From 18426c75529a562716c1db4e45874fb39a8ab379 Mon Sep 17 00:00:00 2001
From: Akemi Izuko <akemi@noway.moe>
Date: Wed, 20 Nov 2024 12:11:10 -0700
Subject: [PATCH] Init

---
 README.md                                     |   17 +
 cifar10-fast-simple/.gitignore                |  129 ++
 cifar10-fast-simple/LICENSE                   |   21 +
 cifar10-fast-simple/README.md                 |   99 +
 .../doc/a100_epoch_vs_validation_error.png    |  Bin 0 -> 19529 bytes
 cifar10-fast-simple/doc/plot.py               |   28 +
 cifar10-fast-simple/logs/A100.txt             | 2006 +++++++++++++++++
 cifar10-fast-simple/logs/P100.txt             | 2006 +++++++++++++++++
 cifar10-fast-simple/model.py                  |  141 ++
 cifar10-fast-simple/train.py                  |  278 +++
 10 files changed, 4725 insertions(+)
 create mode 100644 README.md
 create mode 100644 cifar10-fast-simple/.gitignore
 create mode 100644 cifar10-fast-simple/LICENSE
 create mode 100644 cifar10-fast-simple/README.md
 create mode 100644 cifar10-fast-simple/doc/a100_epoch_vs_validation_error.png
 create mode 100644 cifar10-fast-simple/doc/plot.py
 create mode 100644 cifar10-fast-simple/logs/A100.txt
 create mode 100644 cifar10-fast-simple/logs/P100.txt
 create mode 100644 cifar10-fast-simple/model.py
 create mode 100644 cifar10-fast-simple/train.py

diff --git a/README.md b/README.md
new file mode 100644
index 0000000..91ea718
--- /dev/null
+++ b/README.md
@@ -0,0 +1,17 @@
+## Sources
+
+ - [cifar10-fast-simple](https://github.com/99991/cifar10-fast-simple)
+
+## Setup
+
+Get miniconda [here](https://docs.anaconda.com/miniconda/install/#quick-command-line-install)
+
+```bash
+conda create --name mia_distilled python=3.11.2
+conda activate mia_distilled
+conda install pytorch torchvision torchaudio pytorch-cuda=12.1 -c pytorch -c nvidia
+```
+
+We've found that CUDA 12.2 will still run without issue on `pytorch-cuda=12.1`.
+There is also a `pytorch-cuda=12.4`. Check your system CUDA version with
+`nvidia-smi`.
diff --git a/cifar10-fast-simple/.gitignore b/cifar10-fast-simple/.gitignore
new file mode 100644
index 0000000..b6e4761
--- /dev/null
+++ b/cifar10-fast-simple/.gitignore
@@ -0,0 +1,129 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
diff --git a/cifar10-fast-simple/LICENSE b/cifar10-fast-simple/LICENSE
new file mode 100644
index 0000000..f4e780e
--- /dev/null
+++ b/cifar10-fast-simple/LICENSE
@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2021 Thomas Germer
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
diff --git a/cifar10-fast-simple/README.md b/cifar10-fast-simple/README.md
new file mode 100644
index 0000000..aaab251
--- /dev/null
+++ b/cifar10-fast-simple/README.md
@@ -0,0 +1,99 @@
+# Description
+
+This project is a simplified version of David Page's amazing blog post [How to Train Your ResNet 8: Bag of Tricks](https://myrtle.ai/learn/how-to-train-your-resnet-8-bag-of-tricks/), where a modified ResNet is trained to reach 94% accuracy in 26 seconds on a V100 GPU.
+
+**Update:** Also check out https://github.com/tysam-code/hlb-CIFAR10 for even faster training!
+
+# Usage
+
+```bash
+git clone https://github.com/99991/cifar10-fast-simple.git
+cd cifar10-fast-simple
+python3 train.py
+```
+
+# Example output
+
+* Timing results using an A100 GPU only including training and excluding preprocessing and evaluation. The first run still includes some PyTorch/CuDNN initialization work and takes 15.49 sec.
+
+```
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1557
+    2      194                2.86                 0.7767
+    3      291                4.29                 0.8756
+    4      388                5.73                 0.8975
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9204
+    7      679               10.02                 0.9294
+    8      776               11.45                 0.9373
+    9      873               12.88                 0.9401
+   10      970               14.32                 0.9427
+
+84 of 100 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409949999999995 +- 0.0012262442660416419
+```
+
+### Epoch vs validation accuracy
+
+![epoch vs validation accuracy](https://raw.githubusercontent.com/99991/cifar10-fast-simple/main/doc/a100_epoch_vs_validation_error.png)
+
+* Timing results using a P100 GPU.
+
+```
+Preprocessing: 3.03 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97               10.07                 0.2460
+    2      194               18.60                 0.7690
+    3      291               27.13                 0.8754
+    4      388               35.65                 0.8985
+    5      485               44.18                 0.9107
+    6      582               52.70                 0.9195
+    7      679               61.23                 0.9272
+    8      776               69.75                 0.9337
+    9      873               78.28                 0.9397
+   10      970               86.81                 0.9428
+```
+
+Train time does not include preprocessing, evaluating validation accuracy or importing the pytorch library.
+
+The total time, i.e. what `time python3 train.py` would report, was 42.125 and 103.699 seconds respectively.
+
+* Timing results on a V100 GPU ([thanks to @ZipengFeng](https://github.com/99991/cifar10-fast-simple/issues/1#issuecomment-1057876448))
+
+```
+Preprocessing: 4.78 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                4.24                 0.2051
+    2      194                7.09                 0.7661
+    3      291                9.93                 0.8749
+    4      388               12.78                 0.8982
+    5      485               15.62                 0.9139
+    6      582               18.48                 0.9237
+    7      679               21.33                 0.9301
+    8      776               24.18                 0.9348
+    9      873               27.04                 0.9396
+   10      970               29.90                 0.9422
+```
+
+* Timing results on an RTX 3060 Laptop GPU (6 GB VRAM)
+
+```
+Files already downloaded and verified
+Preprocessing: 4.67 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97               10.50                 0.2578
+    2      194               19.47                 0.7549
+    3      291               28.21                 0.8737
+    4      388               36.97                 0.9013
+    5      485               45.72                 0.9127
+    6      582               54.62                 0.9213
+    7      679               63.39                 0.9286
+    8      776               72.17                 0.9348
+    9      873               80.95                 0.9395
+   10      970               89.74                 0.9412
+```
diff --git a/cifar10-fast-simple/doc/a100_epoch_vs_validation_error.png b/cifar10-fast-simple/doc/a100_epoch_vs_validation_error.png
new file mode 100644
index 0000000000000000000000000000000000000000..e0d93ba4e2f1493d8d45c9d31d87c4fd2a3c0101
GIT binary patch
literal 19529
zcmeIabyStz_BOmhQ3OO#1VlijJER*4DM2~~q)Qq$9TunvNJ)ouBa+gBhzLjtNP~cM
zN%uQ%o^#GP-x%-jkMA4b```P-fX&|f-fP`!tvRoGUDw>NRFq{e<B;K?P^ioDa?)xj
z6lMSlg)xbJ0sg|<Gdcr51YKoxT-6;cT-}YG%~49mu8xl#TpwGTP@&D8U927Kx!K`g
zRw^r3S4S5i4i3Bjynx-o*^=X1Ol2Zm<dUPDt_uov#TfY?BS$>j8ii7<ke9xz;gPs9
z=INuc{r>cHmo=_U=$vueg(n}ry_2|3m_A^Lm06{hI}$d)sN-NKO<b;RL->iE(Qf0l
zRxLADE|qqn&d(GzX~lj_>epAbREat+ypC&=q;KokKM6OZVvTrCYx!<)b0hX*TO2pT
znkPXLe~OD(iYXUfGXab#EwFzXexT-ch$K*`07fZ{^C;BIJOB5=|CxJW$}cP-fO{U3
znVDHlO)Y|zIC#hjg_7VV4SfIiuXAR0cDqf;j}lItNhU^kMOU(DXD9{=rDr@*@4MKO
z9yG*t|1O3Imb|=t#G?O`hwL$Ju!SUCJn9U$mxs9b1Svxp1F};$%8J`((@>Zw0*kc$
zzlHN68<$Zif)>UASZCEhOl){y{?7w?6Wr8^wZV=!?&#>~%&IDZ?UfPj-3k8&yk_iG
zQ_{fIx?RJ}%uFVBcI9;0uz|T0e|r-f*d{w9C0Og>vXnhfv2eG(Tzg=3xU=Fl6E4f+
zzWPw~WN-Sj@1Gw@`TLu5OS##}y@8i+M!kKT%4PaBr@DIh3v!-T+D*s>X^HQ$*DT-3
z%*e2sYI^b6Zn`y0`{BdHWd9S-8-ee)wp_=YS}CeGe+bjbQ-;YhF)``YdXyB^tX4;!
z;1v(DRF1HPwKm`3M4_Hv5^JFGIKQ6g@}no7*S5@dkX^nvJH;;#reb8;8X8TzGE^!{
zM@L5{>gBr5YhS5PBOCg&+HEP+`+iCnL-F9q4BTY`dusIU`I@zQeZykya~JWWIxQmM
z<~+M1mUHPEy<CKXf&#5Pj6%sUCL-dh&#INnpN^j8q0$k40ayg#;9rD|Bt|}}iZSf{
zAI;mhHd=2jeVR+|yLGxJE4n-CQYiN4%?<9E4UeZ`zK1*I#wKtdLkUZ00^G;kVpi)9
zS0~GT1O){T4wi~}Eq{>D&dxIN@@htC=;`H^m6c5{DJUv#r{6lBuWEG_N^%xTT1lor
zp;AR<ma!jcYDO0fWI1$Mlrj-u%SC+r_%SLzK9ARS;C#xiSKaUj<7dt;F2X!4FihFk
zFwAz2hKE<g_EPU~bMu;gBk=v(#t_f%kW8<f$e4R_p<6v$x&Pc1k<^ppLlzbmMQ+O<
z8eU%2UqfzM+U!)~C~O^A%8-dN%1kSFaq{9qv=AZGN8R#2a_USJRBzbtbSU#b^@FXv
zB!0Rn?zUF9dtrBf#lb)S`*(#g=XN$`7M4EGxuopMaWBP^k`j9Qm<kf_?E}je=V3b}
z>~#O3MpHC~b{j5X7mjP?8L{olol}b_vl-sELq9~J8ifMX&KsR0{o#CPG$NyGH)csC
zWMYN~g`(<JK}o5+ZqF@xrHqAV?R_a&=^QGcLIwlxE^)Ao;EipmUaS|Ck$rI6<)kq1
zREI8SKdJONi+6Uxn51cBxR34<W(Q!PBu3IufjN`XZ+o3J2k#I<JS?NYdGu1jrIlIm
z^ZV+6_Y`!q7Ok*h2%p2-Sc%)G`Wo+_hsSm=F-e<ZFX5q3!{@L#pYs}Cta^c*wqXKz
zbXmTbBcMj1eKy~e8vbT=lc<qkAl)TVsHHV$z`;se{>51&c%F(1{BR)r)RBx7K9>zX
z*H+~6eIzK`y}%^>O(u(~BvwTsgM=4WeZ{OwuX@PqhVjp|Brx!tL?fqXUYF;8z13Z$
zwu+63LMc;Vaei@<g<#)3z!+ehy20a!F^GYQ3gCm6pBws|`KXqj0Y)DlfJu6R40m$J
zq}b;8Y(Ck6@E0$-R+ctvZ6yPVnq3#<DvmaAl}4L#E4e_$tGJSIkJV){@U9IlbexBT
zY4;41l-}7(<=;bm_4O6i$m_bivq=kJ5jAp#_tV|IgrAUSd+y-3MCRF~Nz+k*m-745
zZZCu^VK|@7oDz%kjiVwQtZriXINEI4V>M1>cQ`V`!we^;*g_e@{d0k5@vjQjD_m^s
z>h-k`3o3<1-u1JgRB$xh;FVu2%B>v9+>zBmW@ACd@bvNiS6k<+aNLXF`M|hF6lzjK
z<}t?dSWP9YC0E$dk&nEZn_Kzc&CUKyMG`uCdYhxY4Q4hrUcv3QSu-Bx^Batf(#Y+^
zfql<!KbrIV_wPy5?Zy6xjQrwyy$Bw=VV=63VOy_-%$PFtir)3>*BA2gi|%kcmu{`R
zOZ`Qf&un`%yh(93DmF{V2;K_=h5C_exBT{ukpL{d-}V6FM4981B%6N4eG)^D_q5hM
zX)<IyR&NK2jfzXV#75qH>olp<<z?*5e6aH_H?=#xe(Y*zC~T9a`(Hv)l(|`1_qxPS
zYmp-z+>)l0z?WN5p#?D<f0f&DQo`KA!e*@6jX@-|Yc6td$?h+w6#m(sP$#&Kwcf97
zq^YYLXJ5N<PhL%4{uMcXkdWKrPqT)oD5B`7sKj*vh<xEAEBInwJIh0%xer(cuEio7
zUGFwC&ScXITwe6D4g`RdUMGu=*xO;P*L;>oDvLv|R@`@_mU|!b3R#^N>K(Gidq)1p
z<>P*bc6EQhvwXJ8&CSir%~jRa)s0kb?Cn*{%*sk$*HO+^SI#aKo#s<)SbgREtB`rO
ze_`pF=J&<;M`&7D$9vR%=N;kpERNUJu-9$%PF74c2eHu8-?JSkxMV-_$t8Pdbu52$
z^dYxNh1~sSX@XZv$nI>HamBzpkEvH?75?|{U$e?vl*_I)WAQxJ5m^sXirgoBwG;U4
zdX@OcGDb&7-zH=DWF+F@z{stVUW#EI?Tk3su8r5l3poAe>JmMePyeV_NG9%A>m!!(
z_05fRh7_OP=TU``Nk7hZiV$%#XRJ)O0wfklK9J1VWAdf^(Rbg=#)zKOKP*A}3N4xF
zvcAJV-iK_)!Y$q)B_o}J8ETfDN!w;z`d!~3+xOy?IEst#EFsRLShx#hvF~VK^}+5<
zg<Nj@82fDbv>+&yOg`<FISZsdaN6Fvd|{i4Z9j)X#Y2$$s;sD_@uoBFW1gPxrs%c1
zkPtRiP|dM@_I=h&+t=<2T!!F<yzoz`2=Z$;Ib0%!f;GS}G7{pZ<6R6qmk_KE0cR^1
z2cK*I$R=W0D;QsEAm^959hw}L;{yDdo$qVb3DYPNJ9%zlablWf4^Dim4nX!u{1Z%4
z3}-L$dVDxSlV*$ogp=Rm{)wLqQW$Sg&(sV)>SZ{v_%LrDbcDjVCCT9eUj!d7P9&Xu
ztS7wX_qT0WLX;mi7N_Kl8Rs>Kl{{DAcQcm<k6_gIi5juF1nT^#znwUz?0A2ofzK7-
z!g=)krTee`O8qq2+|-w<5)e#G;DuL3E+2&q&cIHhic&?<$T`w;(geWJE~>x-e8tZ&
zZ;Mpay|<qi*mYUeO0xAPv>eakAlzQ;i1{*>eA}nA@1DVM817-b;S7PSq6UL&yaeyR
zkAD^*BjNqA5Ztca)xm)nNe;K93D3HU!O?i=Hbp@JmmuIXY#F%R^y`pI%IA#*VLpx`
z@Hnc|?cYODL3<rrKe~rpc>}Jjf?xR@ZjLxr8tx-w<M)0W!A3cNJYEL}t1Byqiou4Q
z+zd3HANO9<)YFTH2*u;IW#xOaKSxh*e^ZJKXR`iTkcL*v+3wW%Lfp94aB4W0>|JDH
zVzTuo!NF#v!tS6i<s{=rlJLddIAQmqKV2!s*~wlCH~Ai`NJ>6MHu>VB&9%rBA@_%G
z?nF-0m5{@R5qLMYj9QrmXs8O{wcM&FZL*@EpkQ(k#^Qf6=f8D0acZb-=$T1t-4;Qo
zATQqoAv15OsHUiyyD~EouX2a7K<#e0?A-R6&Dm~}q(+~2oc$gxwAsP!w%xCt@4dS!
