": 12288000000, "training_flops": 23363447685120000000, "tokens_per_param": 38.8, "gpu_hours": 25.1, "path": "350m/trunk/checkpoint-000060000.pt", "top1_move_match_pct": 54.373682906716716, "mean_nll_legal": 1.3976302064650983, "perplexity_legal": 4.045601358033151, "final_val_loss": "" }, { "model": "350M", "params": 316887040, "phase": "post", "budget_t": 150000, "checkpoint_iter": 149999, "total_training_tokens": 30719795200, "training_flops": 58408229822005248000, "tokens_per_param": 96.9, "gpu_hours": 62.7, "path": "350m/anneal/T150000/checkpoint-000149999.pt", "top1_move_match_pct": 56.49483230001957, "mean_nll_legal": 1.317104624559824, "perplexity_legal": 3.7325984427332224, "final_val_loss": 1.3407398462295532 }, { "model": "350M", "params": 316887040, "phase": "pre", "budget_t": 150000, "checkpoint_iter": 120000, "total_training_tokens": 24576000000, "training_flops": 46726895370240000000, "tokens_per_param": 77.6, "gpu_hours": 50.2, "path": "350m/trunk/checkpoint-000120000.pt", "top1_move_match_pct": 54.6461791145061, "mean_nll_legal": 1.3879074987910138, "perplexity_legal": 4.0064577579103995, "final_val_loss": "" }, { "model": "350M", "params": 316887040, "phase": "post", "budget_t": 300000, "checkpoint_iter": 299999, "total_training_tokens": 61439795200, "training_flops": 116816849034805248000, "tokens_per_param": 193.9, "gpu_hours": 125.5, "path": "350m/anneal/T300000/checkpoint-000299999.pt", "top1_move_match_pct": 56.77705647537636, "mean_nll_legal": 1.3062968320815889, "perplexity_legal": 3.6924745094983553, "final_val_loss": 1.3332316875457764 }, { "model": "350M", "params": 316887040, "phase": "pre", "budget_t": 300000, "checkpoint_iter": 240000, "total_training_tokens": 49152000000, "training_flops": 93453790740480000000, "tokens_per_param": 155.1, "gpu_hours": 100.4, "path": "350m/trunk/checkpoint-000240000.pt", "top1_move_match_pct": 54.86358787804748, "mean_nll_legal": 1.3808047559006091, "perplexity_legal": 3.9781017405160255, "final_val_loss": "" }, { "model": "350M", "params": 316887040, "phase": "pre", "budget_t": 600000, "checkpoint_iter": 480000, "total_training_tokens": 98304000000, "training_flops": 186907581480960000000, "tokens_per_param": 310.2, "gpu_hours": 200.8, "path": "350m/trunk/checkpoint-000480000.pt", "top1_move_match_pct": 54.99548101971723, "mean_nll_legal": 1.3765396770902847, "perplexity_legal": 3.961170954311486, "final_val_loss": "" }, { "model": "350M", "params": 316887040, "phase": "trunk-final", "budget_t": 600000, "checkpoint_iter": 599999, "total_training_tokens": 122879795200, "training_flops": 233634087460405248000, "tokens_per_param": 387.8, "gpu_hours": 251.0, "path": "350m/trunk/checkpoint-000599999.pt", "top1_move_match_pct": 57.012111319621425, "mean_nll_legal": 1.2984858923993905, "perplexity_legal": 3.6637451614952985, "final_val_loss": 1.3280322551727295 }, { "model": "700M", "params": 742200320, "phase": "post", "budget_t": 18750, "checkpoint_iter": 18749, "total_training_tokens": 3839795200, "training_flops": 17099383357046784000, "tokens_per_param": 5.2, "gpu_hours": 16.8, "path": "700m/anneal/T018750/checkpoint-000018749.pt", "top1_move_match_pct": 55.41638529086057, "mean_nll_legal": 1.3589848522814982, "perplexity_legal": 3.8922400969369915, "final_val_loss": 1.372852087020874 }, { "model": "700M", "params": 742200320, "phase": "pre", "budget_t": 18750, "checkpoint_iter": 15000, "total_training_tokens": 3072000000, "training_flops": 13680236298240000000, "tokens_per_param": 4.1, "gpu_hours": 13.4, "path": "700m/trunk/checkpoint-000015000.pt", "top1_move_match_pct": 