DIGITS-SOFTMAX-8X8

digits-softmax-8x8

Training run	digits-2026-11-01
Model	digits-softmax-8x8
Started	2026-11-01 05:19 UTC
Finished	2026-11-01 05:32 UTC
Training step	60
loss	0.373516
train accuracy	0.942143
test accuracy	0.942065
Weights saved	2026-11-01 05:32 UTC

A handwritten-digit recogniser: one softmax layer on 8x8 greyscale digit images (scikit-learn's 1,797-image digits set). Every matrix multiply of its training ran on Training@Home volunteers' computers.

Architecture

input: 64 numbers (8x8 pixels, each 0..16, divided by 16)
layer: logits = x @ W   (W is 64 x 10, no bias)
output: 10 scores; the largest is the predicted digit

Weights and biases

Open weights, free to download and use:

File	Size	SHA-256
weights.npz	5,376 bytes	05bdc2877cdda5b4...
weights.json	13,409 bytes	f6ab51dc70220341...

Tensor	Shape	Type	Notes
W	[64, 10]	float64	weights, input pixels/16 -> 10 scores; no bias show values [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.024083, -0.048708, 0.070619, 0.047625, -0.020813, 0.079192, -0.031877, -0.004785, -0.048648, -0.018489, -0.068349, -0.332328, 0.433669, 0.289702, -0.389422, 0.664663, -0.478545, 0.038226, -0.161667, 0.003766, 0.165882, -0.286384, 0.210364, 0.307113, -0.494491, 0.104089, -0.11433, 0.26309, -0.254212, 0.098161, -0.075258, -0.474958, -0.110502, 0.503742, -0.090931, 0.282676, -0.255359, 0.338946, -0.022185, -0.096907, -0.357338, 0.092682, -0.309454, 0.329189, -0.623966, 0.952848, -0.410647, 0.574163, -0.048909, -0.198909, -0.167455, -0.056251, -0.129262, -0.058695, -0.244373, 0.446452, -0.116375, 0.511018, -0.239098, 0.053987, -0.016873, -0.025187, -0.011795, -0.015275, -0.032231, -0.035261, -0.007958, 0.166301, -0.027553, 0.005872, -0.000661, -0.002278, -0.000257, -0.00039, -0.000683, -0.000723, -0.000931, -0.000798, 0.008197, -0.001426, -0.128458, -0.306646, 0.351205, 0.246128, -0.132122, 0.248685, -0.232296, -0.07453, -0.046646, 0.074593, 0.200706, -0.976124, 0.34235, 0.228893, -0.653868, 0.580932, -0.398477, 0.093675, 0.284842, 0.296427, 0.187868, 0.048397, 0.008267, -0.280664, 0.071821, 0.085908, 0.206838, 0.074279, -0.203392, -0.200063, 0.223707, 0.434234, 0.335624, 0.109608, -0.284436, -0.169996, -0.483435, 0.3066, -0.390921, -0.081655, 0.390237, 0.046773, -0.237006, 0.482778, -0.803109, -0.351187, -0.758053, 0.417893, 0.531417, 0.279759, -0.112118, -0.146153, -0.137048, 0.107649, -0.179091, -0.008142, -0.161678, 0.371624, 0.123215, 0.141673, -0.016094, -0.02539, -0.008025, -0.008167, 