zbiCEK-ND^;K4qfRG7MlD%x!*>uX3a?oJJOjYptSL|B7qanrB&w;VR%&Ks`SJul{WC
z_b=2dR{;ExZ{Ki?MzC2#1|T|lHMmsdAi&Y{DZ;~vf-W=YnrC6jqTUs22?AK3Us0Xq
zpQx8?0Z$|(EEAmqDZS=zC;1&LRL{QC-^w>SR(K$~t&txlX!;e;QzUuP6hOOOY?a+*
zYaN_FOKW*Ef*Q8YWN<<hs4GxenVHt>lb@9;E|c-d!JMLLA=oKI0LV*NR@c)b19%B>
z55hznjfR`DV~*^cfPAmk**24X`3uuy_0ycq+>a!sik0DV#X7I;2o3Suw15|pm?yTj
z?Nx5vd|~#_AGFoLPLnm92}cB^Wh*6QZ7&Tjy$_=7T^*|_e>`+Y6z=8suC5qbT~=bH
zBq7Z!BI^NKMGtc!@yrL_v{(5T@U!3JbM~N?n^)F?eNMq)67%E+m&tRj0-g6axGm-z
zP7fQ90^)Q(<+SJdrE5<Qk9H~>G|kO3?LW;D!!rxDn?L&W9qKZF{nEJ>S(ya5C7KX9
zfaimaNdt#&H$@LFMUSk!Yn{H-wuEs-YheD(p|)Z~#i)Bpgo){W<@2b3eaYj^8XD<k
zrblvZm+;M9Z4$B1CZ-aGNjgOq^URnU4!M;y#v8}86_-s%-K$!r9KDJUp+N@zI<&A2
zTfl)ub1I+v?7RUM!5(C5%hk{>z>^v_>_Co!1jZYj&sgsprg%6Zu%hn0SjMx!1ccLk
zzv0@q@6jy!NAbbz)pAZwPAhNe@64KUA`y<<!VBZ^g0DLpaj@9wk<0c%U$!r%OzKz6
zy_P39n*BeS#*Gh*gi$C?Rk=h=b|xlODgvtk8Zoi)<tpRs8Wydw$V8m83o)L@d1Pp4
z7_2;Z@|&dL<?!K75)S=Sh(GtTNE&5=iJ|t$9XjOD`5<9XgSKmx4$mz?5j!FgFU|0q
zA)c1Lem*?Wwzg~A7=Qs4f`u%O?LRl*j_>((4~PY#+pA;xBo7^*oJ@<KNQrvy{Deg@
zM;Ck;s3Pi^J7m!?SmH-+4yU>tQKJM^_sn6gdbZ6%Z{~1=e*-?HNX|t{4+-Sp$S+V!
z<*{fPT%-7V5zp=n+^MqB;!J)Er`d+W=VC)K<Z%lNi}k~fj{BP0+6jbgCOWL|Cfx$`
z1tGCWXyCrGMWuNBzBJ~u*?G|Rz%LECa;ZgU@A_0T@1sUk|G<DvYbb@+(MD@$=#3o!
z`v-GB&aq-&MoC2BzQPo6o{_T~tx|QF`_YfQ1M1F|(NF3SoZdpo_L7h$Dm>g6-SL5+
zwr(SoDiQMniYi_gubHz0-zkHmeowODWV?9Q`P0_kbQp`M=-9Fx!_2nVEV^cP!=uT>
zuBqvxl+f;!yIG~42B-HM6Vt~QvY4pU<`*M#?%voiaysj!fx?54i4*#UiQ@iz?WNZ&
zMcrZkN2a3tGm#du!K4eTI&=&I8YG-$r5U5k0Am|zZ>yr_Iuo)9Y3|Km@!vx#G9-pE
za&SC=QFk9~FGVLL<ioLxrQlwM!wB`lY0McYBdC*1LqQzYcBtJ5MDSmj_=(QupUy~H
zx0PW{#doYd{-?YCOYycvT!Q9ikM3GQh`EZ@&55U5xUKq_FCny^bt5g9y%#Wv$-T0d
z*H{D2;6a+ETnA=7gkIM+x)`fpAB=~~ZDSy$y*J%?6%xXGTX)9#A+DpP{3jnbKfx8G
zAfOt4E{y>vS(w+NgLLIri4>HpWCD(<L*dj~VLzOlTU)-l<Y?6>Fh}(@+ab`b`!kGO
zK@Ad;v$JzYDB%Y19k7yK{=1dTjDL|D;iY7j(K!L%16NHgEk5s4B$!Noezp?J_?I{N
zf(1RC!k)G*%)MWxVYzvEc|(V%`(6H8fDaF`USkg3*jD_E<o<wpvje6_&77{o*4b?)
zB|NsL@7@sq-XpWbTw`R{E)-8CO;|XEIDn$~wuM;*If=M|o{lb)B0@T>W5(l?$;w}?
zL2GSm&vi;j_EbW#hN#WIe>*}6++esLxknFnNhSbcBb6f6cgIx@1ui)C0rL=4)P_Fm
zH+d>4LzN{&hQ+DCt9*XQa9J}+{1Fo8yTc*Qj{}hWsKcs6DLZbQQC+RKzrVlOn;Crd
z+BIIkBai5J@8V;tOt%$6wbT08jSM7G)`~fWVW{&Gx7AU$wzlS#Bbq8I-4?Cncg$Ww
za(@;El!o};Xqk22*060cR2YW6ux^?f8X5UT)p?zM=$ukLHnNBxUI8eSG)<o+2LJ0<
zuizwR9r^huQ1sbXvheb5$F2Yb#U)|?2|R`C)C&@YJfl#H?$o<_4Ssb$#15DF+6lhH
zIt_oAOU1zvtv={TKC5>rXwB}S=oXT_S9{j{c5C=*`bU5fDgvSgVxH_Z`upAEb>4LH
z0Grk9><aINQBsV$Pc}aJ-qC?@Ey^Au|KeH*e~_)6>$s1^EyZtdPzt@*9*Ddjt8s^u
zXaK29O;S=aa!wp*B#T?NGvC4y*5?8qY3lFaCY{Nm`r;=WVQ^1UA!7@Y<3nAN+hz+~
znueO1bCc<hIBwZe%P#vKZyfWGrBgfGSZtFyq#Bfd@s>GY`fG@7sc8s>FyG#G_j@T`
z2qI9nDMCO<1B9gK)&cv^>R@Z(^IDx@vDXgXgnG7bH9g~JV|DWoBO^ck+&;C44y4Eu
zl90RJs9S1LK2jp=wwTeUEUZ-HzBV93VvrFZPX-B(_XZX#8=L4V@3W9Z<?KXd%Np}#
zD*)J|2kZIIhW~~H0Th;q!ftc%LyC;eF7aZxlZGiif8`U%OViRMjeIs<^!E1h`0P2h
zeEDM48PEH9&V6+>x6XT)nV-K>$<MHgG(uJUS}an^yrjI%i&+Jvivdtc2Xo@5H^}3F
zxaq?pd6+ue9+R(MY2PU}kmNp|ix5A^9adEhx9E(gOX$p6DsC`@k|dH|sdsCkFFG-i
zHKskYEs}wSpP%2x0>y9hkig595Q_j6U`41#6A;PpK+DtfQ-%Fl-m{BW7U8TiGcvv%
zT<Tw`oET3t_1)~`uRb|gLL#_<OD8X~)gfZtVhs-yJIF!EzK~nuhA`E7HFfpv?<~b-
zts^+rLAa^fCjrK1XNHh9(D->rcKyi8z=%UbAw)}_p_8egTZ_MgIpE;O34T^?T-yG=
zf^MJQ_%`btN0F_0gdsIls)bbk4iChQj;mcLs8CcNWt!%8b8&JBA&sTR#>Sdfw0`-*
zNHFPFzVaF21cZ=!i9ks*4`t{v^&<X``D^&+=?D%kE-#etA!NMYznFy=0#|a^*n{IS
zq{MiJ44g^WsJDv;^I@CkJ51CsLQxA`8)S}c$Rh~=i_9(0N`o|cxlYX0q0(%zlReYA
z<DHS=64M}BYdB#__CtXcF=HQyIx47(F)^gH@>^!40cwc8xQ|g#Sf~l~C!fQ30Rl2e
z3|zy9!qe_G*uQ!mQj*uCon2g?uaRQ^3du#N6-F~<3}i~&-0FS}6U&UMt@fohDr87a
z_mUWb$A8YLFolRmH$2ZET<1iMjjW^`$(`(g%maBU*t1#y@VvGcGoL?yPT%DZ*_NIz
z*hlaz+S@$Wz#!+hPaGml)0)0NS@`1Yuf51f^YbHHOnSb&y4IgAd!=+*I_ZSzUzNG_
z+GpnN)p*ChBp9fG4@7qcQNt_f<EbnDDh38gL#k<NZ#wht=5AL$nMNR1IGY(xs)j$q
z+j85%cnTUD2>R@dXUuh*<R79Bv~X0lAb?g@JRVAeqW%{5r5VVyoJ_i|0zQfYNHreL
z3d`c)w|P^(x7Kqr)boc)$%lK7gxke_CpQKfe=|9Qh2e}D$`}_hpIj4T^j}#ki=q{L
ztmTQ{Z!_^(>MXFW(85K+^YZgMA?TB%^Dpq+KR+cTfkIJh%Uzd%6o0tZV07cg4Tbyn
zf5sU3G0DdZx?~ZvKeTeLo`2i<v>U=pErb_@QYGdvEcoOy_kGIoe5IwO<<gL43NM%L
z*0&!y4Gf%Ti}UcdCMt$iHx2k{J&2iaJ1ubSZ}B~jMKAv{Hu*U3G;<$P*?naw?uMlx
z$}}}e0%Tu{U4f#byu6&A_O;kKwMalFTfUlwIPI}O@k!9)<YEeVtUs#T9rtd1?7u%7
z<94u+o$!|NKwyP0>0K`;;d6v3kj{9kcX1sl2@MUCVW(fmKEDNL&<g4Zy^m(E4E>MY
zN~dK)pbD$F?p>bQot|Qmea1PImAoaxs0E5<8_79v)U}`jM6?0BiF$+Lx}A9T>UnD9
zt;>`_B+sG34QI$eGDWNSNr_gmp&j~74th(P0HGr#%r4~;N<0sIk}LmFm+heF<GmJ2
z;u(Nvz}`B&`9+#=;Qk=$3xdU|A?mcc%^<}Kfi7c!0K5IAiztbiG}P~^Gpd&R2}C}g
z+q}5H0Pt@7ZqX~X2-#ViC`MrGzWcamYlc~mrDMFJR<LX8XttU|x!S{$v!i2&tacQW
zdYQ+Di2;yw-+%nD*eAT!f=~>i^1_Pk<+C@#)Gnwak(CHpEDob0gxHUH==^F(mC9N!
zFuxu}?8W2d`1*{!ym=|t7Da&{1Q5;=jmKk^c&WDhB|=q5N5H0VA9V_%^j_w=5onAi
zns>x7LqJV99Si|BRdJ?bOddp%>Z8ByZvC3tyq;wP%~!-(fwBYOeW4PNGmE&mkwLYq
zs)<S3*RWf~kQDelNWx83;uVR5H5Rw6{RI&okxK3PPu_qeA-CQxznxM((Lcb}ZzE+H
znUJeh<Vs(bQr}I7+V?SrUh)v_E630YMb(SUzU!Ya+8t~*>ptx(0{*bv+y6xcu~)w@
z)!9Zp4|6AB)BBm#nxB)?4Q1UH;MMsj7b&srcRfV(svO_HW7E&V<*4P{E4_G`ECZ=5
zmrJ|s`^cwQ8JY{n4q3`z58+-w^CO7H>fCKNoT{ED=PquY9Ih=L?@kQD?|Qz8z_%i+
zfJifuEB{~zyT}#LUo-%<053YYm_qWZ3ZZb$b`>NxUeCW~Y`Pz5tFX&DE&49=kEXWX
z@_)Fou>o1`?`%6#fY?H+Zxzv2Q=<-Qey^M)#5<H65s?-V5pj*OZ*H?~ZWF6v*Nf(q
z{l?G&yZk7EA?2<@dh5%6FbHr8+2*+g1^3O(%@2UED1zv;7%IGKwd1t2GLp@to@E6k
z(d!Vqnx%n19dR_3BNRScAOv-WQsf1iRhgH_wItj4>Q4#pgE}Jd4L28aud?A-2?1Qc
zN=%{bm~sAufbRJ?%j6zN$)n5DkB7_T4|mrFZ&)VYt{G#m8ACqP=HzJq<H!oIB&teE
zA>Y4$kFPB%E7Jg$ltE<mzqO>D!oU#eYy(|`YqVTA04y@FbJZI)H8mCF1$RQP0kYDD
z2c==$Ha{8dhCXJtRV3I~%flW1>3=-vpKIBb3>Xm+&UgTl%J)L4QSQEG2<$=fdeuMV
z>yRL|47yt1h%!e5ILlM6GXn5=4Q%gEpFY7G-@Rkiy(;47J>PvBAll)@;GrJ4*HKAH
zV`lmgDym%<u8xXX_tGk*h#9oCwK;!!7%$@a5lWyq3?FjA8}@}qWpY?A!u>mG0Ps`6
z*)R<g%|i4-*)U2nh>f%heGkMnA3rY8D%9)G5?|_(y%pVW)4#VdJ-fNj1JwvpciA0q
zQ)8V)1y*93FRL|Lnwpt#zK}99BIE9}3)W4~F9C$+791+6H7p*ruN0vS0||A=X*i^^
zT1#r}Fo<>`GL9tI<!LjLlgkXteEd$|Lgb6e*xU0sUes>5BYYur@^y&n8vDk<Vhyt$
z)7Jb^Uv(VCB@`+Vvk!L?7y^X)Mi3iNR^nA_*R>z^+|6|?TgI!bKPXbbIxD=Ik`}NW
zN2{EefeP4N9pkkht-7{qS)wzudnn9K8vm`$acR5jwJtD)0oad#TBx0F?J=S$LnwlP
zZ-i>NcR6%FX0#O!koEukX$34AJ*{N=^40-n)!O83&paC?By}juK<YqNnX;>d=00eN
z)eBh(_vN17TiqiyryraM{`3z6+N28!u9?%UL??9Uq?3Q*+X_12Ew=3UA2}(^T}bF{
zN>~^KgualS!XDg04RJ{|{a%~BIC(n+U10bQ79RN9m`KWP4`Q60>MJk?073ofEn`4P
zsXss97+69^E-tRU#O5g#V9THFF2C?}CZm7)I1>fvYXGTh<9QV$e%81<7<jLge?COf
zl!3sq4>#8{-*)zO2$vusgpQE6)fYH9IKuvtGzl>XMD^sJyz?oz2pfG565ssp;Wi{1
zRm&u|cYrDZ!j8^hY*L6_Ugi~qg|RAxT?!?^0Kfy+mGZ&w(;EoHuU^c{H!Ih3oABL@
zqqq{`?Dc^PG{h-(*n7R1ib!QX4=4WZnEl}Ppi!}9imw*r)?&c24rmcMvIxTlB0NMZ
z#5RH;si>e36&Xn*5AXnw9mjXbJZ2$;f1Hz>jEpB4^W^V{LbUdoa@M2@i$W>tl~Yxv
z6Y;as!RUDU*6CYXgvU_M?(49!>Za?-b#*_MPO_7Eh)hoarngs`a(o|Hy>MEkWD#9N
zi#<I#c5A-OZ}t1jEBb7y^kuxFb5NNkzW5>(5JKs<H#IppXy{!C+HSf1n9%mxgb_fN
zPp{e>Ghi2<loUoGB?GoPj63WdFb<%Ll$uU5h74j5JK{hb{9xvV@$i$gdFgK5@7FBD
zT$keWH;C3rE;4H@<f}jJ+cJV9+Krjz@611bu`0a9B;1&^^>O4P$?L%VYC6H_-_ChI
zm-;G9itlY%gxkaZd9@HJf%!yYeFNwbDe;4zFoa&t`zN@1LAl`e^@inqVdeN8Vdl_o
z4>FH-{<Eo9aRFr|hbYnN4@5J8bKt15>xkug9TqmoRymrsGE$ikix2zMacO1cP*X)3
zDb7b@#V|e2=p9IeLKZbpLuecM_D^{_ngDx?(1KLU2&M(*5|MZyl}=WG$~5LSXdtvy
zP<sD<U^-I4*EL9L^19Jy1JlVysD`!9;lCdYaU(h|F4yn)AlInD4=P#>Al(rJh!!Eg
z-EuwOUj9sTcKURN2yl^Pe0;Fo2m2-q=pozxgrKeYp3iZO*c1@+>Pk%;pt>d-&o4cD
z6ZZ=ZjHtyJpHF+ggWWb-ND?Tp7{ZfQdxeE}To56&=?#Pp$Os7AOmy=m;;uNItgoE#
z^S+oV>0-u>goh7iT%m?;pPoVDT1XEQQ;T6Yh>DJOYdQudqaH8@{(ZFbe`!$d%K{Wd
zzg*-#xr~Q*+Vl(znG)SJ<rW5Ul;QeH%7hyB8x&MAb5%w(Tbb1+{k%Qz9awD2>VD<-
zaeo5&Ccu$Etu%np5H22=;R3R4fTQUF_9WDi33Y35++g$Q(Et)Cuffkr?Y2b|A<_&`
znGeMFGjtD2@a|h7Brd`OGzBeSt&h1+@Csup-$ma$ZYc$&T+?955+QYssZ5G79)5WA
zBoLp1Ii{Ts3RKv%b^Cu3KF{Un=SQcc6tf#tL!40s1Brt(t`N_Id)WeRp<${9jsFtC
zB?LHOJ6YZG-3mAK*^pTotI_7=<yA`do{iRy_f3Q|IS)0a7Mv_|(m>!h1PwfAqYq{x
zAMk)Z0>o1#gnBddTA=$hW<Q1=u|uNA*5`}2%C_}iTY<Y^2$pkqzj9b@zNecq>5L*!
