[D0 M3] samples=1 steps=1000 sinusoidal_PE=False lr=0.001 [D0 M3 step 100/1000] total=0.0007 align=4.0156 span=0.0000 band=0.0001 grad-band(Q/K/Z)=5.83e-05/9.12e-05/NA [D0 M3 step 200/1000] total=0.0006 align=3.9095 span=0.0000 band=0.0001 grad-band(Q/K/Z)=6.81e-06/1.21e-05/NA [D0 M3 step 300/1000] total=0.0005 align=3.8992 span=0.0000 band=0.0001 grad-band(Q/K/Z)=8.11e-05/0.000166/NA [D0 M3 step 400/1000] total=0.0001 align=3.9841 span=0.0000 band=0.0000 grad-band(Q/K/Z)=1.48e-06/2.4e-06/NA [D0 M3 step 500/1000] total=0.0001 align=3.9884 span=0.0000 band=0.0000 grad-band(Q/K/Z)=1.06e-06/1.77e-06/NA [D0 M3 step 600/1000] total=0.0001 align=3.9927 span=0.0000 band=0.0000 grad-band(Q/K/Z)=8.53e-07/1.48e-06/NA [D0 M3 step 700/1000] total=0.0001 align=3.9967 span=0.0000 band=0.0000 grad-band(Q/K/Z)=7.48e-07/1.34e-06/NA [D0 M3 step 800/1000] total=0.0001 align=4.0002 span=0.0000 band=0.0000 grad-band(Q/K/Z)=7.18e-07/1.28e-06/NA [D0 M3 step 900/1000] total=0.0001 align=4.0011 span=0.0000 band=0.0000 grad-band(Q/K/Z)=7.89e-07/1.34e-06/NA [D0 M3 step 1000/1000] total=0.0001 align=4.0008 span=0.0000 band=0.0000 grad-band(Q/K/Z)=1.06e-06/1.7e-06/NA [done D0 M3] final total=0.0001 align=4.0008; metrics=3 [D0 M4] samples=1 steps=1000 sinusoidal_PE=False lr=0.001 [D0 M4 step 100/1000] total=0.0010 align=2.6409 span=0.0000 band=0.0001 grad-band(Q/K/Z)=2.19e-05/2.77e-05/0.00010367462527938187 [D0 M4 step 200/1000] total=0.0008 align=2.6416 span=0.0000 band=0.0001 grad-band(Q/K/Z)=4.86e-05/1.53e-05/8.66068949108012e-05 [D0 M4 step 300/1000] total=0.0007 align=2.6525 span=0.0000 band=0.0001 grad-band(Q/K/Z)=8.89e-06/5.47e-06/2.158571260224562e-05 [D0 M4 step 400/1000] total=0.0007 align=2.6726 span=0.0000 band=0.0001 grad-band(Q/K/Z)=2.33e-06/3.2e-06/1.1499292668304406e-05 [D0 M4 step 500/1000] total=0.0007 align=2.6930 span=0.0000 band=0.0001 grad-band(Q/K/Z)=1.66e-06/2.27e-06/8.077261554717552e-06 [D0 M4 step 600/1000] total=0.0007 align=2.7104 span=0.0000 band=0.0001 grad-band(Q/K/Z)=1.24e-06/1.71e-06/6.006500825606054e-06 [D0 M4 step 700/1000] total=0.0007 align=2.7257 span=0.0000 band=0.0001 grad-band(Q/K/Z)=9.71e-07/1.35e-06/4.651487415685551e-06 [D0 M4 step 800/1000] total=0.0007 align=2.7392 span=0.0000 band=0.0001 grad-band(Q/K/Z)=7.81e-07/1.11e-06/3.715925686265109e-06 [D0 M4 step 900/1000] total=0.0007 align=2.7513 span=0.0000 band=0.0001 grad-band(Q/K/Z)=6.43e-07/9.34e-07/3.0371497814485338e-06 [D0 M4 step 1000/1000] total=0.0007 align=2.7618 span=0.0000 band=0.0001 grad-band(Q/K/Z)=5.44e-07/8.07e-07/2.539506340326625e-06 [done D0 M4] final total=0.0007 align=2.7618; metrics=3 [D1 M3] samples=1 steps=1000 sinusoidal_PE=False lr=0.001 [D1 M3 step 100/1000] total=0.3147 align=0.3147 span=0.0644 band=0.0039 grad-align(Q/K/Z)=0.0166/0.0208/NA [D1 M3 step 200/1000] total=0.2819 align=0.2819 span=0.0518 band=0.0030 grad-align(Q/K/Z)=0.025/0.0356/NA [D1 M3 step 300/1000] total=0.2723 align=0.2723 span=0.0480 band=0.0028 grad-align(Q/K/Z)=0.0308/0.0511/NA [D1 M3 step 400/1000] total=0.2684 align=0.2684 span=0.0457 band=0.0027 grad-align(Q/K/Z)=0.0308/0.0479/NA [D1 M3 step 500/1000] total=0.2661 align=0.2661 span=0.0444 band=0.0026 grad-align(Q/K/Z)=0.025/0.0495/NA [D1 M3 step 