SOURCE C:\Users\marco\.codex\sessions\2026\05\25\rollout-2026-05-25T20-54-01-019e61c6-391a-75b0-adee-0128be6f2a9d.jsonl line 940 2026-05-26 02:52 UTC NOTE Literal terminal output, unedited. [ ... ] markers are the only editorial additions. No image was rendered for this extract, so there is no amber marking: every [ ... ] line is ours. ---------------------------------------------------------------------------------------------------- Chunk ID: 57531d Wall time: 0.1470 seconds Process exited with code 0 Original token count: 10217 Output: Total output lines: 240 [0] tensor=blk.0.ffn_gate_exps.weight backend=CUDA0 used=145 hits=0 misses=145 reuse_pct=0.0 base_groups=105 copy_groups=105 base_mb=77.082 copy_mb=77.082 saved_mb=0.000 [1] tensor=blk.0.ffn_up_exps.weight backend=CUDA0 used=145 hits=0 misses=145 reuse_pct=0.0 base_groups=105 copy_groups=105 base_mb=77.082 copy_mb=77.082 saved_mb=0.000 [2] tensor=blk.0.ffn_down_exps.weight backend=CUDA0 used=145 hits=0 misses=145 reuse_pct=0.0 base_groups=105 copy_groups=105 base_mb=77.082 copy_mb=77.082 saved_mb=0.000 [3] tensor=blk.1.ffn_gate_exps.weight backend=CUDA0 used=125 hits=0 misses=125 reuse_pct=0.0 base_groups=96 copy_groups=96 base_mb=66.453 copy_mb=66.453 saved_mb=0.000 [4] tensor=blk.1.ffn_up_exps.weight backend=CUDA0 used=125 hits=0 misses=125 reuse_pct=0.0 base_groups=96 copy_groups=96 base_mb=66.453 copy_mb=66.453 saved_mb=0.000 [5] tensor=blk.1.ffn_down_exps.weight backend=CUDA0 used=125 hits=0 misses=125 reuse_pct=0.0 base_groups=96 copy_groups=96 base_mb=66.453 copy_mb=66.453 saved_mb=0.000 [6] tensor=blk.2.ffn_gate_exps.weight backend=CUDA0 used=182 hits=0 misses=182 reuse_pct=0.0 base_groups=117 copy_groups=117 base_mb=96.745 copy_mb=96.745 saved_mb=0.000 [7] tensor=blk.2.ffn_up_exps.weight backend=CUDA0 used=182 hits=0 misses=182 reuse_pct=0.0 base_groups=117 copy_groups=117 base_mb=96.745 copy_mb=96.745 saved_mb=0.000 [8] tensor=blk.2.ffn_down_exps.weight backend=CUDA0 used=182 hits=0 misses=182 reuse_pct=0.0 base_groups=117 copy_groups=117 base_mb=96.745 copy_mb=96.745 saved_mb=0.000 [9] tensor=blk.3.ffn_gate_exps.weight backend=CUDA0 used=148 hits=0 misses=148 reuse_pct=0.0 base_groups=105 copy_groups=105 base_mb=78.676 copy_mb=78.676 saved_mb=0.000 [10] tensor=blk.3.ffn_up_exps.weight backend=CUDA0 used=148 hits=0 misses=148 reuse_pct=0.0 base_groups=105 copy_groups=105 base_mb=78.676 copy_mb=78.676 saved_mb=0.000 [11] tensor=blk.3.ffn_down_exps.weight backend=CUDA0 used=148 hits=0 misses=148 reuse_pct=0.0 base_groups=105 copy_groups=105 base_mb=78.676 copy_mb=78.676 saved_mb=0.000 [12] tensor=blk.4.ffn_gate_exps.weight backend=CUDA0 used=150 hits=0 misses=150 reuse_pct=0.0 base_groups=104 copy_groups=104 base_mb=79.738 copy_mb=79.738 saved_mb=0.000 [13] tensor=blk.4.ffn_up_exps.weight backend=CUDA0 used=150 hits=0 misses=150 reuse_pct=0.0 base_groups=104 copy_groups=104 base_mb=79.738 copy_mb=79.738 saved_mb=0.000 [14] tensor=blk.4.ffn_down_exps.weight backend=CUDA0 used=150 hits=0 misses=150 reuse_pct=0.0 base_groups=104 copy_groups=104 base_mb=79.738 copy_mb=79.738 saved_mb=0.000 [15] tensor=blk.5.ffn_gate_exps.weight backend=CUDA0 used=151 hits=0 misses=151 reuse_pct=0.0 base_groups=109 copy_groups=109 base_mb=80.272 copy_mb=80.272 saved_mb=0.000 [16] tensor=blk.5.ffn_up_exps.weight backend=CUDA0 used=151 hits=0 misses=151 reuse_pct=0.0 base_groups=109 copy_groups=109 base_mb=80.272 copy_mb=80.272 saved_mb=0.000 [17] tensor=blk.5.ffn_down_exps.weight backend=CUDA0 used=151 hits=0 misses=151 reuse_pct=0.0 base_groups=109 copy_groups=109 base_mb=80.272 copy_mb=80.272 saved_mb=0.000 [18] tensor=blk.6.ffn_gate_exps.weight backend=CUDA0 used=197 hits=0 misses=197 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=104.719 copy_mb=104.719 saved_mb=0.000 [19] tensor=blk.6.ffn_up_exps.weight backend=CUDA0 used=197 hits=0 misses=197 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=104.719 copy_mb=104.719 saved_mb=0.000 [20] tensor=blk.6.ffn_down_exps.weight backend=CUDA0 used=197 hits=0 misses=197 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=104.719 copy_mb=104.719 