zrUnR%Ku4D!g0dr(4uh|+vuFTATc}^T2&Fj4r}jN1><vfG2F`7CixUkE%EBu=qI;9i
zkRk)1BruVP0Nk0t{{Rty1FzgF0t$Fx;`F2vqR@o$8oTrCS<Es<u%y%#(G1yJz8VM^
z2e}pjoCrZT_|PQ;0)f8oLnvYpD&XTsHOSb*W!8~4wzj;8Y$EnwNQ>BkF7GVOtY3wF
zAIwqT4g3<byt4o~9M~wR_z>D@tM5VOo`-(LV+4`{x#4%bWMtd&`Z}42#|IElW*h~6
z0(PZ8%+AQj$;g;_-nzN(^Vh3RiBdoWv#K=XKW>G{OSzc8KDk&%#|%E8{r2sf%H6w9
zXc5!}ScOe&JOJZdndhd3-LDUK5yBZH+A_P53az;WZp+Y7d-wGoi0N9lSR%IO$r(U_
z`p2?BMGog%<n(B>+-uuLz88e-LO^Tt+6@QMMGYrjP}b+Qw7Lk8FAf%`Aa_*I<qyJq
zdM%f<B957VaW_K^|3+Ek|BbQ))M@ilu+<vK@v@=x!1le(g<F-HDls%P<n{X%gNA_t
zn|w7$XhheprTd>AP9SQqAN)-D+jZ$Y=HGoEGBKYjkk`sAI>`_gB%;h^P8r#WupkUR
zb$ts_eG4M9<l&6M&4pLc{;fz2)Fc5s{ClY&n=da3t6{^tL9CLGDA3AB{6bo{TtI~*
z6Gj)n5<^K>WZwR6A?5Ta8$c2rh+*=P^wyB+6qJ?Uwa@m>v_&EWGk6DTz*eJSjFcf$
zG_=YX#RfH6Qh|6q6z-$%%gT5FD=a`ADV%A#I?-UnKkE1l;oczYTLYxlE4PV8YCZX0
zh%R~U>^1k{$PMn#r4;}C`IESi`(@V0z9q1bL^$^!#39ulV<~B%aV$leDriB-$zwCD
z)d5FPGHmc{7#|Y{$J)i>PVV4;%Rk}mvv-6;bMczHyN6CyJl&GUtByr3%R3BJhoDfd
zL}Cq6X6HYsN=iPpuU?=pgpt>vSI(AM9GGE*XavM%j21aU1-D!!#~Ri9_7@`O^N#=l
zWRW!o)^cKDx8bx7(COCKn+~xQml(qPo?*R*^}KkS>(Y$X>A_&bR!qZ@jQ`=t1OgB`
zc?Yd}GvqB=o8<B2EtyDRQSN>tYD`O`M&=Ss!Z83!n%fV-IV&VK|B1H+tJ@>A2#*v{
zCMzw5{rl%Wj!2>dZlD5kh_7NfF8?a=I$WvjFSpemb8Nm0{(&^OzmL$YWh!gzmlRxw
zL7(bdJ$IcfmMn>606@|rfTXWqzRb_Y7!8=m7|<tlTA<0xBKmE5CS7C1MqO0&>p>5O
z@vGi;3bhB=Sq*hMlRsg=*`$0ybs!XciuK`TCWKjRO(Kc1+~q*bs={w9{#0fIq+lqJ
zO+}?CnY$^e)HY{HmHYWRr8Xgg#iRgZfR#F5K3x?h)frb75070y-PR=_Lj`n8V@N01
z)-s$)<<IGHsD>X-i$1AD+)@E=k>{31*i@ak_%!?X%I8KDN~b}^N>@Sk{CRxY{d_oL
zGmrqDC(Xc&pGblpBrXCkxH`!YqE-dGmBEm%5pbrK`=`nvf{V)l3FbUpn77`)kiT|D
zc`&|*cmJ}}x~woQBmpJ?APZLLv!{d_AM7IbC~P{IX#%&m`#W0Eh7;SuF}Fs<BbJK1
z?9a#d3)MUfO1vil_PT$=2mJSwhQeWdFalY=e3>xALmC5t4>dk+7)pAzstLQ%aTK_)
zda^Jo@LhO0H{B4*b|BFl*mG-cx;(5k!z)Y-zRD2HmuQlkrhVM}iMdmCLhWGLFk&A~
zg(|os6bmKs6J{$#E3k6V9zqk9G}Y@K%^SbI?6q=S^nF<er~*11Dxo*R?DrJuLw{N`
zy#n(CkMWe(b#MHCK3obzI!Hm;aeiN(roTq9c85YF<5HJ>_Y>!op%Nhcn297@t5%D2
z*pB;$zRwe7b6n9_jKSs4DaiYJcJ<3RRQJPsn3TIFMAXJ25{FgQifUzUY8o2YqFI53
z$HXGYYFms?&EyZ$nA;=utuDSFN$j5?#p_pUih&~7U<}~ww`P(*Sjq@ja30>C(!S`{
zZX_O29C{T%s&^Wy$rTbHTe5&8tT>VftL=su^0qIYP5jI2Uzh^mXAB|bk_6u>V$uO_
zB0EXgoo^`ne=CSY2~=NRi!QL-{1Me`-@j;=S-qM0VD<{BMJpzXiW;n14{Okc;La-d
zJN7CxtkVNltQs&ASe9;#T0&~`0+heitH@yu(m42Rdo{HU%uN>O(e@2RUW+7_-|Wlh
z*AjwVyI!;1!Be0D%UY+2bQ{Amun|A&yxozYf)LbD9D-<(g~U`~)IZT$1BN>@U=`|T
zLl8a8Xx;LvbFLZt>mdT=?8-55+Si*Y*KPme-eP$q$DiD%U&?+9cpMYB!4l6WH-(?d
zqQd@cc+kp2{Tq?-KNzyI%%tlSgR-nwyMp8+h))k$ju&p#Ao90_SxEv#D$RLJ3`C?y
zj2a(HOX=jn(hA8LBoOsTU-f@9kkRQqAF+L(pf$azZLbkFm$JkCbxI)(_3SwRi9Ck!
zu|k0$rG1huZe-IE-e$gs+T7ek<P<;$;Me+ZIwR!jZx6q?5q<8j_))aS#{J4{2am$q
zUcq+5BQ!!e)U5E;0oCudJLV27aM<L|&z~Alk>-HiqwNGt!wO1DqN^qUZ2%KZm+JLD
z$tccxv&bPIe>}{0Jn^-To&i?%2v(H`PYU6s5KjTHcd{VzDJ9?fPs)s+tbi?J14ZQb
z&Q8%)t|aB`hwR#s7l$W)_8f2L-x@qU9y|U{eC+G;9vLFG1EKfjM}B_w2u0O!db|e6
z1jlf24qbDdgoNRLNRA@~HyY;!0M0ekn7z%jInt{(F3xJ$%l&QgxZ2~}hO7;EZ>|LY
zGJm98_V`<xOt9;82tP<?*auK4b^{MamC}_B74Li-LyE{0(_}hVNUbRrX5gyNf_mLw
z`S;{}`N^Ha>T@Xrop2}vIGg9(&yzHUHgc%6=5Jo)*?;k1%i``U6Vn2{J5CutzANy-
z5>dTb->3LIh@!qTWTz#H9$PpF!x+;rxY_-s<>vN6{hhP(sl^z8EsS5I`QX^3>c?ti
z0gh3U{a$+zcpBp0fct?2i@=x(3|W?&_vW`cl>d_F^>dKecSA{cc0hGfQPhv=gjV!?
z`gF+>|MXT~hse}%VMSQS90?ARl<@XW3X=2&Wg*4mYKFWo`h}-!L)*+78VN{Te0vW=
z`qxs#{LkJ5S!QlQ!$W+^-shHsOCvZFZAK6tqS8>*)EQMK<y(Kk72I1FgU>}D%r;$_
z&CNTDo3~$KzI@pO??Xb9!L_7e&`-zHt{^TAWQX1kf~R}4_0y6Y_F3IthxPpNHIFaz
zNXPHr>qNq;+<O%DZ@l|%)@@%(qvp$Oo0Fd%Y;Pva<fos<M0LM}H}jOqGO$%!KZixs
zt62)L_n*23Ir%J>(5j#aOYUtc4?xUj3hK<$)K_$yQ%w1OYjR)^g@Px0ek@IBK2R2r
zLg&jA+q3ts8I4glr^1!fZekG-I&4i6iF&PsnU%<K_*6aEZKzLzc=*zWNJ9EWdr;<5
z_=Xgny#M~AFiL~Ec7#!ig~{YNb;gPE9=^UEk+HwFJn+=t&)At7h3dGAA+6CSoLzac
zlzom)KCHG}sterLF6S^%k1%nl(tZ@4_Gm|PEHlW*mpkVh9$BRzyBulSkj`8nNy>UP
zXCWW4dNN|^wbDl&c?|>A?Zg=HQUUfdH^<G4rLON9jc>D#CoBN^*&wSbbRLU9wcVs-
z{)y8N*>1XPKNj9B$Fa_u|4AW|P^{eH!n_InY2t>W_e}#ca==bp{LUvM#E60iS(cU8
zFcVAgpPnA{ZAbQ1BjOuB+i*@-J=58;tw5K79Qe5*-Zt~zVbQ%_FEq9ubEQ1}%UQx1
z#=;`7BAEMNJrzQ;m7biWp{1E`By#w^UPQ$h8Qm~B7J-0dT3^wey^!Lw-iHSjP1#7g
ztAD16EL^IRD(Xpk)wJo>puA<SL*@DF-g&%JLzkJiwa&IAA4v6~>9%R=iVkA>;D;{T
zeH2YEQgc_vJm_r4Cww>qrPL2S>kOfw^Aw@{^pL%DI+bEsSSK?-MK3!dkg$=*y#TcT
z@5=5h!p(>=t-8iUN%<1W3R(@A0<xbQrOvN|y+_HQUKdM-l47ml9Z&$XW>wKit9cwj
zH|#S%nVOgA2#K+xCNuVqvQWua{hp*Ldb$->ki^(3zL(X<6-BB=ID|w3Km6|+TqP&#
zRl#u~Q!oZ6MU*588^gy>9rfT+gG5eVZr{#t=j%jX8r>>5;uA)<Po4DM15_BgQADGB
zyQZH-&`^sdlK>@jHD!XRQN4+SAT+yrpg4){GR+H_?M%eNVRPJilKpO`%yykH{W;`n
zfO~I4)UZ)2fn5&JYcfTG-KnWg(p}UW7~Vu3Rzc)d4n=-^btBi|SK-HGiReAACb$I$
zXvMc<Uo(7BR=EEjoWS++lK}Kuu?o<_`oXoxlt9?T?>`6|8nmvoQPG>0Gxil+23-YT
z7>20#x@!%aJ>U#v2Z<luMd*GfrxGOQJ57JZO5w+oLOej+T_WipKi*05Uj3v8b^pQc
z+Qf3oXAXklc=1Q`?dPt8c_b%w-I1way`Q4AxY?q=v5fYzpox4u=<^7Jaa$1e{EAou
zUCj{2up(3*(14Ql>KX$S?L|U=_CcOQm~}6(&;dD;8OLpD1I*HJg<XP)sqWzAKOc12
z7E>ZJP5|(LX>0^kS|eZW+^5tIjvwV=n&XQ?zB?hZDZbN}>c;^~ltpDcx=kK2H89CT
z0#odkM*P=^4dn)s(iRpLs;5JRUFVYA6(RsrfK%a2cLJrpQ;7ZaX#lvAU|$E36)}iB
z#Aztb6SXjcbLTg9l{e!((ayrV9N36zV<mvQbmK2b{G2*N68_)7mI#iAw4QX?i<Pj4
zL&ycyz>;K(#yH<$Qlb+zL}wUh@Y_vZIX>e+O=l+8{*FRbW_Lc=YKL|8ijp#Cy=JGI
z=v(C-e?={ZYQ_r}O?ByHR+&7*;_9Tx{;|OoC5?K?sj7=moB#B@R6scown$2_aHMpR
z9E~q+S3HnNy$;mR9ibMcboIoqc=DI+IkvSf4l=SbZY?IP<!k&d)Y8=~1hYG0^sg6C
z@obtFS{^u%BxMbBQhIF?FRW#7HHp>W9LkuM8TXNG=Re*JKH&@9)J#POANg!l`ux|)
z^EUD!<(PEcj`qmYT+Y>cm90lNzBsu%p|z(0dv<U_a{<hi%6r`4LsomZogSFGs+0N2
zB#czomARiHdsSY!<wMkE)br;#tO2>X+vzk2Hc?c>wdY`C>ksqWO21Zb4|R0Bn6KBm
zH;DU);Zt8ZRTT6<u$ZS76kTE%&CX&^RdaBYQZ3MWHK+1{k=Q#=bFt9>Cl=oIcNA*I
zL+`Iqwwr`k(+MRc8rqAjBo3>Psy{5buY(bQPY{j82~9L0+JHc*^kE%h&c8T~rAtP0
zu0pe}Lc)x6kHmt;@{-#&7URhMn|v3KFnsk@`&?-^a+-7&3$+R-x)CFD%;DAoA$LsJ
zyTnO>->xoweLr$9n{$1>5q-LZ8Nhd~<dmckggPikpG~Ymzf%IggO;=NM<54y-4-=O
z*Bj3P&E)TRBX(<Ri{2IsQ$Vn$lT2jWoBBz>`)8WGvWYhKLJ_|_n5(P#QlCZh3hC&k
z+qE#%f8}EuD1zYT_BvQ~&Le1aW7{7Mzh<A8b@a~OBXzU!U>`R&y@QM50blAKC^HLB
zaoMr`COj7Tz}tot$PWI8`rz+ce5Y>@-u}2B!fPUR+XMOq7}#a!pz8zrb=;I(nT4Bu
zM}6^{-_EJneITGLW_<_Cy_&D?>2{n>XY!cX?X^3(CYkACCG^XvCrTwvo8rf`1&RyO
zU8cG!(;^hG^1|(TK&ZB%K|V;m^@TaAu=T#@*=V41v?jFszo~1U&h2WJe=cs%v(cOu
z-$`!ooGsH&ctekV?ULgn+2Qovtk1Vf=@vot-cjfzcvHifnE39=mvAi()_ES@415oz
zakM6d?p0e&^t|Jfqxup#zEyC$nQNBk;gbIC$6A!7t?Os|9yj@cg3{oPgC1$vk(gl5
zdwZCiqVt_$?9^tc(xrJoLGcJO*ubyp6>i}@bPF`jY0@54xq9>7wJW@AuVsnly52X>
z_Hbr=Cu?A~_4BaU>Uns1kZ(ZT<XOuHZ%6%O8o}|~EKBT|&`KkL62M1j4!h;J(W5~C
znzj85%ptAB-;l*qSyc}ar*d#hs_l0Zui~uK@b_+Dzl87RF8GYDL|<)sutceX+mx(v
zW8MD(%1TPhw%IJwBgkv%i!Cc<8z-LUmHq_@nR3{2u9Ne!EmOq}`32$Pjvn$!k+0l;
zg)#=UhSH-I-)L6SW>?b+>EzseNe$!Si~~KO#`uDB!>YV^l|x~JUvA>DVG6#6#;r=m
zRgPWjhQVyOk3b4M@M4HLz*BRsozCg`k`=ak`9;rlOG|kQst%~$sos3~Rq%P_P`=ej
zEn79Scr}*r%kgY~E74a6mPOg?Fq9rBpP`67dDiOotDrTidB!8?Y9g9W!0xsCujpUv
zmeaeJE*mijtN>>~B-dyLLwl*R{i362T->onJolu!Wi8((KH0Cx%8RVJM8GS5Nn`p-
z;?{t__!pb-tK{ewexj!Huu7o^#nTjgfQ^;f)BA{7-^$uQaHaI4$h>NG1H32Fg0Okn
zVSVLYZB&M@tmzvp4OU)P3jXa^{J>n3H?LqLKQBu;TC9QD*G;LuI4@USGsGx@h*P5>
zwX#^bx2@Rp@qo7#WAvpTfhL~Pxdj|?vN5tFC3L0l-_W-V#3%|6N>e{Y$c#?no2Kev
zKQ<P7_P^`~iG@ovs%RD(X<nZF(5r2Pv(#>ql;?ADAZ~jE{0P>rt!M7zn^1_;wG7{+
z<O*s)5BMi}d_ikB55D@5b;y0^m5L0AC9c>ocxlVyc|CpWB{kitd`&mI<0Mna+-^Ao
z69q#93|CD{$<-;Z-K<CRJ}%}6;cDebDB)6=6dSO*2=`Osmra<O?$VW<v?;&7@kKq&
z^K0S;`F|IMH1`%O`Nt!xYpeOH{2f!`g>D3DEb+U3_;~Lki>=-%CENC(OLZt?ppxQr
z_=^q4HrL@ous5|?SAHMbvup-Wp=9nGXZQeAD1m8dp376Iq-*4bV)v*X<$TO4l9y){
z&y%CnC9jjc-to0~3G?3Z!b;%r@aukR!4*D1(*mX7w7L>bz3Vr9>a9IRs#S!-eC(Px
z648*_<0*@s5}LnSZ>UJX{8K}83xn2Z`}E)hLcDV}>g)N6j9%lGB~kC-mMwTO`^z!*
z7mwg*Lz=;+6n;c5)?yb(({8$~N>?iH`KDwsTS|?WZ;rM!vfBh1u;kko6V=4#EGV!U
zbZ^z_0&oC|$wxLy(kQ9nJ?({=e9_v%lf!_#r6bvuTPkiVmnwMcD>ghnn+}Rwk_MIX
z)GK$A_AmGGs+zheA0MYPA^s*+=(J-Tuu*^fMoDc+mR`S^6R-au#8ko9G`v78PvyYo
zIM04>^NnGWFpq6nDyXUfXuzE_AN1kBXd+uQncUU<hu{~JMs_SzWAEW5|9?CHXw?*F
zOd3I3Jb2bk*ZnXgWh5kJy5}XP2hR8Qb~<@*m-Cznu({CL3=pN%G`7h^+T~bO2AoVX
zGBQw+N3@t&_nCU1buX(!3vp!otWBS(JMFlv)s*jZ;(va;+wr*ti&R_t5U|uoz4T2R
zfqSS2Px@^%!ze|Ew=Jt47JWFmQbo8%Q?@Q9iTlqFMen*WD~f$vqNH*X6pHFKa*R-E
zcvzg$f=SBYrc?yMb(t~y`7WWJ$F27-{`o<tHZGfE0yP}mC*$)#zf*ljm{sI`|BnZp
zUyhcHPJ4loMZB-z)vh~O%(sxWDM91}VCrFJYX1oP0$@t-Th?K#c=Q?bGKCON6}Gj#
zebKnjrc$YC(#N1%emYR0fyl`#tPl+H*FR2S5M+3SwepFGi5E#4_FjN_A`LpTXJ=;S
zfpotktQEYqI(8V_a0OBccxOt}%U}NQTDbjn#u+N$roq)Uiw*0DNJz2~CF&->J&*9f
z6``TIWMk55(6qsVoCTK;p%m3i@jv+_XiNi5RbMo;v}`~aIoM1&eT<IM(*K_rH*0hm
z@d7laQ$p*a9e7v}g^j&-T@umHhO6E3hkEqvKsPv;I6XFk8}M_Pv5*fWH6URw-h!5+
z`k`>WW$-}_LZftkY%D1v>LEb<WGh=d7rgxzvbNzNH~IPYYjo<o%4ruC^NaZeb=yV$
zexpGOgJftRKrCosB7a{Z4d-A38eO(xLxZer3&!39SfxCqkp;Y?1M@xUnmRgG=(1}t
z_wmtBF1-FHKBC({)sgn-t5>fgE)uX?xjXh^n^Kv{-+`LM82v`Q0&KNX5hc*ZS(={s
z5v(ycf%_{M;?m2ktsR{Q*{%eXrfCN8{a0M(--$|T%6}E;mcB7=4veFijskbhag5Wr
z=Uf~`dZ$v{otAlUj1pp>?+ERs1}j!Bm(%Yr^dpE9%M`pR^mH~!d3nsu7Ns|eL9B{g
zL3)gwx<q$lqw>r{Pe$Qu1Mb<_<ZVFVbi1msprFH1DCsNEPLCv;zJ11~8cu4R4yUDK
zU`R;J*go=Wcn8`ev|NLYCmu|)z?Zs#TigZ&XbV|EI%KJP`}+E`AKX%k<-7|GNW8*o
z!G#Y$saYm_W);;R*to5Y8-PWzI@$Bj4cM#^ppA)wu!aa00Q!fV^k)92ZhKx>p`ZQz
z(1*gp$5+v|Q&m;<ft)EO0|EjFf*Eb6mEgU(7fPP%wlokI%X&1_i68Kd82daApj4#f
z13w<0uyQasm1IB~%PT6<0-+Hp5kEh>m_U)=f=morAOvQLv4rA~jxR9(k#Rl>0F9p4
zd*|`-UaPnkXo(>4bc7B*HuBrgBxcn~g|3_>_lCnHK4{nlXR!f(kd&vVr$wu6G<XD{
zRmr@1<Lk}N&|VY7ZwmgP3;5)W37trP$HiMdDqz&)6xIu-rK4lw=Rf;~gbUz!gMQe$
z^(I`sO8cZU^IxqNwCIG>LI)WS;y1hMgI@^5eGT+iY2Uqj_t(ji3pC~o*Ls#hA!z8g
zZw^`;LQ)%RV($LSR|dPb^$p(h1R7a0P`mSBc6r*xhF)L?;062CwaD`OR0Q2%8_Nx%
z^pcBk1(#<&NRUY1*)QV87q}!gpf4#rc)$<wqz*W7y=vE7gaN%R6=($=Bk@9Rd9dou
zt83H6{>S;y9Gvs!4c^dWz9grwHxS=;xpg0XY5MIi-~?IuQ!!%S2lnu@R#;K5Ely$O
zd;xNNaQQuG4!SZ3{c${xe{qWL{4%v@wFw78+ZyC^q3Ji=W|bq*wond7rZYv{2wh`<
z_>*524&<scB%z0`hwHp6hg9`w+1OTM{iX59co<n9SSt1o8u=H3nL(d_!lx9To`vsu
z=!KA22n{OA$k!swa`@+3_hniCNEC#qP*U%EXalz{4+<Xmbqmz{JZNzCJ=o$B*2~ZM
zcPrgvsPQg~1r2~$)CpfHF=Ah%#G+NuExcBnUQoaar!`K33Rqi9^c&(Eyf#sY%@_pt
zi?@(NKz~0r6&|!(l*Un%^n;MF)k0##$i~*W%*f^p+yP}NPj8Vy&Ch&oj>xM?@gUCU
z!ZE4_<pFjPGl<w9^%DHZhadA{VPe1ws-HhU50&Qv67BLd^IZ|=0Y$Fne|=X&<v*q2
z?Z?Hf_}3FL@wuRJT<IP+v{<LZI9LP)b>rgV=;>|iA?hQ2;fTl#?bt-M%FQci7IyX@
zGp?Qd(0hoOrs0*$92|ql_fT~CpQz_+72f%0`%<_Zn}d9cIP%bg<`9~a5br5?R5;+%
ztOp8oKMp-cc0y<905#$o6`V_i=<X$lI0?j#)}@esK`+>L@Vp7+@nB*$H82w|@2nrm
zeS!;ZPd>ZC43>06P=<b^G|$aH*5Fe@R!lQr3kr+Nq?{5+s|2(vWP;+dn3Z6EM|k*B
z(z1&K$QWLePc9<;nQ$DfL5hH#JXqy4-GNJSsUtb?P1d;Yu7Ol9S6FSWNNGCn$i`pA
zJH%{yW{`v8AT0I3B$}rjNudFh`(6l{`4PCQIEA$mum{Z_71izN(ZJd{3XtEscQ4Qh
zBqA~x1|Q_vT4>e3helUSxeA5Ah{zs@AD2N|GVjmR45T#$Z?uK1H7}r?`w+jsb#(BU
zn3{%&`5ur$+i2L-=6(m<2xz}LwFKKsD{^@V(cvj64nQr%Q7m1AiE$8MlY(!H2dE#s
z>^|!^WW3A;{u7cfS~!~=b6rs2G$b4Re~Ze>ZbSA*%uUOrsP$)z0VwLPU^#<Uf9H<r
zZ1LmVp&k@f#`C$5n;4Vl{h*Q82KwWbDggUL_M#-LOwJQXNIMB7fkb2XcYgF!X~scc
zDK0T<H`r4lHD|;05WU&{?*WoMn{YWapf3TzA^P^8WRP9Cv)^uXk_J)b+kz{6GlCKb
zH~Nut4aYnB$IF)D>OMZTz}Xc?vuH!>?1cbgaP+@qhFkCx44d0vCA5KTSG`)jROWk#
zmRAFU4(TR_+={q%k#9}dg6?{LCJz{q3P?Ce4=IuiShR~Y8vKsqkKPQGS|aUkJ&3(Z
zk^Pf2(s~5RL3qm-v+>Fu6qV%6?CcG`gibj*Ijg@OU~;|;uvz{q#JnZ?;v0g_8BR-u
zm1Kd&q{v$LfCnWV<ovxOa7Y{L6l$esWPIVoYnFpB3dcP8n#EhQqm$i+D}F|*s>X76
z0KIkcd8eQp^QtbuZTvL(6u0cr=cn`|5s{J4jOPCQ;qzpN90BahJ6Z&sGzVT*hmNN8
zeU~m#Z8$Q*?yF#>QH2&9*reOfZa|zTI}*Bm`!0KoB3OVOR`h|)k;OPqpbDrYRFG~&
z3wo2{%Z={ezaO~<J*Hr$^OnX8(3&EXK&hI8%dFK$+^U;OzBl#mvn{x}dBB|R9YNBf
z%J0v^ICp`OiAf&Xh6cc<E5yT%D&vJ7a%m!vsvtqqk1V5Ws?jUV^2lv>{S#-oh;-^A
z4oHZoHt?-H5gG8^H&zfA0MqyR?v7;xI<@*c({>h*A^6DmB7ZajM)yqF4W$65jx*vb
z2ZOJ3YrQJ!jHn=-1&5!So}ON$-MI^R-Jul1wEbP(2K7Fl{5BINAgC&F)UH3haD|dp
z9&z|VeB=Q%g-FKW>PCz-e>xLLM6@8%K$FHc;NV3dku(O247LHF)=rO8p~^7+=_E#_
z(&(@f7wVKq!d~JQ#zDotw*7Am?|7LiAm{US4wOn5(t!N`?F+9qPf;H?zU~<;uNcDd
PMaj!3OBYE#diuWr<epcA

literal 0
HcmV?d00001

diff --git a/cifar10-fast-simple/doc/plot.py b/cifar10-fast-simple/doc/plot.py
new file mode 100644
index 0000000..ef79b95
--- /dev/null
+++ b/cifar10-fast-simple/doc/plot.py
@@ -0,0 +1,28 @@
+import matplotlib.pyplot as plt
+
+result = """
+    1       97                4.37                 0.2109
+    2      194                7.77                 0.7620
+    3      291               11.16                 0.8764
+    4      388               14.54                 0.8979
+    5      485               17.93                 0.9098
+    6      582               21.32                 0.9177
+    7      679               24.71                 0.9280
+    8      776               28.09                 0.9332
+    9      873               31.48                 0.9395
+   10      970               34.86                 0.9430
+"""
+
+rows = []
+for row in result.strip().split("\n"):
+    numbers = [float(x) for x in row.split()]
+    rows.append(numbers)
+
+epoch, batch, t, accuracy = map(list, zip(*rows))
+
+plt.plot(epoch, [100 - 100 * x for x in accuracy])
+plt.xticks(epoch)
+plt.xlabel("Epoch")
+plt.ylabel("Validation error [%]")
+plt.savefig("a100_epoch_vs_validation_error.png")
+plt.show()
diff --git a/cifar10-fast-simple/logs/A100.txt b/cifar10-fast-simple/logs/A100.txt
new file mode 100644
index 0000000..a408c9d
--- /dev/null
+++ b/cifar10-fast-simple/logs/A100.txt
@@ -0,0 +1,2006 @@
+File   : train.py f6582527f1a15333cfbdec0506623bdfbc7b219b9d739d381c6949426e24aa1c
+Model  : model.py 0cf3352991a284cbf63f3f0e44c07fc3bffe61c9d720298ad8c18e57681b88c3
+PyTorch: 1.11.0+cu113
+Device : A100-SXM4-40GB
+Dtype  : torch.float16
+
+Files already downloaded and verified
+Preprocessing: 6.35 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                2.61                 0.2200
+    2      194                4.04                 0.7673
+    3      291                5.47                 0.8711
+    4      388                6.90                 0.8979
+    5      485                8.34                 0.9107
+    6      582                9.77                 0.9170
+    7      679               11.20                 0.9266
+    8      776               12.63                 0.9331
+    9      873               14.06                 0.9368
+   10      970               15.49                 0.9404
+
+1 of 1 runs >= 94.0 % accuracy
+Min  accuracy: 0.9404
+Max  accuracy: 0.9404
+Mean accuracy: 0.9404 +- 0.0
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1750
+    2      194                2.86                 0.7731
+    3      291                4.29                 0.8747
+    4      388                5.72                 0.8983
+    5      485                7.16                 0.9131
+    6      582                8.59                 0.9201
+    7      679               10.02                 0.9289
+    8      776               11.45                 0.9351
+    9      873               12.88                 0.9390
+   10      970               14.31                 0.9399
+
+1 of 2 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9404
+Mean accuracy: 0.94015 +- 0.00024999999999997247
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1826
+    2      194                2.86                 0.7749
+    3      291                4.29                 0.8765
+    4      388                5.73                 0.8994
+    5      485                7.16                 0.9108
+    6      582                8.59                 0.9192
+    7      679               10.02                 0.9280
+    8      776               11.45                 0.9334
+    9      873               12.88                 0.9386
+   10      970               14.31                 0.9417
+
+2 of 3 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9406666666666667 +- 0.0007586537784494201
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1823
+    2      194                2.86                 0.7603
+    3      291                4.29                 0.8788
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9131
+    6      582                8.59                 0.9219
+    7      679               10.02                 0.9288
+    8      776               11.45                 0.9332
+    9      873               12.88                 0.9373
+   10      970               14.32                 0.9417
+
+3 of 4 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.940925 +- 0.000794905654779253
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1855
+    2      194                2.86                 0.7723
+    3      291                4.29                 0.8809
+    4      388                5.73                 0.8989
+    5      485                7.16                 0.9124
+    6      582                8.59                 0.9221
+    7      679               10.02                 0.9278
+    8      776               11.45                 0.9340
+    9      873               12.88                 0.9373
+   10      970               14.32                 0.9406
+
+4 of 5 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.94086 +- 0.0007227724399837298
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2350
+    2      194                2.86                 0.7790
+    3      291                4.29                 0.8760
+    4      388                5.73                 0.9009
+    5      485                7.16                 0.9133
+    6      582                8.59                 0.9229
+    7      679               10.02                 0.9301
+    8      776               11.45                 0.9367
+    9      873               12.88                 0.9399
+   10      970               14.31                 0.9415
+
+5 of 6 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9409666666666666 +- 0.000701585505994987
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1847
+    2      194                2.86                 0.7648
+    3      291                4.29                 0.8751
+    4      388                5.73                 0.8963
+    5      485                7.16                 0.9097
+    6      582                8.59                 0.9185
+    7      679               10.02                 0.9262
+    8      776               11.45                 0.9317
+    9      873               12.88                 0.9378
+   10      970               14.32                 0.9402
+
+6 of 7 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9408571428571427 +- 0.0007027642214999471
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1709
+    2      194                2.86                 0.7574
+    3      291                4.29                 0.8743
+    4      388                5.73                 0.9000
+    5      485                7.16                 0.9133
+    6      582                8.59                 0.9208
+    7      679               10.02                 0.9288
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9379
+   10      970               14.32                 0.9401
+
+7 of 8 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9407625 +- 0.0007034513131695864
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2391
+    2      194                2.86                 0.7724
+    3      291                4.30                 0.8747
+    4      388                5.73                 0.8976
+    5      485                7.16                 0.9073
+    6      582                8.59                 0.9189
+    7      679               10.02                 0.9273
+    8      776               11.45                 0.9321
+    9      873               12.88                 0.9378
+   10      970               14.32                 0.9400
+
+8 of 9 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9406777777777777 +- 0.0007051836290054496
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2016
+    2      194                2.86                 0.7633
+    3      291                4.29                 0.8720
+    4      388                5.73                 0.8971
+    5      485                7.16                 0.9122
+    6      582                8.59                 0.9195
+    7      679               10.02                 0.9273
+    8      776               11.45                 0.9318
+    9      873               12.88                 0.9376
+   10      970               14.31                 0.9414
+
+9 of 10 runs >= 94.0 % accuracy
+Min  accuracy: 0.9399000000000001
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9407499999999999 +- 0.0007032069396699712
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1618
+    2      194                2.86                 0.7815
+    3      291                4.29                 0.8726
+    4      388                5.73                 0.8981
+    5      485                7.16                 0.9126
+    6      582                8.59                 0.9182
+    7      679               10.02                 0.9271
+    8      776               11.45                 0.9331
+    9      873               12.88                 0.9367
+   10      970               14.31                 0.9393
+
+9 of 11 runs >= 94.0 % accuracy
+Min  accuracy: 0.9393
+Max  accuracy: 0.9417000000000001
+Mean accuracy: 0.9406181818181817 +- 0.0007894971723343771
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1863
+    2      194                2.86                 0.7682
+    3      291                4.29                 0.8746
+    4      388                5.73                 0.8963
