vision-bnn-benchmarks-hf
/
checkpoints
/shallow003_nas_2b8b_noSAM_3lyrDeep
/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: -0.10468257 , 0.075563245 , 0.0014071614 | |
weight | |
total # of elements, shape: 1728 , [64, 3, 3, 3] | |
# of unique elements: 1728 | |
min, max, mean: -0.8909477 , 0.7126475 , -0.0017649194 | |
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: -0.37964904 , 0.22851238 , 0.09636331 | |
weight | |
total # of elements, shape: 2048 , [32, 64, 1, 1] | |
# of unique elements: 2048 | |
min, max, mean: -0.6206478 , 0.48035166 , -0.018839637 | |
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: -0.25792903 , 0.4632549 , 0.08173839 | |
weight | |
total # of elements, shape: 18432 , [64, 32, 3, 3] | |
# of unique elements: 18426 | |
min, max, mean: -0.5974413 , 0.44250032 , -0.005892037 | |
conv2_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: 32 , [32] | |
# of unique elements: 32 | |
min, max, mean: -0.45214012 , 0.36729175 , 0.0695677 | |
weight | |
total # of elements, shape: 18432 , [32, 64, 3, 3] | |
# of unique elements: 18428 | |
min, max, mean: -0.1820527 , 0.17661493 , -5.3904045e-05 | |
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: -0.3718062 , 0.43213025 , 0.06030646 | |
weight | |
total # of elements, shape: 2048 , [64, 32, 1, 1] | |
# of unique elements: 2048 | |
min, max, mean: -1.2500217 , 0.9969034 , -0.020823663 | |
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: -0.54298604 , 0.8285807 , 0.08188485 | |
weight | |
total # of elements, shape: 73728 , [128, 64, 3, 3] | |
# of unique elements: 73684 | |
min, max, mean: -0.17556906 , 0.23762546 , -0.0002926274 | |
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: -0.4131362 , 0.6732363 , 0.14912458 | |
weight | |
total # of elements, shape: 16384 , [128, 128, 1, 1] | |
# of unique elements: 16379 | |
min, max, mean: -0.69071037 , 0.51138675 , -0.013018153 | |
conv4_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: 64 , [64] | |
# of unique elements: 64 | |
min, max, mean: -0.62044334 , 0.785705 , -0.09482862 | |
weight | |
total # of elements, shape: 73728 , [64, 128, 3, 3] | |
# of unique elements: 73663 | |
min, max, mean: -0.12503509 , 0.13774472 , 0.00063377403 | |
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: -0.22714604 , 0.26259202 , 0.039029084 | |
weight | |
total # of elements, shape: 73728 , [128, 64, 3, 3] | |
# of unique elements: 73702 | |
min, max, mean: -0.28169575 , 0.27129906 , -0.00039011563 | |
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: -0.48107234 , 0.52189195 , 0.090940334 | |
weight | |
total # of elements, shape: 16384 , [128, 128, 1, 1] | |
# of unique elements: 16383 | |
min, max, mean: -0.43410164 , 0.54937434 , 7.039722e-06 | |
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: -0.21428631 , 0.19852366 , -0.004559571 | |
weight | |
total # of elements, shape: 51200 , [100, 512] | |
# of unique elements: 51184 | |
min, max, mean: -2.0640674 , 1.576186 , -0.16064821 | |