600/1000] total=0.2649 align=0.2649 span=0.0432 band=0.0026 grad-align(Q/K/Z)=0.0457/0.113/NA [D1 M3 step 700/1000] total=0.2634 align=0.2634 span=0.0416 band=0.0025 grad-align(Q/K/Z)=0.0192/0.047/NA [D1 M3 step 800/1000] total=0.2627 align=0.2627 span=0.0401 band=0.0024 grad-align(Q/K/Z)=0.0205/0.0313/NA [D1 M3 step 900/1000] total=0.2623 align=0.2623 span=0.0384 band=0.0024 grad-align(Q/K/Z)=0.0247/0.05/NA [D1 M3 step 1000/1000] total=0.2630 align=0.2630 span=0.0392 band=0.0025 grad-align(Q/K/Z)=0.0776/0.22/NA [done D1 M3] final total=0.2630 align=0.2630; metrics=3 [D1 M4_noPE] samples=1 steps=1000 sinusoidal_PE=False lr=0.001 [D1 M4_noPE step 100/1000] total=0.2883 align=0.2883 span=0.1017 band=0.0080 grad-align(Q/K/Z)=0.00228/0.00318/0.004216424189507961 [D1 M4_noPE step 200/1000] total=0.2269 align=0.2269 span=0.0498 band=0.0041 grad-align(Q/K/Z)=0.00118/0.00213/0.002313559642061591 [D1 M4_noPE step 300/1000] total=0.2019 align=0.2019 span=0.0365 band=0.0030 grad-align(Q/K/Z)=0.000522/0.000949/0.0009414505329914391 [D1 M4_noPE step 400/1000] total=0.1950 align=0.1950 span=0.0318 band=0.0026 grad-align(Q/K/Z)=0.000319/0.000536/0.0005940448609180748 [D1 M4_noPE step 500/1000] total=0.1915 align=0.1915 span=0.0293 band=0.0025 grad-align(Q/K/Z)=0.000635/0.00405/0.0008793329470790923 [D1 M4_noPE step 600/1000] total=0.1890 align=0.1890 span=0.0280 band=0.0024 grad-align(Q/K/Z)=0.000517/0.00358/0.000659892160911113 [D1 M4_noPE step 700/1000] total=0.1869 align=0.1869 span=0.0271 band=0.0023 grad-align(Q/K/Z)=0.000668/0.00481/0.0008425512933172286 [D1 M4_noPE step 800/1000] total=0.1850 align=0.1850 span=0.0265 band=0.0023 grad-align(Q/K/Z)=0.000353/0.00158/0.0004897429025731981 [D1 M4_noPE step 900/1000] total=0.1840 align=0.1840 span=0.0258 band=0.0022 grad-align(Q/K/Z)=0.000835/0.00496/0.0009656989714130759 [D1 M4_noPE step 1000/1000] total=0.1832 align=0.1832 span=0.0251 band=0.0022 grad-align(Q/K/Z)=0.00106/0.00433/0.001278208801522851 [done D1 M4_noPE] final total=0.1832 align=0.1832; metrics=3 [D1 M4_sinPE] samples=1 steps=1000 sinusoidal_PE=True lr=0.001 [D1 M4_sinPE step 100/1000] total=0.2745 align=0.2745 span=0.0882 band=0.0067 grad-align(Q/K/Z)=0.00297/0.00435/0.0032130267936736345 [D1 M4_sinPE step 200/1000] total=0.2118 align=0.2118 span=0.0447 band=0.0032 grad-align(Q/K/Z)=0.00107/0.00175/0.001607082667760551 [D1 M4_sinPE step 300/1000] total=0.1967 align=0.1967 span=0.0337 band=0.0026 grad-align(Q/K/Z)=0.000582/0.000954/0.0008562988368794322 [D1 M4_sinPE step 400/1000] total=0.1902 align=0.1902 span=0.0296 band=0.0023 grad-align(Q/K/Z)=0.000328/0.000718/0.0005115939420647919 [D1 M4_sinPE step 500/1000] total=0.1872 align=0.1872 span=0.0272 band=0.0022 grad-align(Q/K/Z)=0.00029/0.000911/0.00046330280019901693 [D1 M4_sinPE step 600/1000] total=0.1851 align=0.1851 span=0.0252 band=0.0021 grad-align(Q/K/Z)=0.00186/0.00929/0.0019628412555903196 [D1 M4_sinPE step 700/1000] total=0.1835 align=0.1835 span=0.0235 band=0.0020 grad-align(Q/K/Z)=0.000776/0.00248/0.0006809488986618817 [D1 M4_sinPE step 800/1000] total=0.1828 align=0.1828 span=0.0233 band=0.0020 grad-align(Q/K/Z)=0.00177/0.00702/0.002244245493784547 [D1 M4_sinPE step 900/1000] total=0.1822 