+    5      485                7.16                 0.9096
+    6      582                8.59                 0.9203
+    7      679               10.02                 0.9262
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9392
+   10      970               14.32                 0.9424
+
+10 of 12 runs >= 94.0 % accuracy
+Min  accuracy: 0.9393
+Max  accuracy: 0.9424
+Mean accuracy: 0.9407666666666664 +- 0.000902157906848292
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2488
+    2      194                2.86                 0.7649
+    3      291                4.30                 0.8752
+    4      388                5.73                 0.8987
+    5      485                7.16                 0.9114
+    6      582                8.59                 0.9201
+    7      679               10.02                 0.9285
+    8      776               11.45                 0.9343
+    9      873               12.88                 0.9371
+   10      970               14.32                 0.9400
+
+11 of 13 runs >= 94.0 % accuracy
+Min  accuracy: 0.9393
+Max  accuracy: 0.9424
+Mean accuracy: 0.940707692307692 +- 0.0008905154414603322
+
+Files already downloaded and verified
+Preprocessing: 1.16 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1639
+    2      194                2.86                 0.7551
+    3      291                4.29                 0.8769
+    4      388                5.73                 0.8994
+    5      485                7.16                 0.9134
+    6      582                8.59                 0.9239
+    7      679               10.02                 0.9295
+    8      776               11.45                 0.9362
+    9      873               12.88                 0.9427
+   10      970               14.32                 0.9437
+
+12 of 14 runs >= 94.0 % accuracy
+Min  accuracy: 0.9393
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9409214285714284 +- 0.001153366854925543
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1894
+    2      194                2.86                 0.7581
+    3      291                4.29                 0.8751
+    4      388                5.73                 0.8988
+    5      485                7.16                 0.9123
+    6      582                8.59                 0.9181
+    7      679               10.02                 0.9268
+    8      776               11.45                 0.9326
+    9      873               12.88                 0.9372
+   10      970               14.31                 0.9401
+
+13 of 15 runs >= 94.0 % accuracy
+Min  accuracy: 0.9393
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408666666666664 +- 0.0011329411085999024
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1954
+    2      194                2.86                 0.7762
+    3      291                4.30                 0.8809
+    4      388                5.73                 0.9007
+    5      485                7.16                 0.9116
+    6      582                8.59                 0.9191
+    7      679               10.02                 0.9281
+    8      776               11.45                 0.9331
+    9      873               12.89                 0.9382
+   10      970               14.32                 0.9399
+
+13 of 16 runs >= 94.0 % accuracy
+Min  accuracy: 0.9393
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408062499999997 +- 0.0011216443007923751
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1811
+    2      194                2.86                 0.7517
+    3      291                4.29                 0.8720
+    4      388                5.73                 0.8982
+    5      485                7.16                 0.9123
+    6      582                8.59                 0.9206
+    7      679               10.02                 0.9275
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9369
+   10      970               14.32                 0.9386
+
+13 of 17 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.940676470588235 +- 0.001205638425831252
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2231
+    2      194                2.86                 0.7598
+    3      291                4.29                 0.8750
+    4      388                5.73                 0.8980
+    5      485                7.16                 0.9117
+    6      582                8.59                 0.9219
+    7      679               10.02                 0.9281
+    8      776               11.45                 0.9345
+    9      873               12.88                 0.9390
+   10      970               14.31                 0.9400
+
+14 of 18 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9406388888888887 +- 0.0011818718152381616
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1511
+    2      194                2.86                 0.7385
+    3      291                4.30                 0.8708
+    4      388                5.73                 0.8956
+    5      485                7.16                 0.9111
+    6      582                8.59                 0.9205
+    7      679               10.02                 0.9294
+    8      776               11.45                 0.9353
+    9      873               12.88                 0.9393
+   10      970               14.32                 0.9429
+
+15 of 19 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.940757894736842 +- 0.0012562751079339411
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1677
+    2      194                2.86                 0.7383
+    3      291                4.29                 0.8695
+    4      388                5.73                 0.8973
+    5      485                7.16                 0.9098
+    6      582                8.59                 0.9175
+    7      679               10.02                 0.9266
+    8      776               11.45                 0.9333
+    9      873               12.88                 0.9389
+   10      970               14.31                 0.9418
+
+16 of 20 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408099999999999 +- 0.0012453513560437605
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2051
+    2      194                2.86                 0.7653
+    3      291                4.29                 0.8757
+    4      388                5.73                 0.9009
+    5      485                7.16                 0.9141
+    6      582                8.59                 0.9214
+    7      679               10.02                 0.9283
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9407
+   10      970               14.32                 0.9419
+
+17 of 21 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408619047619047 +- 0.0012373074416714115
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2169
+    2      194                2.86                 0.7749
+    3      291                4.29                 0.8800
+    4      388                5.73                 0.8976
+    5      485                7.16                 0.9109
+    6      582                8.59                 0.9191
+    7      679               10.02                 0.9282
+    8      776               11.45                 0.9330
+    9      873               12.88                 0.9379
+   10      970               14.32                 0.9408
+
+18 of 22 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408590909090908 +- 0.0012089285605290504
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1385
+    2      194                2.86                 0.7481
+    3      291                4.29                 0.8719
+    4      388                5.73                 0.8993
+    5      485                7.16                 0.9103
+    6      582                8.59                 0.9226
+    7      679               10.02                 0.9296
+    8      776               11.45                 0.9358
+    9      873               12.88                 0.9399
+   10      970               14.32                 0.9421
+
+19 of 23 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9409130434782608 +- 0.0012091334834162717
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1937
+    2      194                2.86                 0.7671
+    3      291                4.29                 0.8770
+    4      388                5.73                 0.8984
+    5      485                7.16                 0.9132
+    6      582                8.59                 0.9207
+    7      679               10.02                 0.9302
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9376
+   10      970               14.31                 0.9388
+
+19 of 24 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408249999999999 +- 0.0012567318727556903
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1742
+    2      194                2.86                 0.7835
+    3      291                4.29                 0.8715
+    4      388                5.73                 0.9001
+    5      485                7.16                 0.9076
+    6      582                8.59                 0.9186
+    7      679               10.02                 0.9259
+    8      776               11.45                 0.9344
+    9      873               12.88                 0.9387
+   10      970               14.31                 0.9422
+
+20 of 25 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9408799999999999 +- 0.0012604761005271072
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1694
+    2      194                2.86                 0.7748
+    3      291                4.29                 0.8763
+    4      388                5.73                 0.8980
+    5      485                7.16                 0.9105
+    6      582                8.59                 0.9214
+    7      679               10.02                 0.9286
+    8      776               11.45                 0.9353
+    9      873               12.88                 0.9401
+   10      970               14.32                 0.9432
+
+21 of 26 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9409692307692308 +- 0.0013140569097760556
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2075
+    2      194                2.86                 0.7711
+    3      291                4.29                 0.8775
+    4      388                5.73                 0.9025
+    5      485                7.16                 0.9130
+    6      582                8.59                 0.9185
+    7      679               10.02                 0.9285
+    8      776               11.45                 0.9347
+    9      873               12.88                 0.9377
+   10      970               14.31                 0.9400
+
+22 of 27 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9409333333333334 +- 0.001302419401074905
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1658
+    2      194                2.86                 0.7553
+    3      291                4.29                 0.8735
+    4      388                5.73                 0.8976
+    5      485                7.16                 0.9133
+    6      582                8.59                 0.9219
+    7      679               10.02                 0.9284
+    8      776               11.45                 0.9345
+    9      873               12.88                 0.9379
+   10      970               14.32                 0.9405
+
+23 of 28 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9409178571428571 +- 0.0012814761497776687
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1990
+    2      194                2.86                 0.7628
+    3      291                4.29                 0.8770
+    4      388                5.73                 0.8993
+    5      485                7.16                 0.9130
+    6      582                8.59                 0.9223
+    7      679               10.02                 0.9284
+    8      776               11.45                 0.9336
+    9      873               12.88                 0.9385
+   10      970               14.31                 0.9424
+
+24 of 29 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9437000000000001
+Mean accuracy: 0.9409689655172414 +- 0.001287902183123281
+
+Files already downloaded and verified
+Preprocessing: 1.30 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1776
+    2      194                2.86                 0.7540
+    3      291                4.29                 0.8721
+    4      388                5.73                 0.8991
+    5      485                7.16                 0.9127
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9278
+    8      776               11.45                 0.9356
+    9      873               12.88                 0.9391
+   10      970               14.31                 0.9438
+
+25 of 30 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410633333333334 +- 0.0013644249908131912
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1706
+    2      194                2.86                 0.7568
+    3      291                4.29                 0.8783
+    4      388                5.73                 0.9007
+    5      485                7.16                 0.9121
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9264
+    8      776               11.45                 0.9336
+    9      873               12.88                 0.9386
+   10      970               14.32                 0.9402
+
+26 of 31 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410354838709678 +- 0.0013508774656345519
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1975
+    2      194                2.86                 0.7504
+    3      291                4.29                 0.8703
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9133
+    6      582                8.59                 0.9207
+    7      679               10.02                 0.9271
+    8      776               11.45                 0.9335
+    9      873               12.88                 0.9376
+   10      970               14.31                 0.9404
+
+27 of 32 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.941015625 +- 0.0013341920249255808
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2214
+    2      194                2.86                 0.7783
+    3      291                4.29                 0.8785
+    4      388                5.73                 0.9015
+    5      485                7.16                 0.9117
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9281
+    8      776               11.45                 0.9332
+    9      873               12.88                 0.9373
+   10      970               14.32                 0.9387
+
+27 of 33 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409454545454546 +- 0.001372476396902734
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1526
+    2      194                2.86                 0.7857
+    3      291                4.30                 0.8825
+    4      388                5.73                 0.9006
+    5      485                7.16                 0.9116
+    6      582                8.59                 0.9216
+    7      679               10.02                 0.9292
+    8      776               11.45                 0.9344
+    9      873               12.88                 0.9377
+   10      970               14.32                 0.9409
+
+28 of 34 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409441176470589 +- 0.001352164099088963
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1339
+    2      194                2.86                 0.7638
+    3      291                4.30                 0.8749
+    4      388                5.73                 0.8977
+    5      485                7.16                 0.9115
+    6      582                8.59                 0.9203
+    7      679               10.02                 0.9270
+    8      776               11.45                 0.9335
+    9      873               12.89                 0.9372
+   10      970               14.32                 0.9387
+
+28 of 35 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9408799999999999 +- 0.0013841552348944548
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1608
+    2      194                2.86                 0.7766
+    3      291                4.29                 0.8748
+    4      388                5.73                 0.8990
+    5      485                7.16                 0.9151
+    6      582                8.59                 0.9228
+    7      679               10.02                 0.9298
+    8      776               11.45                 0.9352
+    9      873               12.89                 0.9407
+   10      970               14.32                 0.9431
+
+29 of 36 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409416666666667 +- 0.0014127150770374389
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1548
+    2      194                2.86                 0.7627
+    3      291                4.29                 0.8766
+    4      388                5.73                 0.8994
+    5      485                7.16                 0.9129
+    6      582                8.59                 0.9186
+    7      679               10.02                 0.9283
+    8      776               11.45                 0.9340
+    9      873               12.88                 0.9376
+   10      970               14.32                 0.9406
+
+30 of 37 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409324324324325 +- 0.0013945945945945983
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1331
+    2      194                2.86                 0.7475
+    3      291                4.29                 0.8756
+    4      388                5.73                 0.8987
+    5      485                7.16                 0.9111
+    6      582                8.59                 0.9212
+    7      679               10.02                 0.9299
+    8      776               11.45                 0.9353
+    9      873               12.88                 0.9381
+   10      970               14.32                 0.9400
+
+31 of 38 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409078947368421 +- 0.0013841930156987322
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1920
+    2      194                2.86                 0.7703
+    3      291                4.30                 0.8717
+    4      388                5.73                 0.8999
+    5      485                7.16                 0.9127
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9291
+    8      776               11.45                 0.9335
+    9      873               12.88                 0.9372
+   10      970               14.32                 0.9409
+
+32 of 39 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409076923076923 +- 0.001366332282179358
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2233
+    2      194                2.86                 0.7744
+    3      291                4.29                 0.8774
+    4      388                5.73                 0.9004
+    5      485                7.16                 0.9103
+    6      582                8.59                 0.9180
+    7      679               10.02                 0.9274
+    8      776               11.45                 0.9324
+    9      873               12.88                 0.9362
+   10      970               14.32                 0.9401
+
+33 of 40 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9408875 +- 0.0013550253687662117
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1517
+    2      194                2.86                 0.7652
+    3      291                4.30                 0.8746
+    4      388                5.73                 0.9017
+    5      485                7.16                 0.9127
+    6      582                8.59                 0.9223
+    7      679               10.02                 0.9287
+    8      776               11.45                 0.9352
+    9      873               12.88                 0.9385
+   10      970               14.32                 0.9417
+
+34 of 41 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409073170731707 +- 0.0013442543030141032
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1398
+    2      194                2.86                 0.7663
+    3      291                4.29                 0.8771