saved_mb=0.000 [21] tensor=blk.7.ffn_gate_exps.weight backend=CUDA0 used=215 hits=0 misses=215 reuse_pct=0.0 base_groups=131 copy_groups=131 base_mb=114.282 copy_mb=114.282 saved_mb=0.000 [22] tensor=blk.7.ffn_up_exps.weight backend=CUDA0 used=215 hits=0 misses=215 reuse_pct=0.0 base_groups=131 copy_groups=131 base_mb=114.282 copy_mb=114.282 saved_mb=0.000 [23] tensor=blk.7.ffn_down_exps.weight backend=CUDA0 used=215 hits=0 misses=215 reuse_pct=0.0 base_groups=131 copy_groups=131 base_mb=114.282 copy_mb=114.282 saved_mb=0.000 [24] tensor=blk.8.ffn_gate_exps.weight backend=CUDA0 used=190 hits=0 misses=190 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=101.000 copy_mb=101.000 saved_mb=0.000 [25] tensor=blk.8.ffn_up_exps.weight backend=CUDA0 used=190 hits=0 misses=190 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=101.000 copy_mb=101.000 saved_mb=0.000 [26] tensor=blk.8.ffn_down_exps.weight backend=CUDA0 used=190 hits=0 misses=190 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=101.000 copy_mb=101.000 saved_mb=0.000 [27] tensor=blk.9.ffn_gate_exps.weight backend=CUDA0 used=181 hits=0 misses=181 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=96.216 copy_mb=96.216 saved_mb=0.000 [28] tensor=blk.9.ffn_up_exps.weight backend=CUDA0 used=181 hits=0 misses=181 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=96.216 copy_mb=96.216 saved_mb=0.000 [29] tensor=blk.9.ffn_down_exps.weight backend=CUDA0 used=181 hits=0 misses=181 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=96.216 copy_mb=96.216 saved_mb=0.000 [30] tensor=blk.10.ffn_gate_exps.weight backend=CUDA0 used=163 hits=0 misses=163 reuse_pct=0.0 base_groups=115 copy_groups=115 base_mb=86.650 copy_mb=86.650 saved_mb=0.000 [31] tensor=blk.10.ffn_up_exps.weight backend=CUDA0 used=163 hits=0 misses=163 reuse_pct=0.0 base_groups=115 copy_groups=115 base_mb=86.650 copy_mb=86.650 saved_mb=0.000 [32] tensor=blk.10.ffn_down_exps.weight backend=CUDA0 used=163 hits=0 misses=163 reuse_pct=0.0 base_groups=115 copy_groups=115 base_mb=86.650 copy_mb=86.650 saved_mb=0.000 [33] tensor=blk.11.ffn_gate_exps.weight backend=CUDA0 used=164 hits=0 misses=164 reuse_pct=0.0 base_groups=110 copy_groups=110 base_mb=87.178 copy_mb=87.178 saved_mb=0.000 [34] tensor=blk.11.ffn_up_exps.weight backend=CUDA0 used=164 hits=0 misses=164 reuse_pct=0.0 base_groups=110 copy_groups=110 base_mb=87.178 copy_mb=87.178 saved_mb=0.000 [35] tensor=blk.11.ffn_down_exps.weight backend=CUDA0 used=164 hits=0 misses=164 reuse_pct=0.0 base_groups=110 copy_groups=110 base_mb=87.178 copy_mb=87.178 saved_mb=0.000 [36] tensor=blk.12.ffn_gate_exps.weight backend=CUDA0 used=162 hits=0 misses=162 reuse_pct=0.0 base_groups=106 copy_groups=106 base_mb=86.114 copy_mb=86.114 saved_mb=0.000 [37] tensor=blk.12.ffn_up_exps.weight backend=CUDA0 used=162 hits=0 misses=162 reuse_pct=0.0 base_groups=106 copy_groups=106 base_mb=86.114 copy_mb=86.114 saved_mb=0.000 [38] tensor=blk.12.ffn_down_exps.weight backend=CUDA0 used=162 hits=0 misses=162 reuse_pct=0.0 base_groups=106 copy_groups=106 base_mb=86.114 copy_mb=86.114 saved_mb=0.000 [39] tensor=blk.13.ffn_gate_exps.weight backend=CUDA0 used=176 hits=0 misses=176 reuse_pct=0.0 base_groups=108 copy_groups=108 base_mb=93.552 copy_mb=93.552 saved_mb=0.000 [40] tensor=blk.13.ffn_up_exps.weight backend=CUDA0 used=176 hits=0 misses=176 reuse_pct=0.0 base_groups=108 copy_groups=108 base_mb=93.552 copy_mb=93.552 saved_mb=0.000 [41] tensor=blk.13.ffn_down_exps.weight backend=CUDA0 used=176 hits=0 misses=176 reuse_pct=0.0 base_groups=108 copy_groups=108 base_mb=93.552 copy_mb=93.552 saved_mb=0.000 [42] tensor=blk.14.ffn_gate_exps.weight backend=CUDA0 used=212 hits=0 misses=212 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=112.691 copy_mb=112.691 saved_mb=0.000 [43] tensor=blk.14.ffn_up_exps.weight backend=CUDA0 used=212 hits=0 misses=212 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=112.691 copy_mb=112.691 saved_mb=0.000 [44] tensor=blk.14.ffn_down_exps.weight backend=CUDA0 used=212 hits=0 misses=212 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=112.691 copy_mb=112.691 