0.006641, -0.027253, -0.009221, 0.102374, -0.020026, 0.0052, -0.000519, 0.003932, 1.2e-05, 0.000357, -0.000665, -0.001972, -0.001132, -0.000378, 0.001182, -0.000766, 0.113593, -0.147674, 0.194202, -0.105022, -0.129801, 0.389839, -0.249311, -0.240487, 0.027192, 0.147342, 0.402886, -0.335478, -0.21326, -1.048029, 0.116032, 0.625915, 0.230076, -0.620874, 0.356892, 0.48524, -0.116092, 1.230766, -0.4367, -0.555747, 0.465644, -0.084296, 0.218324, -0.582335, -0.01989, -0.120098, -0.512206, 1.007409, 0.386517, 0.658439, -0.086356, -0.772156, -0.708685, 0.021076, -0.141398, 0.146934, 0.523083, -0.090122, -0.056978, -0.000439, -0.164882, -1.234252, -0.906179, 0.273333, 0.623273, 1.032717, 0.18727, -0.166289, -0.052896, -0.138665, 0.203671, -0.389777, -0.223684, 0.178262, 0.109008, 0.293036, -0.006679, -0.011852, -0.002129, -0.004358, 0.034002, -0.011602, -0.006539, 0.014962, -0.007916, 0.002157, -0.00021, 0.001756, -0.000263, -6.8e-05, 0.001369, -0.001067, -0.000548, -0.000254, -0.000506, -0.00016, 0.294858, 0.023819, -0.170215, -0.28956, 0.309554, 0.514403, -0.060203, -0.276311, -0.239735, -0.106733, 0.332192, 0.086717, -0.938225, -1.08243, 0.652831, 0.80509, 0.4048, -0.575566, -0.135581, 0.449628, -0.798157, 0.739238, -0.8731, -0.066232, -0.254157, 0.440133, -0.051961, -0.422135, 0.687756, 0.598057, -1.179081, 0.390915, 0.059318, 0.495492, -0.090942, -0.054572, -0.585871, 0.141824, 0.307436, 0.514854, 0.074564, -0.120759, -0.152003, -0.642614, 0.33608, -0.453012, -0.586423, 0.636566, -0.092639, 0.999793, 0.476474, -0.275767, -0.15733, -0.383611, 0.539423, -0.313939, -0.323834, 0.454212, -0.333772, 0.318037, -0.000725, -0.000516, -0.000101, -0.000135, 0.003724, -0.00045, -0.000538, -0.000287, -0.000314, -0.000608, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.388598, -0.145789, -0.194872, -0.208938, 0.784077, 0.004625, 0.108149, -0.088578, -0.334409, -0.312965, 0.465119, -0.180646, -0.734871, -0.693496, 0.719742, 0.112866, 0.72503, 0.278152, -0.184475, -0.507865, -0.887826, 0.146764, -0.201621, -0.334261, -0.017717, -0.187556, 0.427845, 0.503172, 0.755807, -0.20515, -1.149041, 0.242982, -0.081543, 0.487302, 0.366575, -0.219773, 0.178929, 0.503118, 0.348969, -0.678156, -0.146535, -0.40719, -0.679554, 0.352175, 0.634284, 0.014921, 0.209542, 0.574925, -0.829288, 0.276166, 0.461438, -0.359208, -0.332162, -0.120011, 0.359664, -0.044154, -0.037011, 0.336436, -0.529931, 0.264797, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.001608, -0.001127, -0.000254, -0.000225, 