align=0.1822 span=0.0229 band=0.0020 grad-align(Q/K/Z)=0.0015/0.00715/0.0013185743009671569 [D1 M4_sinPE step 1000/1000] total=0.1817 align=0.1817 span=0.0227 band=0.0020 grad-align(Q/K/Z)=0.00147/0.00721/0.0012832485372200608 [done D1 M4_sinPE] final total=0.1817 align=0.1817; metrics=3 [D2 M3] samples=100 steps=500 sinusoidal_PE=False lr=0.001 [D2 M3 step 100/500] total=0.7070 align=0.7070 span=0.2982 band=0.0197 grad-align(Q/K/Z)=0.0544/0.0544/NA [D2 M3 step 200/500] total=0.5588 align=0.5588 span=0.2068 band=0.0079 grad-align(Q/K/Z)=0.0866/0.0974/NA [D2 M3 step 300/500] total=0.5614 align=0.5614 span=0.1948 band=0.0119 grad-align(Q/K/Z)=0.0565/0.0439/NA [D2 M3 step 400/500] total=0.5531 align=0.5531 span=0.1878 band=0.0117 grad-align(Q/K/Z)=0.0504/0.0364/NA [D2 M3 step 500/500] total=0.5194 align=0.5194 span=0.1950 band=0.0098 grad-align(Q/K/Z)=0.0576/0.0496/NA [done D2 M3] final total=0.5194 align=0.5194; metrics=300 [D2 M4_sinPE] samples=100 steps=500 sinusoidal_PE=True lr=0.001 [D2 M4_sinPE step 100/500] total=0.7284 align=0.7284 span=0.2934 band=0.0254 grad-align(Q/K/Z)=0.0157/0.0175/0.015301910229027271 [D2 M4_sinPE step 200/500] total=0.6799 align=0.6799 span=0.2622 band=0.0227 grad-align(Q/K/Z)=0.0163/0.0208/0.011812740005552769 [D2 M4_sinPE step 300/500] total=0.6770 align=0.6770 span=0.2819 band=0.0250 grad-align(Q/K/Z)=0.0158/0.0214/0.013132783584296703 [D2 M4_sinPE step 400/500] total=0.6383 align=0.6383 span=0.2607 band=0.0223 grad-align(Q/K/Z)=0.00825/0.0161/0.007149462588131428 [D2 M4_sinPE step 500/500] total=0.6222 align=0.6222 span=0.2545 band=0.0215 grad-align(Q/K/Z)=0.0108/0.0213/0.00810039509087801 [done D2 M4_sinPE] final total=0.6222 align=0.6222; metrics=300 [D3 M3] samples=100 steps=500 sinusoidal_PE=False lr=0.001 [D3 M3 step 100/500] total=4.1864 align=0.9900 span=0.4564 band=0.0376 grad-align(Q/K/Z)=0.0581/0.0607/NA [D3 M3 step 200/500] total=3.2114 align=0.9275 span=0.4127 band=0.0214 grad-align(Q/K/Z)=0.0618/0.069/NA [D3 M3 step 300/500] total=2.9262 align=0.9328 span=0.3966 band=0.0300 grad-align(Q/K/Z)=0.0504/0.0546/NA [D3 M3 step 400/500] total=2.5046 align=0.9455 span=0.4341 band=0.0324 grad-align(Q/K/Z)=0.0404/0.0427/NA [D3 M3 step 500/500] total=2.3617 align=0.8575 span=0.3872 band=0.0244 grad-align(Q/K/Z)=0.0312/0.0323/NA [done D3 M3] final total=2.3617 align=0.8575; metrics=300 [D3 M4_sinPE] samples=100 steps=500 sinusoidal_PE=True lr=0.001 [D3 M4_sinPE step 100/500] total=4.6514 align=0.8535 span=0.3465 band=0.0307 grad-align(Q/K/Z)=0.0136/0.0326/0.012736599892377853 [D3 M4_sinPE step 200/500] total=3.6226 align=0.8372 span=0.3601 band=0.0305 grad-align(Q/K/Z)=0.0227/0.0515/0.018048657104372978 [D3 M4_sinPE step 300/500] total=3.3903 align=0.8115 span=0.3256 band=0.0298 grad-align(Q/K/Z)=0.0153/0.0393/0.017227167263627052 [D3 M4_sinPE step 400/500] total=3.0099 align=0.7362 span=0.2919 band=0.0249 grad-align(Q/K/Z)=0.0118/0.0297/0.01380767859518528 [D3 M4_sinPE step 500/500] total=2.5226 align=0.7414 span=0.3080 band=0.0257 grad-align(Q/K/Z)=0.0128/0.0334/0.019335519522428513 [done D3 M4_sinPE] final total=2.5226 align=0.7414; metrics=300 [all done] elapsed=103.7s output=/home/gloamxun/modeling_zhaocui/Q1/outputs/alignment_debug