Skip to content

Commit 66dd780

Browse files
jiho264jiho264
committed
update test logs
1 parent b09fc91 commit 66dd780

32 files changed

+1487
-98
lines changed

‎logs/BitLog2_Half_16/deit_base.log

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -943,7 +943,7 @@ self.delta: 217.0
943943
0.1106, 0.1475, 0.2212, 0.2949, 0.4424, 0.5899], device='cuda:0')
944944
Performing scale reparameterization ...
945945
Validating ...
946-
Test: [0/250] Time 4.344 (4.344) Loss 0.6282 (0.6282) Prec@1 89.000 (89.000) Prec@5 98.000 (98.000)
947-
Test: [100/250] Time 1.821 (1.845) Loss 1.5258 (0.9715) Prec@1 66.500 (81.386) Prec@5 88.500 (95.342)
946+
Test: [0/250] Time 4.215 (4.215) Loss 0.6282 (0.6282) Prec@1 89.000 (89.000) Prec@5 98.000 (98.000)
947+
Test: [100/250] Time 1.821 (1.844) Loss 1.5258 (0.9715) Prec@1 66.500 (81.386) Prec@5 88.500 (95.342)
948948
Test: [200/250] Time 1.819 (1.832) Loss 0.6791 (1.1897) Prec@1 90.000 (76.403) Prec@5 98.500 (92.368)
949-
* Prec@1 75.594 Prec@5 92.026 Time 457.457
949+
* Prec@1 75.594 Prec@5 92.026 Time 457.408

‎logs/BitLog2_Half_16/deit_small.log

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -880,7 +880,7 @@ self.delta: 217.0
880880
device='cuda:0')
881881
Performing scale reparameterization ...
882882
Validating ...
883-
Test: [0/250] Time 3.300 (3.300) Loss 0.7690 (0.7690) Prec@1 81.500 (81.500) Prec@5 98.000 (98.000)
884-
Test: [100/250] Time 0.840 (0.864) Loss 1.8733 (1.0682) Prec@1 55.500 (75.767) Prec@5 86.500 (93.307)
885-
Test: [200/250] Time 0.840 (0.852) Loss 0.9113 (1.3726) Prec@1 84.000 (70.229) Prec@5 95.500 (89.423)
886-
* Prec@1 69.432 Prec@5 89.056 Time 212.437
883+
Test: [0/250] Time 3.385 (3.385) Loss 0.7690 (0.7690) Prec@1 81.500 (81.500) Prec@5 98.000 (98.000)
884+
Test: [100/250] Time 0.840 (0.865) Loss 1.8733 (1.0682) Prec@1 55.500 (75.767) Prec@5 86.500 (93.307)
885+
Test: [200/250] Time 0.840 (0.853) Loss 0.9113 (1.3726) Prec@1 84.000 (70.229) Prec@5 95.500 (89.423)
886+
* Prec@1 69.432 Prec@5 89.056 Time 212.565

‎logs/BitLog2_Half_16/deit_tiny.log

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1006,7 +1006,7 @@ self.delta: 217.0
10061006
0.1106, 0.1475, 0.2212, 0.2949, 0.4424, 0.5899], device='cuda:0')
10071007
Performing scale reparameterization ...
10081008
Validating ...
1009-
Test: [0/250] Time 2.966 (2.966) Loss 1.1909 (1.1909) Prec@1 74.000 (74.000) Prec@5 94.500 (94.500)
1010-
Test: [100/250] Time 0.418 (0.444) Loss 2.6884 (1.6558) Prec@1 41.500 (65.228) Prec@5 72.500 (87.827)
1011-
Test: [200/250] Time 0.418 (0.431) Loss 1.3788 (2.0576) Prec@1 80.000 (58.249) Prec@5 91.000 (81.843)
1012-
* Prec@1 57.664 Prec@5 81.360 Time 107.238
1009+
Test: [0/250] Time 2.767 (2.767) Loss 1.1909 (1.1909) Prec@1 74.000 (74.000) Prec@5 94.500 (94.500)
1010+
Test: [100/250] Time 0.418 (0.442) Loss 2.6884 (1.6558) Prec@1 41.500 (65.228) Prec@5 72.500 (87.827)
1011+
Test: [200/250] Time 0.418 (0.430) Loss 1.3788 (2.0576) Prec@1 80.000 (58.249) Prec@5 91.000 (81.843)
1012+
* Prec@1 57.664 Prec@5 81.360 Time 106.979