0.006474, -0.000355, -0.001646, -0.000374, -0.000523, -0.000314, 0.189579, -0.250256, -0.045387, -0.079987, 0.614797, -0.179934, -0.022188, -0.025027, -0.037464, -0.164175, 0.764648, -0.424931, -0.026087, -0.611835, 0.39249, -0.791572, 0.848975, -0.044781, 0.755299, -0.862583, -0.542246, 0.234829, 0.703682, -0.803766, 0.651299, -0.551846, 0.411631, 0.54469, 0.115606, -0.764259, -0.616991, 0.663233, -0.06067, -0.265668, 0.826106, -0.118067, -0.287331, 0.268963, 0.172153, -0.582163, 0.191258, -0.353689, -0.772695, 0.598923, 0.220171, 0.115697, 0.250978, -0.385252, 0.168948, -0.034868, 0.215912, -0.447136, -0.339933, 0.472715, -0.190649, -0.013833, 0.659156, -0.216148, -0.371792, 0.231526, -0.002531, -0.006077, 0.00142, -0.00434, -0.004374, -0.002317, 0.022529, -0.001939, -0.004208, 0.001889, -0.004913, -0.003284, 0.006048, -0.000725, 0.011986, -0.000813, -0.003204, -0.001225, -0.002951, -0.000875, -0.026626, -0.122979, 0.131837, -0.009572, 0.11253, 0.007052, -0.084602, -0.069141, 0.043503, 0.018006, 0.586369, -0.415349, 0.402336, -0.124171, -0.470538, -0.206974, 0.299344, -0.435203, 0.580227, -0.216483, 0.097448, 0.192612, 0.695364, -0.380694, -0.130418, -0.256135, 0.401538, 0.335144, -0.479601, -0.475815, 0.166771, 0.529093, 0.513896, -0.034631, 0.427672, 0.197343, -0.502433, -0.56159, -0.30975, -0.426959, 0.393572, -0.041291, 0.313324, 0.45984, -0.587953, -0.028555, 0.364993, -0.984421, 0.220282, -0.110346, -0.20514, -0.166984, 0.414006, 0.324779, -0.4359, -0.262424, 0.756279, -0.334126, -0.242658, 0.151974, -0.021579, 0.116615, 0.060585, -0.036167, -0.024397, -0.028712, -0.001764, -0.016291, -0.042396, -0.005855, -0.000131, -0.000612, 0.002181, -0.000102, -0.000144, -7.7e-05, -7.6e-05, -0.000219, -0.000669, -0.000101, -0.025869, -0.045035, 0.087778, 0.035395, -0.015744, 0.077278, -0.029194, -0.01818, -0.040032, -0.026361, -0.119508, -0.465884, 0.473456, 0.401236, -0.390589, 0.725708, -0.45312, 0.151816, -0.283917, -0.039504, 0.194616, -0.358038, 0.159077, 0.346309, -0.427741, 0.405082, -0.224961, -0.098102, -0.001225, 0.004238, 0.073589, 0.163016, 0.209198, 0.323213, 0.005089, -0.170442, 0.251523, -1.158597, 0.226356, 0.076305, -0.23057, 0.421467, 0.667785, 0.164012, -0.582476, -0.465492, 0.68364, -0.734353, -0.030475, 0.106082, -0.206416, 0.279705, 0.81356, -0.190362, -0.228077, -0.265898, 0.184597, -0.209524, -0.180057, 0.002398, -0.036434, 0.314058, 0.172567, -0.098423, -0.034534, -0.06852, -0.079826, -0.032487, -0.080703, -0.055667]

