vision-bnn-benchmarks-hf
/
checkpoints
/shallow002_nas_2b8b_noSAM_4lyrDeep
/statistics_training_checkpoint
Top: | |
epoch | |
extras | |
state_dict | |
arch | |
------------------------------------- | |
arch: ai85nascifarnet | |
------------------------------------- | |
extras: None | |
------------------------------------- | |
state_dict: | |
conv1_1 | |
output_shift: [-1.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 64 , [64] | |
# of unique elements: 64 | |
min, max, mean: -0.11245388 , 0.10340247 , 0.001591533 | |
weight | |
total # of elements, shape: 1728 , [64, 3, 3, 3] | |
# of unique elements: 1728 | |
min, max, mean: -0.7012884 , 0.76133716 , -0.00029134436 | |
conv1_2 | |
output_shift: [-1.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 32 , [32] | |
# of unique elements: 32 | |
min, max, mean: -0.40471938 , 0.22852542 , 0.08551041 | |
weight | |
total # of elements, shape: 2048 , [32, 64, 1, 1] | |
# of unique elements: 2048 | |
min, max, mean: -0.59585965 , 0.46755692 , -0.0164743 | |
conv1_3 | |
output_shift: [-2.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 64 , [64] | |
# of unique elements: 64 | |
min, max, mean: -0.43269184 , 0.4354972 , 0.075772524 | |
weight | |
total # of elements, shape: 18432 , [64, 32, 3, 3] | |
# of unique elements: 18431 | |
min, max, mean: -0.55321246 , 0.3909786 , -0.004218949 | |
conv2_1 | |
output_shift: [-3.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 32 , [32] | |
# of unique elements: 32 | |
min, max, mean: -0.3348623 , 0.36877617 , 0.064229615 | |
weight | |
total # of elements, shape: 18432 , [32, 64, 3, 3] | |
# of unique elements: 18431 | |
min, max, mean: -0.23059952 , 0.1768544 , -0.0002033399 | |
conv2_2 | |
output_shift: [-0.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [2.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 64 , [64] | |
# of unique elements: 64 | |
min, max, mean: -0.27203873 , 0.44126716 , 0.070215635 | |
weight | |
total # of elements, shape: 2048 , [64, 32, 1, 1] | |
# of unique elements: 2048 | |
min, max, mean: -1.081907 , 1.2835066 , -0.022990013 | |
conv3_1 | |
output_shift: [-3.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [2.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 128 , [128] | |
# of unique elements: 128 | |
min, max, mean: -0.65168923 , 0.6424291 , 0.046436235 | |
weight | |
total # of elements, shape: 73728 , [128, 64, 3, 3] | |
# of unique elements: 73691 | |
min, max, mean: -0.1757474 , 0.16606377 , 4.4539163e-05 | |
conv3_2 | |
output_shift: [-1.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [2.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 128 , [128] | |
# of unique elements: 128 | |
min, max, mean: -0.37927154 , 0.72376007 , 0.15195552 | |
weight | |
total # of elements, shape: 16384 , [128, 128, 1, 1] | |
# of unique elements: 16380 | |
min, max, mean: -0.61208266 , 0.55124444 , -0.01141741 | |
conv4_1 | |
output_shift: [-3.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 64 , [64] | |
# of unique elements: 64 | |
min, max, mean: -0.62133706 , 0.669571 , -0.09196919 | |
weight | |
total # of elements, shape: 73728 , [64, 128, 3, 3] | |
# of unique elements: 73693 | |
min, max, mean: -0.11380878 , 0.14027645 , 0.00026837227 | |
conv4_2 | |
output_shift: [-2.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 128 , [128] | |
# of unique elements: 128 | |
min, max, mean: -0.3281726 , 0.19478674 , 0.017257225 | |
weight | |
total # of elements, shape: 73728 , [128, 64, 3, 3] | |
# of unique elements: 73688 | |
min, max, mean: -0.2758857 , 0.2754693 , 0.00037729595 | |
conv5_1 | |
output_shift: [-1.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 128 , [128] | |
# of unique elements: 128 | |
min, max, mean: -0.26617277 , 0.6138764 , 0.112761155 | |
weight | |
total # of elements, shape: 16384 , [128, 128, 1, 1] | |
# of unique elements: 16382 | |
min, max, mean: -0.44473115 , 0.56870544 , -0.00018005852 | |
fc | |
output_shift: [1.] | |
adjust_output_shift: [1.] | |
quantize_activation: [1.] | |
shift_quantile: [0.985] | |
weight bits: [8.] | |
bias_bits: [8.] | |
bias | |
total # of elements, shape: 100 , [100] | |
# of unique elements: 100 | |
min, max, mean: -0.17998613 , 0.24179709 , -0.00045737333 | |
weight | |
total # of elements, shape: 51200 , [100, 512] | |
# of unique elements: 51181 | |
min, max, mean: -2.0958986 , 1.6769325 , -0.15933196 | |