‎logs/BitLog2_Half_16/vit_base.log

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -943,7 +943,7 @@ self.delta: 217.0
943943
device='cuda:0')
944944
Performing scale reparameterization ...
945945
Validating ...
946-
Test: [0/250] Time 4.270 (4.270) Loss 0.7381 (0.7381) Prec@1 79.500 (79.500) Prec@5 98.500 (98.500)
947-
Test: [100/250] Time 1.819 (1.844) Loss 1.2946 (1.1717) Prec@1 71.500 (71.847) Prec@5 89.500 (91.450)
946+
Test: [0/250] Time 4.206 (4.206) Loss 0.7381 (0.7381) Prec@1 79.500 (79.500) Prec@5 98.500 (98.500)
947+
Test: [100/250] Time 1.820 (1.844) Loss 1.2946 (1.1717) Prec@1 71.500 (71.847) Prec@5 89.500 (91.450)
948948
Test: [200/250] Time 1.820 (1.832) Loss 0.8806 (1.4349) Prec@1 80.000 (67.816) Prec@5 96.000 (88.045)
949-
* Prec@1 67.482 Prec@5 87.864 Time 457.387
949+
* Prec@1 67.482 Prec@5 87.864 Time 457.417

‎logs/BitLog2_Half_16/vit_small.log

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -880,7 +880,7 @@ self.delta: 218.0
880880
device='cuda:0')
881881
Performing scale reparameterization ...
882882
Validating ...
883-
Test: [0/250] Time 3.149 (3.149) Loss 0.8334 (0.8334) Prec@1 77.000 (77.000) Prec@5 95.000 (95.000)
884-
Test: [100/250] Time 0.840 (0.862) Loss 1.5188 (1.1890) Prec@1 56.500 (69.490) Prec@5 89.000 (90.688)
883+
Test: [0/250] Time 2.945 (2.945) Loss 0.8334 (0.8334) Prec@1 77.000 (77.000) Prec@5 95.000 (95.000)
884+
Test: [100/250] Time 0.840 (0.861) Loss 1.5188 (1.1890) Prec@1 56.500 (69.490) Prec@5 89.000 (90.688)
885885
Test: [200/250] Time 0.840 (0.851) Loss 1.0661 (1.4629) Prec@1 73.500 (65.107) Prec@5 91.500 (86.759)
886-
* Prec@1 64.580 Prec@5 86.384 Time 212.279
886+
* Prec@1 64.580 Prec@5 86.384 Time 212.179

‎logs/BitLog2_Half_17/deit_base.log

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1648,7 +1648,7 @@ self.delta: 326.0
16481648
device='cuda:0')
16491649
Performing scale reparameterization ...
16501650
Validating ...
1651-
Test: [0/250] Time 4.267 (4.267) Loss 0.6341 (0.6341) Prec@1 87.500 (87.500) Prec@5 99.000 (99.000)
1652-
Test: [100/250] Time 1.821 (1.845) Loss 1.5027 (0.9691) Prec@1 67.500 (81.505) Prec@5 88.000 (95.421)
1653-
Test: [200/250] Time 1.819 (1.832) Loss 0.6430 (1.1892) Prec@1 91.500 (76.619) Prec@5 98.500 (92.495)
1654-
* Prec@1 75.836 Prec@5 92.164 Time 457.516
1651+
Test: [0/250] Time 4.231 (4.231) Loss 0.6341 (0.6341) Prec@1 87.500 (87.500) Prec@5 99.000 (99.000)
1652+
Test: [100/250] Time 1.822 (1.844) Loss 1.5027 (0.9691) Prec@1 67.500 (81.505) Prec@5 88.000 (95.421)
1653+
Test: [200/250] Time 1.820 (1.832) Loss 0.6430 (1.1892) Prec@1 91.500 (76.619) Prec@5 98.500 (92.495)
1654+
* Prec@1 75.836 Prec@5 92.164 Time 457.424

‎logs/BitLog2_Half_17/deit_small.log

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1648,7 +1648,7 @@ self.delta: 328.0
16481648
device='cuda:0')
16491649
Performing scale reparameterization ...
16501650
Validating ...
1651-
Test: [0/250] Time 3.380 (3.380) Loss 0.7813 (0.7813) Prec@1 84.000 (84.000) Prec@5 98.000 (98.000)
1652-
Test: [100/250] Time 0.840 (0.865) Loss 1.9291 (1.1192) Prec@1 54.500 (75.683) Prec@5 86.000 (93.124)
1651+
Test: [0/250] Time 3.273 (3.273) Loss 0.7813 (0.7813) Prec@1 84.000 (84.000) Prec@5 98.000 (98.000)
1652+
Test: [100/250] Time 0.840 (0.864) Loss 1.9291 (1.1192) Prec@1 54.500 (75.683) Prec@5 86.000 (93.124)
16531653
Test: [200/250] Time 0.840 (0.852) Loss 0.8887 (1.3988) Prec@1 82.500 (70.192) Prec@5 95.500 (89.537)
1654-
* Prec@1 69.554 Prec@5 89.132 Time 212.535
1654+
* Prec@1 69.554 Prec@5 89.132 Time 212.468