Tensor

Shape

Type

Notes

[64, 10]

float64

weights, input pixels/16 -> 10 scores; no bias

show values

[0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.024083, -0.048708, 0.070619, 0.047625, -0.020813, 0.079192, -0.031877, -0.004785, -0.048648, -0.018489, -0.068349, -0.332328, 0.433669, 0.289702, -0.389422, 0.664663, -0.478545, 0.038226, -0.161667, 0.003766, 0.165882, -0.286384, 0.210364, 0.307113, -0.494491, 0.104089, -0.11433, 0.26309, -0.254212, 0.098161, -0.075258, -0.474958, -0.110502, 0.503742, -0.090931, 0.282676, -0.255359, 0.338946, -0.022185, -0.096907, -0.357338, 0.092682, -0.309454, 0.329189, -0.623966, 0.952848, -0.410647, 0.574163, -0.048909, -0.198909, -0.167455, -0.056251, -0.129262, -0.058695, -0.244373, 0.446452, -0.116375, 0.511018, -0.239098, 0.053987, -0.016873, -0.025187, -0.011795, -0.015275, -0.032231, -0.035261, -0.007958, 0.166301, -0.027553, 0.005872, -0.000661, -0.002278, -0.000257, -0.00039, -0.000683, -0.000723, -0.000931, -0.000798, 0.008197, -0.001426, -0.128458, -0.306646, 0.351205, 0.246128, -0.132122, 0.248685, -0.232296, -0.07453, -0.046646, 0.074593, 0.200706, -0.976124, 0.34235, 0.228893, -0.653868, 0.580932, -0.398477, 0.093675, 0.284842, 0.296427, 0.187868, 0.048397, 0.008267, -0.280664, 0.071821, 0.085908, 0.206838, 0.074279, -0.203392, -0.200063, 0.223707, 0.434234, 0.335624, 0.109608, -0.284436, -0.169996, -0.483435, 0.3066, -0.390921, -0.081655, 0.390237, 0.046773, -0.237006, 0.482778, -0.803109, -0.351187, -0.758053, 0.417893, 0.531417, 0.279759, -0.112118, -0.146153, -0.137048, 0.107649, -0.179091, -0.008142, -0.161678, 0.371624, 0.123215, 0.141673, -0.016094, -0.02539, -0.008025, -0.008167, 0.006641, -0.027253, -0.009221, 0.102374, -0.020026, 0.0052, -0.000519, 0.003932, 1.2e-05, 0.000357, -0.000665, -0.001972, -0.001132, -0.000378, 0.001182, -0.000766, 0.113593, -0.147674, 0.194202, -0.105022, -0.129801, 0.389839, -0.249311, -0.240487, 0.027192, 0.147342, 0.402886, -0.335478, -0.21326, -1.048029, 0.116032, 0.625915, 0.230076, -0.620874, 0.356892, 0.48524, -0.116092, 1.230766, -0.4367, -0.555747, 0.465644, -0.084296, 0.218324, -0.582335, -0.01989, -0.120098, -0.512206, 1.007409, 0.386517, 0.658439, -0.086356, -0.772156, -0.708685, 0.021076, -0.141398, 0.146934, 0.523083, -0.090122, -0.056978, -0.000439, -0.164882, -1.234252, -0.906179, 0.273333, 0.623273, 1.032717, 0.18727, -0.166289, -0.052896, -0.138665, 0.203671, -0.389777, -0.223684, 0.178262, 0.109008, 0.293036, -0.006679, -0.011852, -0.002129, -0.004358, 0.034002, -0.011602, -0.006539, 0.014962, -0.007916, 0.002157, -0.00021, 0.001756, -0.000263, -6.8e-05, 0.001369, -0.001067, -0.000548, -0.000254, -0.000506, -0.00016, 0.294858, 0.023819, -0.170215, -0.28956, 0.309554, 0.514403, -0.060203, -0.276311, -0.239735, -0.106733, 0.332192, 0.086717, -0.938225, -1.08243, 0.652831, 0.80509, 0.4048, -0.575566, -0.135581, 0.449628, -0.798157, 0.739238, -0.8731, -0.066232, -0.254157, 0.440133, -0.051961, -0.422135, 0.687756, 0.598057, -1.179081, 0.390915, 0.059318, 0.495492, -0.090942, -0.054572, -0.585871, 0.141824, 0.307436, 0.514854, 0.074564, -0.120759, -0.152003, -0.642614, 0.33608, -0.453012, -0.586423, 0.636566, -0.092639, 0.999793, 0.476474, -0.275767, -0.15733, -0.383611, 0.539423, -0.313939, -0.323834, 0.454212, -0.333772, 0.318037, -0.000725, -0.000516, -0.000101, -0.000135, 0.003724, -0.00045, -0.000538, -0.000287, -0.000314, -0.000608, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.388598, -0.145789, -0.194872, -0.208938, 0.784077, 0.004625, 0.108149, -0.088578, -0.334409, -0.312965, 0.465119, -0.180646, -0.734871, -0.693496, 0.719742, 0.112866, 0.72503, 0.278152, -0.184475, -0.507865, -0.887826, 