saved_mb=0.000 [45] tensor=blk.15.ffn_gate_exps.weight backend=CUDA0 used=200 hits=0 misses=200 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=106.308 copy_mb=106.308 saved_mb=0.000 [46] tensor=blk.15.ffn_up_exps.weight backend=CUDA0 used=200 hits=0 misses=200 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=106.308 copy_mb=106.308 saved_mb=0.000 [47] tensor=blk.15.ffn_down_exps.weight backend=CUDA0 used=200 hits=0 misses=200 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=106.308 copy_mb=106.308 saved_mb=0.000 [48] tensor=blk.16.ffn_gate_exps.weight backend=CUDA0 used=218 hits=0 misses=218 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=115.875 copy_mb=115.875 saved_mb=0.000 [49] tensor=blk.16.ffn_up_exps.weight backend=CUDA0 used=218 hits=0 misses=218 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=115.875 copy_mb=115.875 saved_mb=0.000 [50] tensor=blk.16.ffn_down_exps.weight backend=CUDA0 used=218 hits=0 misses=218 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=115.875 copy_mb=115.875 saved_mb=0.000 [51] tensor=blk.17.ffn_gate_exps.weight backend=CUDA0 used=229 hits=0 misses=229 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=121.716 copy_mb=121.716 saved_mb=0.000 [52] tensor=blk.17.ffn_up_exps.weight backend=CUDA0 used=229 hits=0 misses=229 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=121.716 copy_mb=121.716 saved_mb=0.000 [53] tensor=blk.17.ffn_down_exps.weight backend=CUDA0 used=229 hits=0 misses=229 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=121.716 copy_mb=121.716 saved_mb=0.000 [54] tensor=blk.18.ffn_gate_exps.weight backend=CUDA0 used=252 hits=0 misses=252 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=133.938 copy_mb=133.938 saved_mb=0.000 [55] tensor=blk.18.ffn_up_exps.weight backend=CUDA0 used=252 hits=0 misses=252 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=133.938 copy_mb=133.938 saved_mb=0.000 [56] tensor=blk.18.ffn_down_exps.weight backend=CUDA0 used=252 hits=0 misses=252 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=133.938 copy_mb=133.938 saved_mb=0.000 [57] tensor=blk.19.ffn_gate_exps.weight backend=CUDA0 used=251 hits=0 misses=251 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=133.404 copy_mb=133.404 saved_mb=0.000 [58] tensor=blk.19.ffn_up_exps.weight backend=CUDA0 used=251 hits=0 misses=251 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=133.404 copy_mb=133.404 saved_mb=0.000 [59] tensor=blk.19.ffn_down_exps.weight backend=CUDA0 used=251 hits=0 misses=251 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=133.404 copy_mb=133.404 saved_mb=0.000 [60] tensor=blk.20.ffn_gate_exps.weight backend=CUDA0 used=258 hits=0 misses=258 reuse_pct=0.0 base_groups=137 copy_groups=137 base_mb=137.129 copy_mb=137.129 saved_mb=0.000 [61] tensor=blk.20.ffn_up_exps.weight backend=CUDA0 used=258 hits=0 misses=258 reuse_pct=0.0 base_groups=137 copy_groups=137 base_mb=137.129 copy_mb=137.129 saved_mb=0.000 [62] tensor=blk.20.ffn_down_exps.weight backend=CUDA0 used=258 hits=0 misses=258 reuse_pct=0.0 base_groups=137 copy_groups=137 base_mb=137.129 copy_mb=137.129 saved_mb=0.000 [63] tensor=blk.21.ffn_gate_exps.weight backend=CUDA0 used=226 hits=0 misses=226 reuse_pct=0.0 base_groups=133 copy_groups=133 base_mb=120.127 copy_mb=120.127 saved_mb=0.000 [64] tensor=blk.21.ffn_up_exps.weight backend=CUDA0 used=226 hits=0 misses=226 reuse_pct=0.0 base_groups=133 copy_groups=133 base_mb=120.127 copy_mb=120.127 saved_mb=0.000 [65] tensor=blk.21.ffn_down_exps.weight backend=CUDA0 used=226 hits=0 misses=226 reuse_pct=0.0 base_groups=133 copy_groups=133 base_mb=120.127 copy_mb=120.127 saved_mb=0.000 [66] tensor=blk.22.ffn_gate_exps.weight backend=CUDA0 used=227 hits=0 misses=227 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=120.660 copy_mb=120.660 saved_mb=0.000 [67] tensor=blk.22.ffn_up_exps.weight backend=CUDA0 used=227 hits=0 misses=227 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=120.660 copy_mb=120.660 saved_mb=0.000 [68] tensor=blk.22.ffn_down_exps.weight backend=CUDA0 used=227 hits=0 misses=227 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=120.660 