+    4      388                5.73                 0.9026
+    5      485                7.16                 0.9138
+    6      582                8.59                 0.9227
+    7      679               10.02                 0.9285
+    8      776               11.45                 0.9327
+    9      873               12.88                 0.9382
+   10      970               14.31                 0.9410
+
+35 of 42 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409095238095239 +- 0.0013282300295272198
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1710
+    2      194                2.86                 0.7369
+    3      291                4.30                 0.8745
+    4      388                5.73                 0.8990
+    5      485                7.16                 0.9128
+    6      582                8.59                 0.9217
+    7      679               10.02                 0.9298
+    8      776               11.45                 0.9341
+    9      873               12.88                 0.9372
+   10      970               14.32                 0.9415
+
+36 of 43 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409232558139534 +- 0.0013157078224644379
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1305
+    2      194                2.86                 0.7610
+    3      291                4.29                 0.8699
+    4      388                5.73                 0.9012
+    5      485                7.16                 0.9139
+    6      582                8.59                 0.9224
+    7      679               10.02                 0.9303
+    8      776               11.45                 0.9345
+    9      873               12.88                 0.9401
+   10      970               14.31                 0.9424
+
+37 of 44 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409568181818181 +- 0.0013191591204718153
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1573
+    2      194                2.86                 0.7738
+    3      291                4.30                 0.8788
+    4      388                5.73                 0.8990
+    5      485                7.16                 0.9100
+    6      582                8.59                 0.9157
+    7      679               10.02                 0.9245
+    8      776               11.45                 0.9314
+    9      873               12.89                 0.9369
+   10      970               14.32                 0.9404
+
+38 of 45 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409444444444444 +- 0.0013069992018284683
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1518
+    2      194                2.86                 0.7656
+    3      291                4.29                 0.8689
+    4      388                5.73                 0.8969
+    5      485                7.16                 0.9113
+    6      582                8.59                 0.9207
+    7      679               10.02                 0.9282
+    8      776               11.45                 0.9349
+    9      873               12.88                 0.9388
+   10      970               14.31                 0.9423
+
+39 of 46 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409739130434782 +- 0.0013077419607540432
+
+Files already downloaded and verified
+Preprocessing: 1.21 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1756
+    2      194                2.86                 0.7715
+    3      291                4.30                 0.8784
+    4      388                5.73                 0.8988
+    5      485                7.16                 0.9130
+    6      582                8.59                 0.9207
+    7      679               10.02                 0.9282
+    8      776               11.45                 0.9339
+    9      873               12.89                 0.9380
+   10      970               14.32                 0.9420
+
+40 of 47 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.940995744680851 +- 0.0013022006654544492
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1981
+    2      194                2.86                 0.7759
+    3      291                4.29                 0.8776
+    4      388                5.73                 0.8999
+    5      485                7.16                 0.9135
+    6      582                8.59                 0.9210
+    7      679               10.02                 0.9299
+    8      776               11.45                 0.9345
+    9      873               12.88                 0.9395
+   10      970               14.31                 0.9415
+
+41 of 48 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410062499999999 +- 0.001290575816254126
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1363
+    2      194                2.86                 0.7775
+    3      291                4.30                 0.8744
+    4      388                5.73                 0.8985
+    5      485                7.16                 0.9094
+    6      582                8.59                 0.9211
+    7      679               10.02                 0.9283
+    8      776               11.45                 0.9341
+    9      873               12.88                 0.9398
+   10      970               14.32                 0.9426
+
+42 of 49 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.941038775510204 +- 0.0012970636109415849
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2250
+    2      194                2.86                 0.7678
+    3      291                4.29                 0.8731
+    4      388                5.73                 0.9009
+    5      485                7.16                 0.9141
+    6      582                8.59                 0.9226
+    7      679               10.02                 0.9290
+    8      776               11.45                 0.9354
+    9      873               12.88                 0.9384
+   10      970               14.32                 0.9398
+
+42 of 50 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410139999999998 +- 0.0012956866905236
+
+Files already downloaded and verified
+Preprocessing: 1.16 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1939
+    2      194                2.86                 0.7838
+    3      291                4.29                 0.8782
+    4      388                5.73                 0.9011
+    5      485                7.16                 0.9123
+    6      582                8.59                 0.9210
+    7      679               10.02                 0.9266
+    8      776               11.45                 0.9322
+    9      873               12.88                 0.9383
+   10      970               14.32                 0.9413
+
+43 of 51 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410196078431371 +- 0.0012835336652158391
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2084
+    2      194                2.86                 0.7610
+    3      291                4.29                 0.8758
+    4      388                5.73                 0.8994
+    5      485                7.16                 0.9115
+    6      582                8.59                 0.9219
+    7      679               10.02                 0.9290
+    8      776               11.45                 0.9323
+    9      873               12.88                 0.9375
+   10      970               14.32                 0.9402
+
+44 of 52 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410038461538459 +- 0.00127610608711115
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1918
+    2      194                2.86                 0.7624
+    3      291                4.29                 0.8777
+    4      388                5.73                 0.8997
+    5      485                7.16                 0.9119
+    6      582                8.59                 0.9217
+    7      679               10.02                 0.9295
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9382
+   10      970               14.31                 0.9391
+
+44 of 53 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409679245283017 +- 0.0012902791251194902
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2120
+    2      194                2.86                 0.7683
+    3      291                4.29                 0.8776
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9099
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9284
+    8      776               11.45                 0.9339
+    9      873               12.88                 0.9380
+   10      970               14.32                 0.9414
+
+45 of 54 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409759259259258 +- 0.001279602829756342
+
+Files already downloaded and verified
+Preprocessing: 1.17 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1534
+    2      194                2.86                 0.7741
+    3      291                4.29                 0.8761
+    4      388                5.73                 0.8978
+    5      485                7.16                 0.9095
+    6      582                8.59                 0.9205
+    7      679               10.02                 0.9260
+    8      776               11.45                 0.9334
+    9      873               12.88                 0.9379
+   10      970               14.31                 0.9407
+
+46 of 55 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.940970909090909 +- 0.0012684525615199195
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1579
+    2      194                2.86                 0.7788
+    3      291                4.29                 0.8765
+    4      388                5.73                 0.8975
+    5      485                7.16                 0.9115
+    6      582                8.59                 0.9200
+    7      679               10.02                 0.9275
+    8      776               11.45                 0.9358
+    9      873               12.88                 0.9399
+   10      970               14.32                 0.9412
+
+47 of 56 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409749999999999 +- 0.0012574421315852723
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1582
+    2      194                2.86                 0.7814
+    3      291                4.29                 0.8789
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9148
+    6      582                8.59                 0.9203
+    7      679               10.02                 0.9287
+    8      776               11.45                 0.9341
+    9      873               12.88                 0.9384
+   10      970               14.31                 0.9397
+
+47 of 57 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409526315789473 +- 0.0012575533560105088
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1567
+    2      194                2.86                 0.7726
+    3      291                4.29                 0.8755
+    4      388                5.73                 0.9026
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9182
+    7      679               10.02                 0.9266
+    8      776               11.45                 0.9323
+    9      873               12.88                 0.9373
+   10      970               14.32                 0.9409
+
+48 of 58 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409517241379309 +- 0.0012466840679496254
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2288
+    2      194                2.86                 0.7648
+    3      291                4.29                 0.8718
+    4      388                5.73                 0.8973
+    5      485                7.16                 0.9102
+    6      582                8.59                 0.9193
+    7      679               10.02                 0.9270
+    8      776               11.45                 0.9327
+    9      873               12.88                 0.9372
+   10      970               14.31                 0.9416
+
+49 of 59 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409627118644067 +- 0.0012389030556846066
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1764
+    2      194                2.86                 0.7733
+    3      291                4.29                 0.8805
+    4      388                5.72                 0.8990
+    5      485                7.16                 0.9105
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9280
+    8      776               11.45                 0.9341
+    9      873               12.88                 0.9378
+   10      970               14.32                 0.9408
+
+50 of 60 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.94096 +- 0.0012287120628256764
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2281
+    2      194                2.86                 0.7761
+    3      291                4.29                 0.8799
+    4      388                5.73                 0.8993
+    5      485                7.16                 0.9090
+    6      582                8.59                 0.9209
+    7      679               10.02                 0.9291
+    8      776               11.45                 0.9350
+    9      873               12.88                 0.9379
+   10      970               14.31                 0.9404
+
+51 of 61 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409508196721311 +- 0.0012206720737954772
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1591
+    2      194                2.86                 0.7880
+    3      291                4.29                 0.8783
+    4      388                5.73                 0.8986
+    5      485                7.16                 0.9093
+    6      582                8.59                 0.9216
+    7      679               10.02                 0.9277
+    8      776               11.45                 0.9344
+    9      873               12.88                 0.9398
+   10      970               14.32                 0.9438
+
+52 of 62 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409967741935483 +- 0.0012628651170454254
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2074
+    2      194                2.86                 0.7707
+    3      291                4.29                 0.8757
+    4      388                5.73                 0.8971
+    5      485                7.16                 0.9089
+    6      582                8.59                 0.9189
+    7      679               10.02                 0.9280
+    8      776               11.45                 0.9325
+    9      873               12.88                 0.9373
+   10      970               14.31                 0.9411
+
+53 of 63 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409984126984127 +- 0.0012528687162493164
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2003
+    2      194                2.86                 0.7585
+    3      291                4.29                 0.8755
+    4      388                5.73                 0.9012
+    5      485                7.16                 0.9132
+    6      582                8.59                 0.9218
+    7      679               10.02                 0.9283
+    8      776               11.45                 0.9340
+    9      873               12.88                 0.9391
+   10      970               14.32                 0.9406
+
+54 of 64 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409921875 +- 0.0012440237999506885
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1777
+    2      194                2.86                 0.7455
+    3      291                4.29                 0.8687
+    4      388                5.73                 0.8964
+    5      485                7.16                 0.9086
+    6      582                8.59                 0.9202
+    7      679               10.02                 0.9305
+    8      776               11.45                 0.9350
+    9      873               12.88                 0.9387
+   10      970               14.31                 0.9414
+
+55 of 65 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409984615384616 +- 0.0012354373013001455
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1714
+    2      194                2.86                 0.7786
+    3      291                4.29                 0.8730
+    4      388                5.73                 0.9006
+    5      485                7.16                 0.9120
+    6      582                8.59                 0.9218
+    7      679               10.02                 0.9298
+    8      776               11.45                 0.9357
+    9      873               12.89                 0.9399
+   10      970               14.32                 0.9408
+
+56 of 66 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409954545454546 +- 0.0012262818671214433
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1950
+    2      194                2.86                 0.7619
+    3      291                4.29                 0.8790
+    4      388                5.73                 0.8987
+    5      485                7.16                 0.9116
+    6      582                8.59                 0.9226
+    7      679               10.02                 0.9288
+    8      776               11.45                 0.9352
+    9      873               12.88                 0.9395
+   10      970               14.31                 0.9415
+
+57 of 67 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410029850746269 +- 0.0012186327240785646
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1849
+    2      194                2.86                 0.7589
+    3      291                4.29                 0.8778
+    4      388                5.73                 0.8996
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9210
+    7      679               10.02                 0.9282
+    8      776               11.45                 0.9347
+    9      873               12.89                 0.9400
+   10      970               14.32                 0.9419
+
+58 of 68 runs >= 94.0 % accuracy
+Min  accuracy: 0.9386
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410161764705882 +- 0.0012144485911815442
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1663
+    2      194                2.86                 0.7818
+    3      291                4.29                 0.8776
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9130
+    6      582                8.59                 0.9203
+    7      679               10.02                 0.9278
+    8      776               11.45                 0.9325
+    9      873               12.88                 0.9367
+   10      970               14.31                 0.9380
+
+58 of 69 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409724637681158 +- 0.0012583499729742719
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1706
+    2      194                2.86                 0.7743
+    3      291                4.29                 0.8780
+    4      388                5.73                 0.8990
+    5      485                7.16                 0.9110
+    6      582                8.59                 0.9216
+    7      679               10.02                 0.9284
+    8      776               11.45                 0.9355
+    9      873               12.88                 0.9381
+   10      970               14.32                 0.9403
+
+59 of 70 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.940962857142857 +- 0.0012518753279519065
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1936
+    2      194                2.86                 0.7535
+    3      291                4.29                 0.8758
+    4      388                5.73                 0.8978
+    5      485                7.16                 0.9110
+    6      582                8.59                 0.9195
+    7      679               10.02                 0.9275
+    8      776               11.45                 0.9349
+    9      873               12.88                 0.9395
+   10      970               14.31                 0.9408
+
+60 of 71 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409605633802814 +- 0.0012431761767932648
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2558
+    2      194                2.86                 0.7785
+    3      291                4.29                 0.8749
+    4      388                5.73                 0.9016
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9228