‎logs/BitLog2_Half_17/deit_tiny.log

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1712,7 +1712,7 @@ self.delta: 326.0
17121712
device='cuda:0')
17131713
Performing scale reparameterization ...
17141714
Validating ...
1715-
Test: [0/250] Time 2.962 (2.962) Loss 1.1925 (1.1925) Prec@1 74.500 (74.500) Prec@5 94.000 (94.000)
1716-
Test: [100/250] Time 0.418 (0.444) Loss 2.6272 (1.6528) Prec@1 41.500 (66.104) Prec@5 74.000 (88.490)
1717-
Test: [200/250] Time 0.418 (0.431) Loss 1.4412 (2.0558) Prec@1 79.500 (59.037) Prec@5 90.000 (82.234)
1718-
* Prec@1 58.346 Prec@5 81.730 Time 107.202
1715+
Test: [0/250] Time 2.679 (2.679) Loss 1.1925 (1.1925) Prec@1 74.500 (74.500) Prec@5 94.000 (94.000)
1716+
Test: [100/250] Time 0.418 (0.441) Loss 2.6272 (1.6528) Prec@1 41.500 (66.104) Prec@5 74.000 (88.490)
1717+
Test: [200/250] Time 0.418 (0.430) Loss 1.4412 (2.0558) Prec@1 79.500 (59.037) Prec@5 90.000 (82.234)
1718+
* Prec@1 58.346 Prec@5 81.730 Time 106.888

‎logs/BitLog2_Half_17/vit_base.log

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -1648,7 +1648,7 @@ self.delta: 325.0
16481648
device='cuda:0')
16491649
Performing scale reparameterization ...
16501650
Validating ...
1651-
Test: [0/250] Time 4.285 (4.285) Loss 0.7226 (0.7226) Prec@1 81.000 (81.000) Prec@5 99.000 (99.000)
1651+
Test: [0/250] Time 4.283 (4.283) Loss 0.7226 (0.7226) Prec@1 81.000 (81.000) Prec@5 99.000 (99.000)
16521652
Test: [100/250] Time 1.820 (1.845) Loss 1.0870 (1.1063) Prec@1 75.500 (73.054) Prec@5 92.000 (92.079)
1653-
Test: [200/250] Time 1.820 (1.832) Loss 0.7462 (1.3450) Prec@1 84.500 (69.264) Prec@5 97.500 (88.948)
1654-
* Prec@1 68.900 Prec@5 88.804 Time 457.509
1653+
Test: [200/250] Time 1.819 (1.832) Loss 0.7462 (1.3450) Prec@1 84.500 (69.264) Prec@5 97.500 (88.948)
1654+
* Prec@1 68.900 Prec@5 88.804 Time 457.477

‎logs/BitLog2_Half_17/vit_small.log

Lines changed: 4 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -1648,7 +1648,7 @@ self.delta: 327.0
16481648
device='cuda:0')
16491649
Performing scale reparameterization ...
16501650
Validating ...
1651-
Test: [0/250] Time 3.315 (3.315) Loss 0.7889 (0.7889) Prec@1 80.500 (80.500) Prec@5 95.000 (95.000)
1652-
Test: [100/250] Time 0.840 (0.864) Loss 1.3933 (1.1430) Prec@1 60.500 (70.470) Prec@5 90.000 (91.059)
1653-
Test: [200/250] Time 0.840 (0.852) Loss 0.8526 (1.3983) Prec@1 81.500 (66.251) Prec@5 95.500 (87.701)
1654-
* Prec@1 65.874 Prec@5 87.406 Time 212.499
1651+
Test: [0/250] Time 3.122 (3.122) Loss 0.7889 (0.7889) Prec@1 80.500 (80.500) Prec@5 95.000 (95.000)
1652+
Test: [100/250] Time 0.840 (0.863) Loss 1.3933 (1.1430) Prec@1 60.500 (70.470) Prec@5 90.000 (91.059)
1653+
Test: [200/250] Time 0.840 (0.851) Loss 0.8526 (1.3983) Prec@1 81.500 (66.251) Prec@5 95.500 (87.701)
1654+
* Prec@1 65.874 Prec@5 87.406 Time 212.327