copy_mb=120.660 saved_mb=0.000 [69] tensor=blk.23.ffn_gate_exps.weight backend=CUDA0 used=276 hits=0 misses=276 reuse_pct=0.0 base_groups=126 copy_groups=126 base_mb=146.687 copy_mb=146.687 saved_mb=0.000 [70] tensor=blk.23.ffn_up_exps.weight backend=CUDA0 used=276 hits=0 misses=276 reuse_pct=0.0 base_groups=126 copy_groups=126 base_mb=146.687 copy_mb=146.687 saved_mb=0.000 [71] tensor=blk.23.ffn_down_exps.weight backend=CUDA0 used=276 hits=0 misses=276 reuse_pct=0.0 base_groups=126 copy_groups=126 base_mb=146.687 copy_mb=146.687 saved_mb=0.000 [72] tensor=blk.24.ffn_gate_exps.weight backend=CUDA0 used=265 hits=0 misses=265 reuse_pct=0.0 base_groups=118 copy_groups=118 base_mb=140.838 copy_mb=140.838 saved_mb=0.000 [73] tensor=blk.24.ffn_up_exps.weight backend=CUDA0 used=265 hits=0 misses=265 reuse_pct=0.0 base_groups=118 copy_groups=118 base_mb=140.838 copy_mb=140.838 saved_mb=0.000 [74] tensor=blk.24.ffn_down_exps.weight backend=CUDA0 used=265 hits=0 misses=265 reuse_pct=0.0 base_groups=118 copy_groups=118 base_mb=140.838 copy_mb=140.838 saved_mb=0.000 [75] tensor=blk.25.ffn_gate_exps.weight backend=CUDA0 used=256 hits=0 misses=256 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=136.062 copy_mb=136.062 saved_mb=0.000 [76] tensor=blk.25.ffn_up_exps.weight backend=CUDA0 used=256 hits=0 misses=256 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=136.062 copy_mb=136.062 saved_mb=0.000 [77] tensor=blk.25.ffn_down_exps.weight backend=CUDA0 used=256 hits=0 misses=256 reuse_pct=0.0 base_groups=129 copy_groups=129 base_mb=136.062 copy_mb=136.062 saved_mb=0.000 [78] tensor=blk.26.ffn_gate_exps.weight backend=CUDA0 used=271 hits=0 misses=271 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=144.033 copy_mb=144.033 saved_mb=0.000 [79] tensor=blk.26.ffn_up_exps.weight backend=CUDA0 used=271 hits=0 misses=271 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=144.033 copy_mb=144.033 saved_mb=0.000 [80] tensor=blk.26.ffn_down_exps.weight backend=CUDA0 used=271 hits=0 misses=271 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=144.033 copy_mb=144.033 saved_mb=0.000 [81] tensor=blk.27.ffn_gate_exps.weight backend=CUDA0 used=275 hits=0 misses=275 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=146.151 copy_mb=146.151 saved_mb=0.000 [82] tensor=blk.27.ffn_up_exps.weight backend=CUDA0 used=275 hits=0 misses=275 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=146.151 copy_mb=146.151 saved_mb=0.000 [83] tensor=blk.27.ffn_down_exps.weight backend=CUDA0 used=275 hits=0 misses=275 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=146.151 copy_mb=146.151 saved_mb=0.000 [84] tensor=blk.28.ffn_gate_exps.weight backend=CUDA0 used=281 hits=0 misses=281 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=149.345 copy_mb=149.345 saved_mb=0.000 [85] tensor=blk.28.ffn_up_exps.weight backend=CUDA0 used=281 hits=0 misses=281 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=149.345 copy_mb=149.345 saved_mb=0.000 [86] tensor=blk.28.ffn_down_exps.weight backend=CUDA0 used=281 hits=0 misses=281 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=149.345 copy_mb=149.345 saved_mb=0.000 [87] tensor=blk.29.ffn_gate_exps.weight backend=CUDA0 used=298 hits=0 misses=298 reuse_pct=0.0 base_groups=127 copy_groups=127 base_mb=158.375 copy_mb=158.375 saved_mb=0.000 [88] tensor=blk.29.ffn_up_exps.weight backend=CUDA0 used=298 hits=0 misses=298 reuse_pct=0.0 base_groups=127 copy_groups=127 base_mb=158.375 copy_mb=158.375 saved_mb=0.000 [89] tensor=blk.29.ffn_down_exps.weight backend=CUDA0 used=298 hits=0 misses=298 reuse_pct=0.0 base_groups=127 copy_groups=127 base_mb=158.375 copy_mb=158.375 saved_mb=0.000 [90] tensor=blk.30.ffn_gate_exps.weight backend=CUDA0 used=301 hits=0 misses=301 reuse_pct=0.0 base_groups=126 copy_groups=126 base_mb=159.967 copy_mb=159.967 saved_mb=0.000 [91] tensor=blk.30.ffn_up_exps.weight backend=CUDA0 used=301 hits=0 misses=301 reuse_pct=0.0 base_groups=126 copy_groups=126 base_mb=159.967 copy_mb=159.967 saved_mb=0.000 [92] tensor=blk.30.ffn_down_exps.weight backend=CUDA0 used=301 hits=0 misses=301 