+    7      679               10.02                 0.9280
+    8      776               11.45                 0.9330
+    9      873               12.88                 0.9382
+   10      970               14.32                 0.9420
+
+61 of 72 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409749999999997 +- 0.0012404916140161704
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2024
+    2      194                2.86                 0.7858
+    3      291                4.29                 0.8788
+    4      388                5.73                 0.8996
+    5      485                7.16                 0.9103
+    6      582                8.59                 0.9211
+    7      679               10.02                 0.9271
+    8      776               11.45                 0.9324
+    9      873               12.88                 0.9382
+   10      970               14.31                 0.9415
+
+62 of 73 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409821917808217 +- 0.0012334762621337761
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1816
+    2      194                2.86                 0.7755
+    3      291                4.29                 0.8742
+    4      388                5.73                 0.8991
+    5      485                7.16                 0.9107
+    6      582                8.59                 0.9207
+    7      679               10.02                 0.9287
+    8      776               11.46                 0.9351
+    9      873               12.89                 0.9388
+   10      970               14.32                 0.9408
+
+63 of 74 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409797297297294 +- 0.001225294198509046
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2062
+    2      194                2.86                 0.7855
+    3      291                4.29                 0.8772
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9197
+    7      679               10.02                 0.9270
+    8      776               11.45                 0.9324
+    9      873               12.88                 0.9362
+   10      970               14.32                 0.9394
+
+63 of 75 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409586666666665 +- 0.00123051136614914
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1848
+    2      194                2.86                 0.7730
+    3      291                4.29                 0.8757
+    4      388                5.73                 0.8957
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9203
+    7      679               10.02                 0.9267
+    8      776               11.45                 0.9350
+    9      873               12.89                 0.9383
+   10      970               14.32                 0.9409
+
+64 of 76 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409578947368419 +- 0.0012224073700666859
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1839
+    2      194                2.86                 0.7664
+    3      291                4.30                 0.8727
+    4      388                5.73                 0.8976
+    5      485                7.16                 0.9112
+    6      582                8.59                 0.9210
+    7      679               10.02                 0.9275
+    8      776               11.45                 0.9340
+    9      873               12.88                 0.9384
+   10      970               14.32                 0.9409
+
+65 of 77 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409571428571426 +- 0.0012144614082215085
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2224
+    2      194                2.86                 0.7789
+    3      291                4.29                 0.8746
+    4      388                5.72                 0.9005
+    5      485                7.16                 0.9117
+    6      582                8.59                 0.9220
+    7      679               10.02                 0.9293
+    8      776               11.45                 0.9324
+    9      873               12.88                 0.9396
+   10      970               14.32                 0.9413
+
+66 of 78 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409615384615382 +- 0.0012072676045655868
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1626
+    2      194                2.86                 0.7808
+    3      291                4.29                 0.8801
+    4      388                5.73                 0.9013
+    5      485                7.16                 0.9120
+    6      582                8.59                 0.9210
+    7      679               10.02                 0.9273
+    8      776               11.45                 0.9309
+    9      873               12.88                 0.9373
+   10      970               14.31                 0.9395
+
+66 of 79 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409430379746833 +- 0.0012106785987387404
+
+Files already downloaded and verified
+Preprocessing: 4.87 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.46                 0.1603
+    2      194                2.89                 0.7822
+    3      291                4.33                 0.8754
+    4      388                5.76                 0.8994
+    5      485                7.19                 0.9122
+    6      582                8.62                 0.9206
+    7      679               10.06                 0.9268
+    8      776               11.49                 0.9343
+    9      873               12.92                 0.9391
+   10      970               14.35                 0.9427
+
+67 of 80 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409649999999997 +- 0.0012188211517692017
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1789
+    2      194                2.86                 0.7684
+    3      291                4.29                 0.8763
+    4      388                5.73                 0.9000
+    5      485                7.16                 0.9121
+    6      582                8.59                 0.9199
+    7      679               10.02                 0.9268
+    8      776               11.45                 0.9335
+    9      873               12.88                 0.9375
+   10      970               14.31                 0.9403
+
+68 of 81 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409567901234565 +- 0.0012134979814625029
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2120
+    2      194                2.86                 0.7553
+    3      291                4.29                 0.8746
+    4      388                5.73                 0.8971
+    5      485                7.16                 0.9116
+    6      582                8.59                 0.9206
+    7      679               10.02                 0.9287
+    8      776               11.45                 0.9344
+    9      873               12.89                 0.9387
+   10      970               14.32                 0.9404
+
+69 of 82 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409499999999996 +- 0.0012076231442534504
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1668
+    2      194                2.86                 0.7615
+    3      291                4.29                 0.8732
+    4      388                5.73                 0.9000
+    5      485                7.16                 0.9112
+    6      582                8.59                 0.9186
+    7      679               10.02                 0.9258
+    8      776               11.45                 0.9341
+    9      873               12.88                 0.9380
+   10      970               14.31                 0.9421
+
+70 of 83 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409638554216864 +- 0.0012068657318690286
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1906
+    2      194                2.86                 0.7623
+    3      291                4.29                 0.8724
+    4      388                5.73                 0.8991
+    5      485                7.16                 0.9102
+    6      582                8.59                 0.9211
+    7      679               10.02                 0.9287
+    8      776               11.45                 0.9357
+    9      873               12.89                 0.9394
+   10      970               14.32                 0.9421
+
+71 of 84 runs >= 94.0 % accuracy
+Min  accuracy: 0.9380000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409773809523805 +- 0.0012059723591632458
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1760
+    2      194                2.86                 0.7717
+    3      291                4.29                 0.8754
+    4      388                5.73                 0.9013
+    5      485                7.16                 0.9114
+    6      582                8.59                 0.9196
+    7      679               10.02                 0.9262
+    8      776               11.45                 0.9319
+    9      873               12.88                 0.9358
+   10      970               14.31                 0.9379
+
+71 of 85 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409411764705877 +- 0.0012439305936423105
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2114
+    2      194                2.86                 0.7834
+    3      291                4.29                 0.8785
+    4      388                5.73                 0.9012
+    5      485                7.16                 0.9110
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9285
+    8      776               11.45                 0.9354
+    9      873               12.88                 0.9391
+   10      970               14.32                 0.9417
+
+72 of 86 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409499999999995 +- 0.0012393499796639477
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1915
+    2      194                2.86                 0.7640
+    3      291                4.29                 0.8795
+    4      388                5.73                 0.9007
+    5      485                7.16                 0.9126
+    6      582                8.59                 0.9205
+    7      679               10.02                 0.9266
+    8      776               11.45                 0.9350
+    9      873               12.88                 0.9394
+   10      970               14.31                 0.9415
+
+73 of 87 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409563218390801 +- 0.0012336005752180202
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1896
+    2      194                2.86                 0.7769
+    3      291                4.29                 0.8762
+    4      388                5.72                 0.8991
+    5      485                7.16                 0.9128
+    6      582                8.59                 0.9217
+    7      679               10.02                 0.9296
+    8      776               11.45                 0.9363
+    9      873               12.88                 0.9396
+   10      970               14.32                 0.9413
+
+74 of 88 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409602272727269 +- 0.0012271122580276247
+
+Files already downloaded and verified
+Preprocessing: 1.19 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1637
+    2      194                2.86                 0.7730
+    3      291                4.29                 0.8763
+    4      388                5.73                 0.8981
+    5      485                7.16                 0.9110
+    6      582                8.59                 0.9217
+    7      679               10.02                 0.9269
+    8      776               11.45                 0.9325
+    9      873               12.88                 0.9384
+   10      970               14.32                 0.9418
+
+75 of 89 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409696629213479 +- 0.0012234051295858975
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1619
+    2      194                2.86                 0.7752
+    3      291                4.29                 0.8743
+    4      388                5.73                 0.9001
+    5      485                7.16                 0.9114
+    6      582                8.59                 0.9221
+    7      679               10.02                 0.9284
+    8      776               11.45                 0.9351
+    9      873               12.89                 0.9388
+   10      970               14.32                 0.9417
+
+76 of 90 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409777777777774 +- 0.0012189957412903276
+
+Files already downloaded and verified
+Preprocessing: 1.21 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1508
+    2      194                2.86                 0.7641
+    3      291                4.29                 0.8788
+    4      388                5.73                 0.9009
+    5      485                7.16                 0.9135
+    6      582                8.59                 0.9217
+    7      679               10.02                 0.9273
+    8      776               11.45                 0.9355
+    9      873               12.88                 0.9369
+   10      970               14.32                 0.9399
+
+76 of 91 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409659340659335 +- 0.001217475299767027
+
+Files already downloaded and verified
+Preprocessing: 2.33 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.45                 0.1757
+    2      194                2.89                 0.7658
+    3      291                4.32                 0.8735
+    4      388                5.75                 0.8987
+    5      485                7.18                 0.9091
+    6      582                8.61                 0.9202
+    7      679               10.05                 0.9292
+    8      776               11.48                 0.9349
+    9      873               12.91                 0.9373
+   10      970               14.34                 0.9400
+
+77 of 92 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409554347826081 +- 0.0012149757673764164
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2004
+    2      194                2.86                 0.7658
+    3      291                4.29                 0.8757
+    4      388                5.73                 0.9004
+    5      485                7.16                 0.9131
+    6      582                8.59                 0.9219
+    7      679               10.02                 0.9304
+    8      776               11.45                 0.9354
+    9      873               12.88                 0.9385
+   10      970               14.32                 0.9410
+
+78 of 93 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409559139784941 +- 0.0012084347260915502
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1777
+    2      194                2.86                 0.7851
+    3      291                4.29                 0.8757
+    4      388                5.73                 0.9008
+    5      485                7.16                 0.9122
+    6      582                8.59                 0.9200
+    7      679               10.02                 0.9287
+    8      776               11.45                 0.9337
+    9      873               12.88                 0.9388
+   10      970               14.32                 0.9422
+
+79 of 94 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409691489361697 +- 0.001208747073814997
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2026
+    2      194                2.86                 0.7886
+    3      291                4.29                 0.8795
+    4      388                5.73                 0.8995
+    5      485                7.16                 0.9102
+    6      582                8.59                 0.9238
+    7      679               10.02                 0.9288
+    8      776               11.45                 0.9331
+    9      873               12.88                 0.9374
+   10      970               14.32                 0.9423
+
+80 of 95 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409831578947364 +- 0.0012100154528442541
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1473
+    2      194                2.86                 0.7702
+    3      291                4.29                 0.8831
+    4      388                5.73                 0.9019
+    5      485                7.16                 0.9123
+    6      582                8.59                 0.9206
+    7      679               10.02                 0.9281
+    8      776               11.45                 0.9342
+    9      873               12.88                 0.9395
+   10      970               14.32                 0.9430
+
+81 of 96 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9410041666666662 +- 0.0012209897510717378
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.2115
+    2      194                2.86                 0.7763
+    3      291                4.30                 0.8766
+    4      388                5.73                 0.8973
+    5      485                7.16                 0.9102
+    6      582                8.59                 0.9186
+    7      679               10.02                 0.9270
+    8      776               11.45                 0.9322
+    9      873               12.89                 0.9379
+   10      970               14.32                 0.9404
+
+82 of 97 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409979381443294 +- 0.0012162117440985179
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1600
+    2      194                2.86                 0.7649
+    3      291                4.29                 0.8754
+    4      388                5.72                 0.9016
+    5      485                7.16                 0.9099
+    6      582                8.59                 0.9211
+    7      679               10.02                 0.9281
+    8      776               11.45                 0.9342
+    9      873               12.88                 0.9391
+   10      970               14.32                 0.9390
+
+82 of 98 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409775510204076 +- 0.0012265374036281819
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1522
+    2      194                2.86                 0.7512
+    3      291                4.29                 0.8724
+    4      388                5.73                 0.8958
+    5      485                7.16                 0.9115
+    6      582                8.59                 0.9215
+    7      679               10.02                 0.9285
+    8      776               11.45                 0.9346
+    9      873               12.88                 0.9387
+   10      970               14.32                 0.9410
+
+83 of 99 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409777777777772 +- 0.0012203291123858678
+
+Files already downloaded and verified
+Preprocessing: 1.18 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                1.43                 0.1557
+    2      194                2.86                 0.7767
+    3      291                4.29                 0.8756
+    4      388                5.73                 0.8975
+    5      485                7.16                 0.9118
+    6      582                8.59                 0.9204
+    7      679               10.02                 0.9294
+    8      776               11.45                 0.9373
+    9      873               12.88                 0.9401
+   10      970               14.32                 0.9427
+
+84 of 100 runs >= 94.0 % accuracy
+Min  accuracy: 0.9379000000000001
+Max  accuracy: 0.9438000000000001
+Mean accuracy: 0.9409949999999995 +- 0.0012262442660416419
+
diff --git a/cifar10-fast-simple/logs/P100.txt b/cifar10-fast-simple/logs/P100.txt
new file mode 100644
index 0000000..c131a8a
--- /dev/null
+++ b/cifar10-fast-simple/logs/P100.txt
@@ -0,0 +1,2006 @@
+File   : train.py 86f88a78d9ab160bcc35ec677f85faf945d1131a3ede2282ebc808596e250907
+Model  : model.py 0cf3352991a284cbf63f3f0e44c07fc3bffe61c9d720298ad8c18e57681b88c3
+PyTorch: 1.11.0+cu113
+Device : Tesla P100-PCIE-16GB
+Dtype  : torch.float16
+
+Files already downloaded and verified
+Preprocessing: 3.73 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                9.96                 0.2183
