checkpoints/maxim000_nas_8b/statistics_hardware_checkpoint

Top:
   epoch
   state_dict
   arch
   optimizer_state_dict
   optimizer_type
   compression_sched
   extras
-------------------------------------
arch: ai85nascifarnet
-------------------------------------
extras: {'current_top1': 64.22, 'best_top1': 64.22, 'best_epoch': 254, 'clipping_method': 'MAX_BIT_SHIFT'}
-------------------------------------
state_dict:
   conv1_1
     output_shift:         [-0.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 64 , [64]
        # of unique elements:       60
        min, max, mean: -1625.0 ,  1508.0 ,  104.171875
     weight
        total # of elements, shape: 1728 , [64, 3, 3, 3]
        # of unique elements:       145
        min, max, mean: -108.0 ,  127.0 ,  -0.11921296
   conv1_2
     output_shift:         [-1.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 32 , [32]
        # of unique elements:       32
        min, max, mean: -9670.0 ,  7746.0 ,  3041.8125
     weight
        total # of elements, shape: 2048 , [32, 64, 1, 1]
        # of unique elements:       200
        min, max, mean: -128.0 ,  115.0 ,  -4.880371
   conv1_3
     output_shift:         [-1.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 64 , [64]
        # of unique elements:       62
        min, max, mean: -10111.0 ,  16383.0 ,  2751.25
     weight
        total # of elements, shape: 18432 , [64, 32, 3, 3]
        # of unique elements:       211
        min, max, mean: -128.0 ,  127.0 ,  -1.8574761
   conv2_1
     output_shift:         [-3.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 32 , [32]
        # of unique elements:       15
        min, max, mean: -16384.0 ,  16383.0 ,  4184.625
     weight
        total # of elements, shape: 18432 , [32, 64, 3, 3]
        # of unique elements:       241
        min, max, mean: -128.0 ,  127.0 ,  -0.30381945
   conv2_2
     output_shift:         [-0.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 64 , [64]
        # of unique elements:       64
        min, max, mean: -6662.0 ,  9154.0 ,  423.60938
     weight
        total # of elements, shape: 2048 , [64, 32, 1, 1]
        # of unique elements:       214
        min, max, mean: -128.0 ,  127.0 ,  -3.1674805
   conv3_1
     output_shift:         [-3.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 128 , [128]
        # of unique elements:       56
        min, max, mean: -16384.0 ,  16383.0 ,  4185.4375
     weight
        total # of elements, shape: 73728 , [128, 64, 3, 3]
        # of unique elements:       256
        min, max, mean: -128.0 ,  127.0 ,  -0.86854386
   conv3_2
     output_shift:         [-1.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 128 , [128]
        # of unique elements:       120
        min, max, mean: -16384.0 ,  16383.0 ,  4592.4453
     weight
        total # of elements, shape: 16384 , [128, 128, 1, 1]
        # of unique elements:       247
        min, max, mean: -128.0 ,  127.0 ,  -4.216797
   conv4_1
     output_shift:         [-3.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 64 , [64]
        # of unique elements:       23
        min, max, mean: -16384.0 ,  16383.0 ,  -5280.2656
     weight
        total # of elements, shape: 73728 , [64, 128, 3, 3]
        # of unique elements:       256
        min, max, mean: -128.0 ,  127.0 ,  0.3366021
   conv4_2
     output_shift:         [-2.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 128 , [128]
        # of unique elements:       127
        min, max, mean: -12892.0 ,  13250.0 ,  669.0703
     weight
        total # of elements, shape: 73728 , [128, 64, 3, 3]
        # of unique elements:       255
        min, max, mean: -128.0 ,  127.0 ,  -0.24837239
   conv5_1
     output_shift:         [-1.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 128 , [128]
        # of unique elements:       128
        min, max, mean: -12592.0 ,  16346.0 ,  3888.5
     weight
        total # of elements, shape: 16384 , [128, 128, 1, 1]
        # of unique elements:       245
        min, max, mean: -128.0 ,  127.0 ,  -0.55474854
   fc
     output_shift:         [2.]
     adjust_output_shift:  [0.]
     quantize_activation:  [1.]
     shift_quantile:       [0.995]
     weight bits:          [8.]
     bias_bits:            [8.]
     bias
        total # of elements, shape: 100 , [100]
        # of unique elements:       98
        min, max, mean: -1793.0 ,  1060.0 ,  -39.89
     weight
        total # of elements, shape: 51200 , [100, 512]
        # of unique elements:       195
        min, max, mean: -128.0 ,  76.0 ,  -7.866582