reuse_pct=0.0 base_groups=126 copy_groups=126 base_mb=159.967 copy_mb=159.967 saved_mb=0.000 [93] tensor=blk.31.ffn_gate_exps.weight backend=CUDA0 used=324 hits=0 misses=324 reuse_pct=0.0 base_groups=120 copy_groups=120 base_mb=172.183 copy_mb=172.183 saved_mb=0.000 [94] tensor=blk.31.ffn_up_exps.weight backend=CUDA0 used=324 hits=0 misses=324 reuse_pct=0.0 base_groups=120 copy_groups=120 base_mb=172.183 copy_mb=172.183 saved_mb=0.000 [95] tensor=blk.31.ffn_down_exps.weight backend=CUDA0 used=324 hits=0 misses=324 reuse_pct=0.0 base_groups=120 copy_groups=120 base_mb=172.183 copy_mb=172.183 saved_mb=0.000 [96] tensor=blk.32.ffn_gate_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=122 copy_groups=122 base_mb=150.403 copy_mb=150.403 saved_mb=0.000 [97] tensor=blk.32.ffn_up_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=122 copy_groups=122 base_mb=150.403 copy_mb=150.403 saved_mb=0.000 [98] tensor=blk.32.ffn_down_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=122 copy_groups=122 base_mb=150.403 copy_mb=150.403 saved_mb=0.000 [99] tensor=blk.33.ffn_gate_exps.weight backend=CUDA0 used=270 hits=0 misses=270 reuse_pct=0.0 base_groups=118 copy_groups=118 base_mb=143.495 copy_mb=143.495 saved_mb=0.000 [100] tensor=blk.33.ffn_up_exps.weight backend=CUDA0 used=270 hits=0 misses=270 reuse_pct=0.0 base_groups=118 copy_groups=118 base_mb=143.495 copy_mb=143.495 saved_mb=0.000 [101] tensor=blk.33.ffn_down_exps.weight backend=CUDA0 used=270 hits=0 misses=270 reuse_pct=0.0 base_groups=118 copy_groups=118 base_mb=143.495 copy_mb=143.495 saved_mb=0.000 [102] tensor=blk.34.ffn_gate_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=128 copy_groups=128 base_mb=150.406 copy_mb=150.406 saved_mb=0.000 [103] tensor=blk.34.ffn_up_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=128 copy_groups=128 base_mb=150.406 copy_mb=150.406 saved_mb=0.000 [104] tensor=blk.34.ffn_down_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=128 copy_groups=128 base_mb=150.406 copy_mb=150.406 saved_mb=0.000 [105] tensor=blk.35.ffn_gate_exps.weight backend=CUDA0 used=293 hits=0 misses=293 reuse_pct=0.0 base_groups=137 copy_groups=137 base_mb=155.723 copy_mb=155.723 saved_mb=0.000 [106] tensor=blk.35.ffn_up_exps.weight backend=CUDA0 used=293 hits=0 misses=293 reuse_pct=0.0 base_groups=137 copy_groups=137 base_mb=155.723 copy_mb=155.723 saved_mb=0.000 [107] tensor=blk.35.ffn_down_exps.weight backend=CUDA0 used=293 hits=0 misses=293 reuse_pct=0.0 base_groups=137 copy_groups=137 base_mb=155.723 copy_mb=155.723 saved_mb=0.000 [108] tensor=blk.36.ffn_gate_exps.weight backend=CUDA0 used=267 hits=0 misses=267 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=141.904 copy_mb=141.904 saved_mb=0.000 [109] tensor=blk.36.ffn_up_exps.weight backend=CUDA0 used=267 hits=0 misses=267 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=141.904 copy_mb=141.904 saved_mb=0.000 [110] tensor=blk.36.ffn_down_exps.weight backend=CUDA0 used=267 hits=0 misses=267 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=141.904 copy_mb=141.904 saved_mb=0.000 [111] tensor=blk.37.ffn_gate_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=127 copy_groups=127 base_mb=150.405 copy_mb=150.405 saved_mb=0.000 [112] tensor=blk.37.ffn_up_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=127 copy_groups=127 base_mb=150.405 copy_mb=150.405 saved_mb=0.000 [113] tensor=blk.37.ffn_down_exps.weight backend=CUDA0 used=283 hits=0 misses=283 reuse_pct=0.0 base_groups=127 copy_groups=127 base_mb=150.405 copy_mb=150.405 saved_mb=0.000 [114] tensor=blk.38.ffn_gate_exps.weight backend=CUDA0 used=292 hits=0 misses=292 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=155.189 copy_mb=155.189 saved_mb=0.000 [115] tensor=blk.38.ffn_up_exps.weight backend=C…217 tokens truncated…ht backend=CUDA0 used=273 hits=0 misses=273 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=145.097 copy_mb=145.097 saved_mb=0.000 [121] tensor=blk.40.ffn_up_exps.weight backend=CUDA0 used=273 hits=0 misses=273 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=145.097 copy_mb=145.097 saved_mb=0.000 [122] tensor=blk.40.ffn_down_exps.weight backend=CUDA0 used=273 hits=0 misses=273 reuse_pct=0.0 base_groups=136 copy_groups=136 base_mb=145.097 copy_mb=145.097 saved_mb=0.000 [123] tensor=blk.41.ffn_gate_exps.weight backend=CUDA0 used=270 hits=0 misses=270 reuse_pct=0.0 base_groups=141 copy_groups=141 base_mb=143.506 copy_mb=143.506 saved_mb=0.000 [124] tensor=blk.41.ffn_up_exps.weight backend=CUDA0 used=270 hits=0 misses=270 reuse_pct=0.0 base_groups=141 copy_groups=141 base_mb=143.506 copy_mb=143.506 saved_mb=0.000 [125] tensor=blk.41.ffn_down_exps.weight backend=CUDA0 used=270 hits=0 misses=270 reuse_pct=0.0 base_groups=141 copy_groups=141 base_mb=143.506 copy_mb=143.506 saved_mb=0.000 [126] tensor=blk.42.ffn_gate_exps.weight backend=CUDA0 used=251 hits=0 misses=251 reuse_pct=0.0 base_groups=128 copy_groups=128 base_mb=133.406 copy_mb=133.406 saved_mb=0.000 [127] tensor=blk.42.ffn_up_exps.weight backend=CUDA0 used=251 hits=0 misses=251 reuse_pct=0.0 base_groups=128 copy_groups=128 base_mb=133.406 copy_mb=133.406 saved_mb=0.000 [128] tensor=blk.42.ffn_down_exps.weight backend=CUDA0 used=251 hits=0 misses=251 reuse_pct=0.0 base_groups=128 copy_groups=128 base_mb=133.406 copy_mb=133.406 saved_mb=0.000 [129] tensor=blk.43.ffn_gate_exps.weight backend=CUDA0 used=275 hits=0 misses=275 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=146.154 copy_mb=146.154 saved_mb=0.000 [130] tensor=blk.43.ffn_up_exps.weight backend=CUDA0 used=275 hits=0 misses=275 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=146.154 copy_mb=146.154 saved_mb=0.000 [131] tensor=blk.43.ffn_down_exps.weight backend=CUDA0 used=275 hits=0 misses=275 reuse_pct=0.0 base_groups=123 copy_groups=123 base_mb=146.154 copy_mb=146.154 saved_mb=0.000 [132] tensor=blk.44.ffn_gate_exps.weight backend=CUDA0 used=235 hits=0 misses=235 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=124.908 copy_mb=124.908 saved_mb=0.000 [133] tensor=blk.44.ffn_up_exps.weight backend=CUDA0 used=235 hits=0 misses=235 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=124.908 copy_mb=124.908 saved_mb=0.000 [134] tensor=blk.44.ffn_down_exps.weight backend=CUDA0 used=235 hits=0 misses=235 reuse_pct=0.0 base_groups=132 copy_groups=132 base_mb=124.908 copy_mb=124.908 saved_mb=0.000 [135] tensor=blk.45.ffn_gate_exps.weight backend=CUDA0 used=225 hits=0 misses=225 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=119.592 copy_mb=119.592 saved_mb=0.000 [136] tensor=blk.45.ffn_up_exps.weight backend=CUDA0 used=225 hits=0 misses=225 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=119.592 copy_mb=119.592 saved_mb=0.000 [137] tensor=blk.45.ffn_down_exps.weight backend=CUDA0 used=225 hits=0 misses=225 reuse_pct=0.0 base_groups=124 copy_groups=124 base_mb=119.592 copy_mb=119.592 saved_mb=0.000 [138] tensor=blk.46.ffn_gate_exps.weight backend=CUDA0 used=215 hits=0 misses=215 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=114.277 copy_mb=114.277 saved_mb=0.000 [139] tensor=blk.46.ffn_up_exps.weight backend=CUDA0 used=215 hits=0 misses=215 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=114.277 copy_mb=114.277 saved_mb=0.000 [140] tensor=blk.46.ffn_down_exps.weight backend=CUDA0 used=215 hits=0 misses=215 reuse_pct=0.0 base_groups=119 copy_groups=119 base_mb=114.277 copy_mb=114.277 saved_mb=0.000 [141] tensor=blk.0.ffn_gate_exps.weight backend=CUDA0 used=107 hits=106 misses=1 reuse_pct=99.1 base_groups=84 copy_groups=1 base_mb=56.885 copy_mb=0.532 saved_mb=56.353 [142] tensor=blk.0.ffn_up_exps.weight backend=CUDA0 used=107 hits=106 misses=1 reuse_pct=99.1 base_groups=84 copy_groups=1 base_mb=56.885 copy_mb=0.532 saved_mb=56.353 [143] tensor=blk.0.ffn_down_exps.weight backend=CUDA0 used=107 hits=106 misses=1 reuse_pct=99.1 base_groups=84 copy_groups=1 base_mb=56.885 copy_mb=0.532 saved_mb=56.353 [144] tensor=blk.1.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [145] tensor=blk.1.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [146] tensor=blk.1.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [147] tensor=blk.2.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [148] tensor=blk.2.