+    2      194               18.35                 0.7741
+    3      291               26.74                 0.8689
+    4      388               35.12                 0.8988
+    5      485               43.51                 0.9121
+    6      582               51.89                 0.9211
+    7      679               60.27                 0.9273
+    8      776               68.65                 0.9318
+    9      873               77.04                 0.9364
+   10      970               85.42                 0.9405
+
+1 of 1 runs >= 94.0 % accuracy
+Min  accuracy: 0.9405
+Max  accuracy: 0.9405
+Mean accuracy: 0.9405 +- 0.0
+
+Files already downloaded and verified
+Preprocessing: 1.35 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.39                 0.1883
+    2      194               16.77                 0.7746
+    3      291               25.16                 0.8727
+    4      388               33.54                 0.9000
+    5      485               41.92                 0.9098
+    6      582               50.31                 0.9203
+    7      679               58.69                 0.9279
+    8      776               67.08                 0.9348
+    9      873               75.46                 0.9393
+   10      970               83.85                 0.9428
+
+2 of 2 runs >= 94.0 % accuracy
+Min  accuracy: 0.9405
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.9416500000000001 +- 0.0011500000000000399
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.39                 0.1840
+    2      194               16.77                 0.7696
+    3      291               25.16                 0.8745
+    4      388               33.54                 0.8971
+    5      485               41.93                 0.9123
+    6      582               50.31                 0.9200
+    7      679               58.69                 0.9282
+    8      776               67.08                 0.9342
+    9      873               75.46                 0.9389
+   10      970               83.85                 0.9414
+
+3 of 3 runs >= 94.0 % accuracy
+Min  accuracy: 0.9405
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.9415666666666667 +- 0.0009463379711052599
+
+Files already downloaded and verified
+Preprocessing: 1.33 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1896
+    2      194               16.77                 0.7562
+    3      291               25.15                 0.8768
+    4      388               33.54                 0.8971
+    5      485               41.92                 0.9150
+    6      582               50.31                 0.9197
+    7      679               58.69                 0.9254
+    8      776               67.07                 0.9341
+    9      873               75.46                 0.9378
+   10      970               83.84                 0.9418
+
+4 of 4 runs >= 94.0 % accuracy
+Min  accuracy: 0.9405
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.941625 +- 0.000825757228245719
+
+Files already downloaded and verified
+Preprocessing: 1.37 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1951
+    2      194               16.77                 0.7719
+    3      291               25.15                 0.8795
+    4      388               33.54                 0.8969
+    5      485               41.92                 0.9098
+    6      582               50.30                 0.9217
+    7      679               58.69                 0.9295
+    8      776               67.07                 0.9363
+    9      873               75.46                 0.9388
+   10      970               83.84                 0.9412
+
+5 of 5 runs >= 94.0 % accuracy
+Min  accuracy: 0.9405
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.94154 +- 0.0007578918128598851
+
+Files already downloaded and verified
+Preprocessing: 1.34 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.39                 0.2589
+    2      194               16.77                 0.7830
+    3      291               25.15                 0.8764
+    4      388               33.54                 0.9010
+    5      485               41.92                 0.9105
+    6      582               50.30                 0.9174
+    7      679               58.69                 0.9261
+    8      776               67.07                 0.9320
+    9      873               75.45                 0.9367
+   10      970               83.84                 0.9390
+
+5 of 6 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.9411166666666667 +- 0.001172485489130782
+
+Files already downloaded and verified
+Preprocessing: 1.35 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1756
+    2      194               16.77                 0.7820
+    3      291               25.15                 0.8783
+    4      388               33.54                 0.8994
+    5      485               41.92                 0.9105
+    6      582               50.30                 0.9194
+    7      679               58.68                 0.9277
+    8      776               67.07                 0.9334
+    9      873               75.45                 0.9382
+   10      970               83.84                 0.9414
+
+6 of 7 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.9411571428571428 +- 0.0010900290203963118
+
+Files already downloaded and verified
+Preprocessing: 1.35 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1676
+    2      194               16.77                 0.7640
+    3      291               25.15                 0.8803
+    4      388               33.53                 0.9012
+    5      485               41.92                 0.9124
+    6      582               50.30                 0.9221
+    7      679               58.68                 0.9266
+    8      776               67.07                 0.9328
+    9      873               75.45                 0.9375
+   10      970               83.83                 0.9399
+
+6 of 8 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.941 +- 0.0011011357772772667
+
+Files already downloaded and verified
+Preprocessing: 1.36 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2273
+    2      194               16.77                 0.7778
+    3      291               25.15                 0.8723
+    4      388               33.54                 0.8976
+    5      485               41.92                 0.9100
+    6      582               50.30                 0.9207
+    7      679               58.68                 0.9291
+    8      776               67.07                 0.9328
+    9      873               75.45                 0.9383
+   10      970               83.83                 0.9421
+
+7 of 9 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9428000000000001
+Mean accuracy: 0.9411222222222223 +- 0.001094204711476465
+
+Files already downloaded and verified
+Preprocessing: 1.33 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2068
+    2      194               16.77                 0.7633
+    3      291               25.15                 0.8706
+    4      388               33.54                 0.8972
+    5      485               41.92                 0.9108
+    6      582               50.30                 0.9190
+    7      679               58.68                 0.9271
+    8      776               67.07                 0.9327
+    9      873               75.45                 0.9389
+   10      970               83.84                 0.9431
+
+8 of 10 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9431
+Mean accuracy: 0.9413199999999999 +- 0.0011956588142108143
+
+Files already downloaded and verified
+Preprocessing: 1.34 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.39                 0.1576
+    2      194               16.77                 0.7786
+    3      291               25.15                 0.8768
+    4      388               33.54                 0.9014
+    5      485               41.92                 0.9126
+    6      582               50.30                 0.9199
+    7      679               58.69                 0.9259
+    8      776               67.07                 0.9328
+    9      873               75.46                 0.9383
+   10      970               83.84                 0.9415
+
+9 of 11 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9431
+Mean accuracy: 0.9413363636363635 +- 0.0011411897517883628
+
+Files already downloaded and verified
+Preprocessing: 1.37 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1953
+    2      194               16.77                 0.7587
+    3      291               25.15                 0.8765
+    4      388               33.53                 0.8995
+    5      485               41.92                 0.9116
+    6      582               50.30                 0.9224
+    7      679               58.68                 0.9285
+    8      776               67.07                 0.9360
+    9      873               75.45                 0.9393
+   10      970               83.83                 0.9423
+
+10 of 12 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9431
+Mean accuracy: 0.9414166666666666 +- 0.001124598693855825
+
+Files already downloaded and verified
+Preprocessing: 1.34 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2340
+    2      194               16.77                 0.7671
+    3      291               25.15                 0.8800
+    4      388               33.53                 0.9000
+    5      485               41.92                 0.9147
+    6      582               50.30                 0.9215
+    7      679               58.68                 0.9283
+    8      776               67.07                 0.9351
+    9      873               75.45                 0.9377
+   10      970               83.83                 0.9394
+
+10 of 13 runs >= 94.0 % accuracy
+Min  accuracy: 0.9390000000000001
+Max  accuracy: 0.9431
+Mean accuracy: 0.9412615384615385 +- 0.0012067365154628795
+
+Files already downloaded and verified
+Preprocessing: 1.34 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1725
+    2      194               16.77                 0.7548
+    3      291               25.15                 0.8766
+    4      388               33.54                 0.8988
+    5      485               41.92                 0.9115
+    6      582               50.30                 0.9207
+    7      679               58.69                 0.9280
+    8      776               67.07                 0.9321
+    9      873               75.45                 0.9359
+   10      970               83.84                 0.9384
+
+10 of 14 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9431
+Mean accuracy: 0.9410571428571428 +- 0.001376701173984028
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1930
+    2      194               16.77                 0.7692
+    3      291               25.15                 0.8733
+    4      388               33.53                 0.8985
+    5      485               41.92                 0.9106
+    6      582               50.30                 0.9204
+    7      679               58.68                 0.9281
+    8      776               67.07                 0.9334
+    9      873               75.45                 0.9397
+   10      970               83.83                 0.9415
+
+11 of 15 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9431
+Mean accuracy: 0.9410866666666666 +- 0.001334599398904246
+
+Files already downloaded and verified
+Preprocessing: 1.35 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1923
+    2      194               16.77                 0.7746
+    3      291               25.15                 0.8775
+    4      388               33.53                 0.8989
+    5      485               41.91                 0.9139
+    6      582               50.30                 0.9206
+    7      679               58.68                 0.9284
+    8      776               67.06                 0.9357
+    9      873               75.45                 0.9407
+   10      970               83.83                 0.9434
+
+12 of 16 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.94123125 +- 0.0014083317924054731
+
+Files already downloaded and verified
+Preprocessing: 1.33 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1765
+    2      194               16.77                 0.7611
+    3      291               25.15                 0.8765
+    4      388               33.53                 0.9004
+    5      485               41.92                 0.9116
+    6      582               50.30                 0.9210
+    7      679               58.68                 0.9281
+    8      776               67.06                 0.9363
+    9      873               75.45                 0.9402
+   10      970               83.83                 0.9421
+
+13 of 17 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412823529411765 +- 0.0013814890917870676
+
+Files already downloaded and verified
+Preprocessing: 1.33 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2296
+    2      194               16.77                 0.7657
+    3      291               25.15                 0.8742
+    4      388               33.53                 0.8968
+    5      485               41.91                 0.9109
+    6      582               50.30                 0.9209
+    7      679               58.68                 0.9290
+    8      776               67.06                 0.9347
+    9      873               75.45                 0.9396
+   10      970               83.83                 0.9422
+
+14 of 18 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9413333333333332 +- 0.0013589211407093054
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1566
+    2      194               16.77                 0.7418
+    3      291               25.15                 0.8733
+    4      388               33.53                 0.8958
+    5      485               41.91                 0.9107
+    6      582               50.30                 0.9218
+    7      679               58.68                 0.9291
+    8      776               67.06                 0.9347
+    9      873               75.44                 0.9391
+   10      970               83.83                 0.9427
+
+15 of 19 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9414052631578946 +- 0.0013574254595064747
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1636
+    2      194               16.77                 0.7496
+    3      291               25.15                 0.8732
+    4      388               33.53                 0.9000
+    5      485               41.92                 0.9122
+    6      582               50.30                 0.9200
+    7      679               58.68                 0.9292
+    8      776               67.06                 0.9356
+    9      873               75.45                 0.9387
+   10      970               83.83                 0.9414
+
+16 of 20 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9414049999999999 +- 0.0013230551764760303
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2156
+    2      194               16.77                 0.7821
+    3      291               25.15                 0.8759
+    4      388               33.53                 0.9004
+    5      485               41.92                 0.9127
+    6      582               50.30                 0.9224
+    7      679               58.68                 0.9300
+    8      776               67.06                 0.9357
+    9      873               75.45                 0.9384
+   10      970               83.83                 0.9414
+
+17 of 21 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.941404761904762 +- 0.0012911700822919939
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2418
+    2      194               16.77                 0.7750
+    3      291               25.15                 0.8798
+    4      388               33.53                 0.8963
+    5      485               41.91                 0.9123
+    6      582               50.30                 0.9205
+    7      679               58.68                 0.9271
+    8      776               67.06                 0.9340
+    9      873               75.44                 0.9384
+   10      970               83.83                 0.9413
+
+18 of 22 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9413999999999999 +- 0.001261672771291431
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1365
+    2      194               16.77                 0.7367
+    3      291               25.15                 0.8721
+    4      388               33.53                 0.8988
+    5      485               41.92                 0.9093
+    6      582               50.30                 0.9197
+    7      679               58.68                 0.9288
+    8      776               67.07                 0.9345
+    9      873               75.45                 0.9382
+   10      970               83.83                 0.9412
+
+19 of 23 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.941391304347826 +- 0.0012346141954960033
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1828
+    2      194               16.77                 0.7584
+    3      291               25.15                 0.8750
+    4      388               33.53                 0.9013
+    5      485               41.91                 0.9130
+    6      582               50.30                 0.9209
+    7      679               58.68                 0.9261
+    8      776               67.06                 0.9321
+    9      873               75.44                 0.9379
+   10      970               83.83                 0.9391
+
+19 of 24 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412958333333332 +- 0.0012924392850558086
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1716
+    2      194               16.77                 0.7875
+    3      291               25.15                 0.8739
+    4      388               33.53                 0.9006
+    5      485               41.91                 0.9120
+    6      582               50.29                 0.9227
+    7      679               58.67                 0.9296
+    8      776               67.05                 0.9349
+    9      873               75.43                 0.9381
+   10      970               83.82                 0.9405
+
+20 of 25 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412639999999999 +- 0.0012758934124761424
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1719
+    2      194               16.77                 0.7652
+    3      291               25.15                 0.8793
+    4      388               33.53                 0.9002
+    5      485               41.91                 0.9111
+    6      582               50.29                 0.9204
+    7      679               58.67                 0.9279
+    8      776               67.06                 0.9337
+    9      873               75.44                 0.9392
+   10      970               83.82                 0.9402
+
+21 of 26 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412230769230768 +- 0.0012677380499589118
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2125
+    2      194               16.77                 0.7705
+    3      291               25.15                 0.8770
+    4      388               33.53                 0.8998
+    5      485               41.91                 0.9107
+    6      582               50.29                 0.9200
+    7      679               58.67                 0.9286
+    8      776               67.05                 0.9344
+    9      873               75.44                 0.9378
+   10      970               83.82                 0.9405
+
+22 of 27 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9411962962962962 +- 0.0012515121169510055
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1776
+    2      194               16.76                 0.7447
+    3      291               25.15                 0.8718
+    4      388               33.53                 0.9007
+    5      485               41.91                 0.9143
+    6      582               50.29                 0.9199
+    7      679               58.67                 0.9270
+    8      776               67.05                 0.9338
+    9      873               75.44                 0.9374
+   10      970               83.82                 0.9419
