{ "model": "EleutherAI/pythia-1b", "seq_len": 4096, "results": { "alpha_1": { "alpha": 1.0, "theta_new": 10000, "bands": { "A_in_train": { "range": [ 960, 1024 ], "ppl": 11.94919679725793 }, "B_boundary": { "range": [ 1984, 2048 ], "ppl": 12.659132687210402 }, "C_1p5x": { "range": [ 3008, 3072 ], "ppl": 620.9496501605687 }, "D_2x": { "range": [ 4032, 4096 ], "ppl": 1424.1467265906977 } } }, "alpha_4": { "alpha": 4.0, "theta_new": 40000, "bands": { "A_in_train": { "range": [ 960, 1024 ], "ppl": 13.877564029792271 }, "B_boundary": { "range": [ 1984, 2048 ], "ppl": 14.549244421859534 }, "C_1p5x": { "range": [ 3008, 3072 ], "ppl": 16.084113267384442 }, "D_2x": { "range": [ 4032, 4096 ], "ppl": 12.365884194357319 } } }, "alpha_16": { "alpha": 16.0, "theta_new": 160000, "bands": { "A_in_train": { "range": [ 960, 1024 ], "ppl": 40.116867227997055 }, "B_boundary": { "range": [ 1984, 2048 ], "ppl": 59.7889171894188 }, "C_1p5x": { "range": [ 3008, 3072 ], "ppl": 72.9693188079783 }, "D_2x": { "range": [ 4032, 4096 ], "ppl": 77.5390166251303 } } }, "alpha_opt": { "alpha": 64.0, "theta_new": 640000, "bands": { "A_in_train": { "range": [ 960, 1024 ], "ppl": 79.96912518546141 }, "B_boundary": { "range": [ 1984, 2048 ], "ppl": 143.51021968773875 }, "C_1p5x": { "range": [ 3008, 3072 ], "ppl": 172.83610058973207 }, "D_2x": { "range": [ 4032, 4096 ], "ppl": 219.85511253338788 } } } } }