‎logs/BitLog2_Single_16/deit_base.log

Lines changed: 130 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,130 @@
1+
Namespace(model='deit_base', dataset='dataset/imagenet/', calib_batchsize=32, val_batchsize=200, num_workers=8, device='cuda', print_freq=100, seed=0, w_bits=4, a_bits=4, log_quant_scheme='BitLog2_Single_16')
2+
Building dataloader ...
3+
Building model ...
4+
Performing initial quantization ...
5+
self.delta: 22492.0
6+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
7+
70, 80, 90, 100, 110, 120, 130, 140],
8+
device='cuda:0', dtype=torch.int32)
9+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
10+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
11+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
12+
16 tensor([0.0000e+00, 4.4460e-05, 8.8921e-05, 1.7784e-04, 3.5568e-04, 7.1136e-04,
13+
1.4227e-03, 2.8455e-03, 5.6909e-03, 1.1382e-02, 2.2764e-02, 4.5527e-02,
14+
9.1055e-02, 1.8211e-01, 3.6422e-01, 7.2844e-01], device='cuda:0')
15+
self.delta: 24500.0
16+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
17+
70, 80, 90, 100, 110, 120, 130, 140],
18+
device='cuda:0', dtype=torch.int32)
19+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
20+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
21+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
22+
16 tensor([0.0000e+00, 4.0816e-05, 8.1633e-05, 1.6327e-04, 3.2653e-04, 6.5306e-04,
23+
1.3061e-03, 2.6122e-03, 5.2245e-03, 1.0449e-02, 2.0898e-02, 4.1796e-02,
24+
8.3592e-02, 1.6718e-01, 3.3437e-01, 6.6873e-01], device='cuda:0')
25+
self.delta: 24528.0
26+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
27+
70, 80, 90, 100, 110, 120, 130, 140],
28+
device='cuda:0', dtype=torch.int32)
29+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
30+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
31+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
32+
16 tensor([0.0000e+00, 4.0770e-05, 8.1539e-05, 1.6308e-04, 3.2616e-04, 6.5232e-04,
33+
1.3046e-03, 2.6093e-03, 5.2185e-03, 1.0437e-02, 2.0874e-02, 4.1748e-02,
34+
8.3496e-02, 1.6699e-01, 3.3399e-01, 6.6797e-01], device='cuda:0')
35+
self.delta: 21922.0
36+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
37+
70, 80, 90, 100, 110, 120, 130, 140],
38+
device='cuda:0', dtype=torch.int32)
39+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
40+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
41+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
42+
16 tensor([0.0000e+00, 4.5616e-05, 9.1233e-05, 1.8247e-04, 3.6493e-04, 7.2986e-04,
43+
1.4597e-03, 2.9194e-03, 5.8389e-03, 1.1678e-02, 2.3356e-02, 4.6711e-02,
44+
9.3422e-02, 1.8684e-01, 3.7369e-01, 7.4738e-01], device='cuda:0')
45+
self.delta: 23644.0
46+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
47+
70, 80, 90, 100, 110, 120, 130, 140],
48+
device='cuda:0', dtype=torch.int32)
49+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
50+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
51+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
52+
16 tensor([0.0000e+00, 4.2294e-05, 8.4588e-05, 1.6918e-04, 3.3835e-04, 6.7670e-04,
53+
1.3534e-03, 2.7068e-03, 5.4136e-03, 1.0827e-02, 2.1655e-02, 4.3309e-02,
54+
8.6618e-02, 1.7324e-01, 3.4647e-01, 6.9295e-01], device='cuda:0')
55+
self.delta: 24428.0
56+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
57+
70, 80, 90, 100, 110, 120, 130, 140],
58+
device='cuda:0', dtype=torch.int32)
59+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
60+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
61+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
62+
16 tensor([0.0000e+00, 4.0937e-05, 8.1873e-05, 1.6375e-04, 3.2749e-04, 6.5499e-04,
63+
1.3100e-03, 2.6199e-03, 5.2399e-03, 1.0480e-02, 2.0960e-02, 4.1919e-02,
64+
8.3838e-02, 1.6768e-01, 3.3535e-01, 6.7071e-01], device='cuda:0')
65+
self.delta: 24110.0
66+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