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [149] tensor=blk.2.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [150] tensor=blk.3.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=1 base_mb=5.313 copy_mb=4.250 saved_mb=1.062 [151] tensor=blk.3.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=1 base_mb=5.313 copy_mb=4.250 saved_mb=1.062 [152] tensor=blk.3.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=1 base_mb=5.313 copy_mb=4.250 saved_mb=1.062 [153] tensor=blk.4.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [154] tensor=blk.4.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [155] tensor=blk.4.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [156] tensor=blk.5.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [157] tensor=blk.5.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [158] tensor=blk.5.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [159] tensor=blk.6.ffn_gate_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [160] tensor=blk.6.ffn_up_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [161] tensor=blk.6.ffn_down_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [162] tensor=blk.7.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [163] tensor=blk.7.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [164] tensor=blk.7.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [165] tensor=blk.8.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [166] tensor=blk.8.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [167] tensor=blk.8.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [168] tensor=blk.9.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [169] tensor=blk.9.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [170] tensor=blk.9.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [171] tensor=blk.10.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.721 saved_mb=1.592 [172] tensor=blk.10.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.721 saved_mb=1.592 [173] tensor=blk.10.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.721 saved_mb=1.592 [174] tensor=blk.11.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [175] tensor=blk.11.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [176] tensor=blk.11.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [177] tensor=blk.12.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [178] tensor=blk.12.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [179] tensor=blk.12.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [180] tensor=blk.13.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [181] tensor=blk.13.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [182] tensor=blk.13.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [183] tensor=blk.14.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [184] tensor=blk.14.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [185] tensor=blk.14.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [186] tensor=blk.15.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [187] tensor=blk.15.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [188] tensor=blk.15.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [189] tensor=blk.16.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [190] tensor=blk.16.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [191] tensor=blk.16.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.720 saved_mb=1.593 [192] tensor=blk.17.ffn_gate_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [193] tensor=blk.17.ffn_up_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [194] tensor=blk.17.ffn_down_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [195] tensor=blk.18.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [196] tensor=blk.18.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [197] tensor=blk.18.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [198] tensor=blk.19.ffn_gate_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [199] tensor=blk.19.ffn_up_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [200] tensor=blk.19.ffn_down_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [201] tensor=blk.20.ffn_gate_exps.weight backend=CUDA0 used=10 hits=5 misses=5 reuse_pct=50.