53.79916724072987, "mean_nll_legal": 1.4208227219136957, "perplexity_legal": 4.140525540430386, "final_val_loss": "" }, { "model": "700M", "params": 742200320, "phase": "post", "budget_t": 37500, "checkpoint_iter": 37499, "total_training_tokens": 7679795200, "training_flops": 34199678729846784000, "tokens_per_param": 10.3, "gpu_hours": 33.6, "path": "700m/anneal/T037500/checkpoint-000037499.pt", "top1_move_match_pct": 56.06284267048546, "mean_nll_legal": 1.3347861838156103, "perplexity_legal": 3.799183532642012, "final_val_loss": 1.353476881980896 }, { "model": "700M", "params": 742200320, "phase": "pre", "budget_t": 37500, "checkpoint_iter": 30000, "total_training_tokens": 6144000000, "training_flops": 27360472596480000000, "tokens_per_param": 8.3, "gpu_hours": 26.9, "path": "700m/trunk/checkpoint-000030000.pt", "top1_move_match_pct": 54.43295563933673, "mean_nll_legal": 1.3967479861986556, "perplexity_legal": 4.04203382043352, "final_val_loss": "" }, { "model": "700M", "params": 742200320, "phase": "post", "budget_t": 75000, "checkpoint_iter": 74999, "total_training_tokens": 15359795200, "training_flops": 68400269475446784000, "tokens_per_param": 20.7, "gpu_hours": 67.2, "path": "700m/anneal/T075000/checkpoint-000074999.pt", "top1_move_match_pct": 56.50399468807724, "mean_nll_legal": 1.3170397282517825, "perplexity_legal": 3.7323562187346946, "final_val_loss": 1.337888479232788 }, { "model": "700M", "params": 742200320, "phase": "pre", "budget_t": 75000, "checkpoint_iter": 60000, "total_training_tokens": 12288000000, "training_flops": 54720945192960000000, "tokens_per_param": 16.6, "gpu_hours": 53.8, "path": "700m/trunk/checkpoint-000060000.pt", "top1_move_match_pct": 54.73633248835754, "mean_nll_legal": 1.3824926166336773, "perplexity_legal": 3.984821891980476, "final_val_loss": "" }, { "model": "700M", "params": 742200320, "phase": "post", "budget_t": 150000, "checkpoint_iter": 149999, "total_training_tokens": 30719795200, "training_flops": 136801450966646784000, "tokens_per_param": 41.4, "gpu_hours": 134.4, "path": "700m/anneal/T150000/checkpoint-000149999.pt", "top1_move_match_pct": 56.880557525657515, "mean_nll_legal": 1.303091969236236, "perplexity_legal": 3.6806595778658946, "final_val_loss": 1.3260610103607178 }, { "model": "700M", "params": 742200320, "phase": "pre", "budget_t": 150000, "checkpoint_iter": 120000, "total_training_tokens": 24576000000, "training_flops": 109441890385920000000, "tokens_per_param": 33.1, "gpu_hours": 107.5, "path": "700m/trunk/checkpoint-000120000.pt", "top1_move_match_pct": 55.02534361783113, "mean_nll_legal": 1.3726845887174977, "perplexity_legal": 3.9459296873144227, "final_val_loss": "" }, { "model": "700M", "params": 742200320, "phase": "post", "budget_t": 300000, "checkpoint_iter": 299999, "total_training_tokens": 61439795200, "training_flops": 273603813949046784000, "tokens_per_param": 82.8, "gpu_hours": 268.8, "path": "700m/anneal/T300000/checkpoint-000299999.pt", "top1_move_match_pct": 57.15022583589824, "mean_nll_legal": 1.2917126068377571, "perplexity_legal": 3.6390134214126957, "final_val_loss": 1.3188841342926023 }, { "model": "700M", "params": 742200320, "phase": "pre", "budget_t": 300000, "checkpoint_iter": 240000, "total_training_tokens": 49152000000, "training_flops": 218883780771840000000, "tokens_per_param": 66.2, "gpu_hours": 215.0, "path": "700m/trunk/checkpoint-000240000.pt", "top1_move_match_pct": 55.28856432166091, "mean_nll_legal": 1.3654216689845482, "perplexity_legal": 3.9173745391218078, "final_val_loss": "" } ]