67+
70, 80, 90, 100, 110, 120, 130, 140],
68+
device='cuda:0', dtype=torch.int32)
69+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
70+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
71+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
72+
16 tensor([0.0000e+00, 4.1477e-05, 8.2953e-05, 1.6591e-04, 3.3181e-04, 6.6363e-04,
73+
1.3273e-03, 2.6545e-03, 5.3090e-03, 1.0618e-02, 2.1236e-02, 4.2472e-02,
74+
8.4944e-02, 1.6989e-01, 3.3978e-01, 6.7955e-01], device='cuda:0')
75+
self.delta: 24275.0
76+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
77+
70, 80, 90, 100, 110, 120, 130, 140],
78+
device='cuda:0', dtype=torch.int32)
79+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
80+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
81+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
82+
16 tensor([0.0000e+00, 4.1195e-05, 8.2389e-05, 1.6478e-04, 3.2956e-04, 6.5911e-04,
83+
1.3182e-03, 2.6365e-03, 5.2729e-03, 1.0546e-02, 2.1092e-02, 4.2183e-02,
84+
8.4367e-02, 1.6873e-01, 3.3747e-01, 6.7493e-01], device='cuda:0')
85+
self.delta: 24151.0
86+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
87+
70, 80, 90, 100, 110, 120, 130, 140],
88+
device='cuda:0', dtype=torch.int32)
89+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
90+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
91+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
92+
16 tensor([0.0000e+00, 4.1406e-05, 8.2812e-05, 1.6562e-04, 3.3125e-04, 6.6250e-04,
93+
1.3250e-03, 2.6500e-03, 5.3000e-03, 1.0600e-02, 2.1200e-02, 4.2400e-02,
94+
8.4800e-02, 1.6960e-01, 3.3920e-01, 6.7840e-01], device='cuda:0')
95+
self.delta: 24589.0
96+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
97+
70, 80, 90, 100, 110, 120, 130, 140],
98+
device='cuda:0', dtype=torch.int32)
99+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
100+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
101+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
102+
16 tensor([0.0000e+00, 4.0669e-05, 8.1337e-05, 1.6267e-04, 3.2535e-04, 6.5070e-04,
103+
1.3014e-03, 2.6028e-03, 5.2056e-03, 1.0411e-02, 2.0822e-02, 4.1645e-02,
104+
8.3289e-02, 1.6658e-01, 3.3316e-01, 6.6631e-01], device='cuda:0')
105+
self.delta: 24586.0
106+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
107+
70, 80, 90, 100, 110, 120, 130, 140],
108+
device='cuda:0', dtype=torch.int32)
109+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
110+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
111+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
112+
16 tensor([0.0000e+00, 4.0674e-05, 8.1347e-05, 1.6269e-04, 3.2539e-04, 6.5078e-04,
113+
1.3016e-03, 2.6031e-03, 5.2062e-03, 1.0412e-02, 2.0825e-02, 4.1650e-02,
114+
8.3299e-02, 1.6660e-01, 3.3320e-01, 6.6640e-01], device='cuda:0')
115+
self.delta: 24545.0
116+
16 tensor([-100000, 0, 10, 20, 30, 40, 50, 60,
117+
70, 80, 90, 100, 110, 120, 130, 140],
118+
device='cuda:0', dtype=torch.int32)
119+
16 tensor([0.0000e+00, 1.0000e+00, 2.0000e+00, 4.0000e+00, 8.0000e+00, 1.6000e+01,
120+
3.2000e+01, 6.4000e+01, 1.2800e+02, 2.5600e+02, 5.1200e+02, 1.0240e+03,
121+
2.0480e+03, 4.0960e+03, 8.1920e+03, 1.6384e+04], device='cuda:0')
122+
16 tensor([0.0000e+00, 4.0741e-05, 8.1483e-05, 1.6297e-04, 3.2593e-04, 6.5186e-04,
123+
1.3037e-03, 2.6075e-03, 5.2149e-03, 1.0430e-02, 2.0860e-02, 4.1719e-02,
124+
8.3439e-02, 1.6688e-01, 3.3375e-01, 6.6751e-01], device='cuda:0')
125+
Performing scale reparameterization ...
126+
Validating ...
127+
Test: [0/250] Time 4.031 (4.031) Loss 0.5960 (0.5960) Prec@1 89.500 (89.500) Prec@5 99.000 (99.000)
128+
Test: [100/250] Time 1.841 (1.864) Loss 1.4360 (0.9697) Prec@1 68.500 (81.079) Prec@5 89.500 (95.277)
129+
Test: [200/250] Time 1.840 (1.852) Loss 0.6625 (1.1916) Prec@1 87.500 (76.239) Prec@5 98.500 (92.348)
130+
* Prec@1 75.458 Prec@5 92.044 Time 462.533

0 commit comments

Comments
 (0)