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.657 saved_mb=2.656 [202] tensor=blk.20.ffn_up_exps.weight backend=CUDA0 used=10 hits=5 misses=5 reuse_pct=50.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.657 saved_mb=2.656 [203] tensor=blk.20.ffn_down_exps.weight backend=CUDA0 used=10 hits=5 misses=5 reuse_pct=50.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.657 saved_mb=2.656 [204] tensor=blk.21.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=2.127 saved_mb=3.186 [205] tensor=blk.21.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=2.127 saved_mb=3.186 [206] tensor=blk.21.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=2.127 saved_mb=3.186 [207] tensor=blk.22.ffn_gate_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.721 saved_mb=1.592 [208] tensor=blk.22.ffn_up_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.721 saved_mb=1.592 [209] tensor=blk.22.ffn_down_exps.weight backend=CUDA0 used=10 hits=3 misses=7 reuse_pct=30.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.721 saved_mb=1.592 [210] tensor=blk.23.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [211] tensor=blk.23.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [212] tensor=blk.23.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [213] tensor=blk.24.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [214] tensor=blk.24.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [215] tensor=blk.24.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [216] tensor=blk.25.ffn_gate_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [217] tensor=blk.25.ffn_up_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [218] tensor=blk.25.ffn_down_exps.weight backend=CUDA0 used=10 hits=2 misses=8 reuse_pct=20.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=4.251 saved_mb=1.062 [219] tensor=blk.26.ffn_gate_exps.weight backend=CUDA0 used=10 hits=5 misses=5 reuse_pct=50.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.658 saved_mb=2.655 [220] tensor=blk.26.ffn_up_exps.weight backend=CUDA0 used=10 hits=5 misses=5 reuse_pct=50.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.658 saved_mb=2.655 [221] tensor=blk.26.ffn_down_exps.weight backend=CUDA0 used=10 hits=5 misses=5 reuse_pct=50.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.658 saved_mb=2.655 [222] tensor=blk.27.ffn_gate_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [223] tensor=blk.27.ffn_up_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [224] tensor=blk.27.ffn_down_exps.weight backend=CUDA0 used=10 hits=4 misses=6 reuse_pct=40.0 base_groups=1 copy_groups=4 base_mb=5.313 copy_mb=3.189 saved_mb=2.124 [225] tensor=blk.28.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [226] tensor=blk.28.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [227] tensor=blk.28.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [228] tensor=blk.29.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [229] tensor=blk.29.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [230] tensor=blk.29.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=3 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [231] tensor=blk.30.ffn_gate_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [232] tensor=blk.30.ffn_up_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [233] tensor=blk.30.ffn_down_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [234] tensor=blk.31.ffn_gate_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [235] tensor=blk.31.ffn_up_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [236] tensor=blk.31.ffn_down_exps.weight backend=CUDA0 used=10 hits=7 misses=3 reuse_pct=70.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=1.595 saved_mb=3.718 [237] tensor=blk.32.ffn_gate_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [238] tensor=blk.32.ffn_up_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187 [239] tensor=blk.32.ffn_down_exps.weight backend=CUDA0 used=10 hits=6 misses=4 reuse_pct=60.0 base_groups=1 copy_groups=2 base_mb=5.313 copy_mb=2.126 saved_mb=3.187