0.146764, -0.201621, -0.334261, -0.017717, -0.187556, 0.427845, 0.503172, 0.755807, -0.20515, -1.149041, 0.242982, -0.081543, 0.487302, 0.366575, -0.219773, 0.178929, 0.503118, 0.348969, -0.678156, -0.146535, -0.40719, -0.679554, 0.352175, 0.634284, 0.014921, 0.209542, 0.574925, -0.829288, 0.276166, 0.461438, -0.359208, -0.332162, -0.120011, 0.359664, -0.044154, -0.037011, 0.336436, -0.529931, 0.264797, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, -0.001608, -0.001127, -0.000254, -0.000225, 0.006474, -0.000355, -0.001646, -0.000374, -0.000523, -0.000314, 0.189579, -0.250256, -0.045387, -0.079987, 0.614797, -0.179934, -0.022188, -0.025027, -0.037464, -0.164175, 0.764648, -0.424931, -0.026087, -0.611835, 0.39249, -0.791572, 0.848975, -0.044781, 0.755299, -0.862583, -0.542246, 0.234829, 0.703682, -0.803766, 0.651299, -0.551846, 0.411631, 0.54469, 0.115606, -0.764259, -0.616991, 0.663233, -0.06067, -0.265668, 0.826106, -0.118067, -0.287331, 0.268963, 0.172153, -0.582163, 0.191258, -0.353689, -0.772695, 0.598923, 0.220171, 0.115697, 0.250978, -0.385252, 0.168948, -0.034868, 0.215912, -0.447136, -0.339933, 0.472715, -0.190649, -0.013833, 0.659156, -0.216148, -0.371792, 0.231526, -0.002531, -0.006077, 0.00142, -0.00434, -0.004374, -0.002317, 0.022529, -0.001939, -0.004208, 0.001889, -0.004913, -0.003284, 0.006048, -0.000725, 0.011986, -0.000813, -0.003204, -0.001225, -0.002951, -0.000875, -0.026626, -0.122979, 0.131837, -0.009572, 0.11253, 0.007052, -0.084602, -0.069141, 0.043503, 0.018006, 0.586369, -0.415349, 0.402336, -0.124171, -0.470538, -0.206974, 0.299344, -0.435203, 0.580227, -0.216483, 0.097448, 0.192612, 0.695364, -0.380694, -0.130418, -0.256135, 0.401538, 0.335144, -0.479601, -0.475815, 0.166771, 0.529093, 0.513896, -0.034631, 0.427672, 0.197343, -0.502433, -0.56159, -0.30975, -0.426959, 0.393572, -0.041291, 0.313324, 0.45984, -0.587953, -0.028555, 0.364993, -0.984421, 0.220282, -0.110346, -0.20514, -0.166984, 0.414006, 0.324779, -0.4359, -0.262424, 0.756279, -0.334126, -0.242658, 0.151974, -0.021579, 0.116615, 0.060585, -0.036167, -0.024397, -0.028712, -0.001764, -0.016291, -0.042396, -0.005855, -0.000131, -0.000612, 0.002181, -0.000102, -0.000144, -7.7e-05, -7.6e-05, -0.000219, -0.000669, -0.000101, -0.025869, -0.045035, 0.087778, 0.035395, -0.015744, 0.077278, -0.029194, -0.01818, -0.040032, -0.026361, -0.119508, -0.465884, 0.473456, 0.401236, -0.390589, 0.725708, -0.45312, 0.151816, -0.283917, -0.039504, 0.194616, -0.358038, 0.159077, 0.346309, -0.427741, 0.405082, -0.224961, -0.098102, -0.001225, 0.004238, 0.073589, 0.163016, 0.209198, 0.323213, 0.005089, -0.170442, 0.251523, -1.158597, 0.226356, 0.076305, -0.23057, 0.421467, 0.667785, 0.164012, -0.582476, -0.465492, 0.68364, -0.734353, -0.030475, 0.106082, -0.206416, 0.279705, 0.81356, -0.190362, -0.228077, -0.265898, 0.184597, -0.209524, -0.180057, 0.002398, -0.036434, 0.314058, 0.172567, -0.098423, -0.034534, -0.06852, -0.079826, -0.032487, -0.080703, -0.055667]

How to use it

Download weights.npz (or weights.json if you do not use numpy). W has shape (64, 10); there is no bias.

    import numpy as np
    from sklearn.datasets import load_digits

    W = np.load("weights.npz")["W"]              # or: np.array(json.load(open("weights.json"))["W"])
    image = load_digits().data[0]                # 64 pixel values, 0..16
    x = image.reshape(1, 64) / 16.0              # scale to 0..1 exactly like training
    scores = x @ W                               # 10 scores
    print(int(scores.argmax()))                  # the predicted digit

Plain Python, no libraries: score[d] = sum(x[i] * W[i][d] for i in range(64)); the digit is the d with the biggest score.

All models

← KE9GDP