+
+23 of 28 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412214285714285 +- 0.0012358794274457724
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.41                 0.1977
+    2      194               16.79                 0.7662
+    3      291               25.17                 0.8749
+    4      388               33.55                 0.8978
+    5      485               41.93                 0.9119
+    6      582               50.31                 0.9204
+    7      679               58.70                 0.9276
+    8      776               67.08                 0.9352
+    9      873               75.46                 0.9395
+   10      970               83.84                 0.9411
+
+24 of 29 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412172413793103 +- 0.0012145863395489032
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1735
+    2      194               16.77                 0.7464
+    3      291               25.15                 0.8715
+    4      388               33.53                 0.9001
+    5      485               41.91                 0.9103
+    6      582               50.29                 0.9219
+    7      679               58.67                 0.9297
+    8      776               67.05                 0.9361
+    9      873               75.43                 0.9395
+   10      970               83.82                 0.9417
+
+25 of 30 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412333333333333 +- 0.0011973118038153528
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1698
+    2      194               16.76                 0.7590
+    3      291               25.15                 0.8752
+    4      388               33.53                 0.8999
+    5      485               41.91                 0.9103
+    6      582               50.29                 0.9205
+    7      679               58.67                 0.9274
+    8      776               67.05                 0.9346
+    9      873               75.43                 0.9393
+   10      970               83.82                 0.9412
+
+26 of 31 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.941232258064516 +- 0.0011778567459788494
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2070
+    2      194               16.77                 0.7572
+    3      291               25.15                 0.8736
+    4      388               33.53                 0.9013
+    5      485               41.91                 0.9129
+    6      582               50.29                 0.9218
+    7      679               58.67                 0.9284
+    8      776               67.05                 0.9349
+    9      873               75.44                 0.9384
+   10      970               83.82                 0.9410
+
+27 of 32 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412249999999999 +- 0.001160010775812025
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2345
+    2      194               16.77                 0.7870
+    3      291               25.15                 0.8761
+    4      388               33.53                 0.8994
+    5      485               41.91                 0.9085
+    6      582               50.30                 0.9184
+    7      679               58.68                 0.9289
+    8      776               67.06                 0.9320
+    9      873               75.44                 0.9382
+   10      970               83.83                 0.9410
+
+28 of 33 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412181818181816 +- 0.0011429506061743373
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1765
+    2      194               16.77                 0.7703
+    3      291               25.15                 0.8779
+    4      388               33.53                 0.8997
+    5      485               41.91                 0.9122
+    6      582               50.30                 0.9198
+    7      679               58.68                 0.9267
+    8      776               67.06                 0.9325
+    9      873               75.45                 0.9367
+   10      970               83.83                 0.9393
+
+28 of 34 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9411617647058822 +- 0.0011717294171393138
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1325
+    2      194               16.77                 0.7603
+    3      291               25.15                 0.8734
+    4      388               33.53                 0.8991
+    5      485               41.91                 0.9128
+    6      582               50.30                 0.9198
+    7      679               58.68                 0.9281
+    8      776               67.06                 0.9325
+    9      873               75.44                 0.9382
+   10      970               83.83                 0.9411
+
+29 of 35 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9411599999999999 +- 0.0011549149628559616
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1759
+    2      194               16.77                 0.7756
+    3      291               25.15                 0.8741
+    4      388               33.53                 0.8987
+    5      485               41.91                 0.9117
+    6      582               50.30                 0.9202
+    7      679               58.68                 0.9284
+    8      776               67.06                 0.9336
+    9      873               75.45                 0.9397
+   10      970               83.83                 0.9432
+
+30 of 36 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412166666666665 +- 0.0011870832602082645
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1695
+    2      194               16.77                 0.7587
+    3      291               25.15                 0.8744
+    4      388               33.53                 0.8971
+    5      485               41.91                 0.9111
+    6      582               50.29                 0.9196
+    7      679               58.68                 0.9278
+    8      776               67.06                 0.9343
+    9      873               75.44                 0.9392
+   10      970               83.83                 0.9417
+
+31 of 37 runs >= 94.0 % accuracy
+Min  accuracy: 0.9384
+Max  accuracy: 0.9434
+Mean accuracy: 0.9412297297297295 +- 0.0011735519806460731
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1254
+    2      194               16.77                 0.7464
+    3      291               25.15                 0.8760
+    4      388               33.53                 0.8988
+    5      485               41.91                 0.9139
+    6      582               50.30                 0.9204
+    7      679               58.68                 0.9270
+    8      776               67.06                 0.9327
+    9      873               75.44                 0.9358
+   10      970               83.83                 0.9366
+
+31 of 38 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9411078947368419 +- 0.0013748457480841016
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1924
+    2      194               16.77                 0.7740
+    3      291               25.15                 0.8756
+    4      388               33.53                 0.8989
+    5      485               41.92                 0.9107
+    6      582               50.30                 0.9201
+    7      679               58.68                 0.9288
+    8      776               67.06                 0.9338
+    9      873               75.45                 0.9387
+   10      970               83.83                 0.9420
+
+32 of 39 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.941130769230769 +- 0.0013644109927005042
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2172
+    2      194               16.77                 0.7571
+    3      291               25.15                 0.8762
+    4      388               33.53                 0.8985
+    5      485               41.92                 0.9115
+    6      582               50.30                 0.9201
+    7      679               58.68                 0.9283
+    8      776               67.06                 0.9348
+    9      873               75.44                 0.9379
+   10      970               83.83                 0.9404
+
+33 of 40 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9411124999999997 +- 0.0013520701719955232
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1521
+    2      194               16.77                 0.7526
+    3      291               25.15                 0.8716
+    4      388               33.53                 0.8968
+    5      485               41.91                 0.9122
+    6      582               50.30                 0.9215
+    7      679               58.68                 0.9288
+    8      776               67.06                 0.9325
+    9      873               75.44                 0.9363
+   10      970               83.83                 0.9385
+
+33 of 41 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9410487804878045 +- 0.0013949599087200878
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1304
+    2      194               16.77                 0.7688
+    3      291               25.15                 0.8754
+    4      388               33.53                 0.8967
+    5      485               41.91                 0.9119
+    6      582               50.29                 0.9200
+    7      679               58.68                 0.9259
+    8      776               67.06                 0.9322
+    9      873               75.44                 0.9380
+   10      970               83.82                 0.9414
+
+34 of 42 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9410571428571425 +- 0.0013792929274058666
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1666
+    2      194               16.77                 0.7524
+    3      291               25.15                 0.8704
+    4      388               33.53                 0.8974
+    5      485               41.91                 0.9109
+    6      582               50.30                 0.9216
+    7      679               58.68                 0.9293
+    8      776               67.06                 0.9339
+    9      873               75.44                 0.9377
+   10      970               83.83                 0.9401
+
+35 of 43 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9410348837209299 +- 0.001370771930028334
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1364
+    2      194               16.77                 0.7617
+    3      291               25.15                 0.8733
+    4      388               33.53                 0.8999
+    5      485               41.91                 0.9102
+    6      582               50.30                 0.9209
+    7      679               58.68                 0.9266
+    8      776               67.06                 0.9318
+    9      873               75.44                 0.9358
+   10      970               83.83                 0.9383
+
+35 of 44 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.940972727272727 +- 0.0014150752688563725
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1609
+    2      194               16.77                 0.7789
+    3      291               25.15                 0.8798
+    4      388               33.53                 0.8992
+    5      485               41.91                 0.9109
+    6      582               50.29                 0.9206
+    7      679               58.67                 0.9276
+    8      776               67.06                 0.9332
+    9      873               75.44                 0.9374
+   10      970               83.82                 0.9400
+
+36 of 45 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409511111111107 +- 0.001406591186319645
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1521
+    2      194               16.77                 0.7581
+    3      291               25.15                 0.8720
+    4      388               33.53                 0.8964
+    5      485               41.91                 0.9126
+    6      582               50.29                 0.9198
+    7      679               58.67                 0.9292
+    8      776               67.06                 0.9351
+    9      873               75.44                 0.9380
+   10      970               83.82                 0.9406
+
+37 of 46 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409434782608692 +- 0.0013921600629352097
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1743
+    2      194               16.77                 0.7826
+    3      291               25.15                 0.8786
+    4      388               33.53                 0.9006
+    5      485               41.91                 0.9125
+    6      582               50.29                 0.9208
+    7      679               58.67                 0.9284
+    8      776               67.06                 0.9344
+    9      873               75.44                 0.9386
+   10      970               83.82                 0.9402
+
+38 of 47 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409276595744678 +- 0.0013814426821661717
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1795
+    2      194               16.77                 0.7730
+    3      291               25.15                 0.8805
+    4      388               33.53                 0.9000
+    5      485               41.91                 0.9122
+    6      582               50.29                 0.9224
+    7      679               58.67                 0.9272
+    8      776               67.06                 0.9333
+    9      873               75.44                 0.9360
+   10      970               83.82                 0.9380
+
+38 of 48 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408666666666664 +- 0.0014295006898292243
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1302
+    2      194               16.77                 0.7725
+    3      291               25.15                 0.8744
+    4      388               33.53                 0.9009
+    5      485               41.91                 0.9104
+    6      582               50.30                 0.9218
+    7      679               58.68                 0.9285
+    8      776               67.06                 0.9344
+    9      873               75.44                 0.9383
+   10      970               83.83                 0.9403
+
+39 of 49 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.940855102040816 +- 0.0014171055905943639
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2323
+    2      194               16.77                 0.7674
+    3      291               25.15                 0.8729
+    4      388               33.53                 0.8974
+    5      485               41.91                 0.9120
+    6      582               50.29                 0.9206
+    7      679               58.67                 0.9291
+    8      776               67.06                 0.9330
+    9      873               75.44                 0.9385
+   10      970               83.82                 0.9413
+
+40 of 50 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408639999999997 +- 0.001404244992869836
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2032
+    2      194               16.77                 0.7778
+    3      291               25.15                 0.8746
+    4      388               33.53                 0.9015
+    5      485               41.91                 0.9117
+    6      582               50.29                 0.9218
+    7      679               58.67                 0.9295
+    8      776               67.06                 0.9343
+    9      873               75.44                 0.9384
+   10      970               83.82                 0.9411
+
+41 of 51 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408686274509801 +- 0.00139079469398437
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2075
+    2      194               16.77                 0.7787
+    3      291               25.15                 0.8763
+    4      388               33.53                 0.8991
+    5      485               41.91                 0.9118
+    6      582               50.29                 0.9220
+    7      679               58.67                 0.9287
+    8      776               67.06                 0.9336
+    9      873               75.44                 0.9393
+   10      970               83.82                 0.9416
+
+42 of 52 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408826923076921 +- 0.0013810142808048684
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1834
+    2      194               16.77                 0.7662
+    3      291               25.15                 0.8784
+    4      388               33.53                 0.8978
+    5      485               41.91                 0.9120
+    6      582               50.29                 0.9216
+    7      679               58.67                 0.9273
+    8      776               67.06                 0.9351
+    9      873               75.44                 0.9386
+   10      970               83.82                 0.9418
+
+43 of 53 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408999999999997 +- 0.0013736056394868924
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1876
+    2      194               16.77                 0.7738
+    3      291               25.15                 0.8746
+    4      388               33.53                 0.9011
+    5      485               41.91                 0.9116
+    6      582               50.29                 0.9196
+    7      679               58.67                 0.9277
+    8      776               67.06                 0.9349
+    9      873               75.44                 0.9384
+   10      970               83.82                 0.9415
+
+44 of 54 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409111111111108 +- 0.0013632296437975523
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1535
+    2      194               16.77                 0.7717
+    3      291               25.15                 0.8765
+    4      388               33.53                 0.9010
+    5      485               41.91                 0.9125
+    6      582               50.30                 0.9234
+    7      679               58.68                 0.9277
+    8      776               67.06                 0.9347
+    9      873               75.44                 0.9389
+   10      970               83.82                 0.9417
+
+45 of 55 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409254545454542 +- 0.0013548858632664769
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1505
+    2      194               16.77                 0.7702
+    3      291               25.15                 0.8777
+    4      388               33.53                 0.8981
+    5      485               41.91                 0.9105
+    6      582               50.30                 0.9200
+    7      679               58.68                 0.9284
+    8      776               67.06                 0.9354
+    9      873               75.44                 0.9385
+   10      970               83.82                 0.9403
+
+46 of 56 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409142857142854 +- 0.0013452865562278372
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1656
+    2      194               16.77                 0.7657
+    3      291               25.15                 0.8764
+    4      388               33.53                 0.9018
+    5      485               41.91                 0.9114
+    6      582               50.29                 0.9185
+    7      679               58.68                 0.9267
+    8      776               67.06                 0.9340
+    9      873               75.44                 0.9377
+   10      970               83.82                 0.9411
+
+47 of 57 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409175438596488 +- 0.0013336564705386735
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1637
+    2      194               16.77                 0.7757
+    3      291               25.15                 0.8769
+    4      388               33.53                 0.9035
+    5      485               41.91                 0.9120
+    6      582               50.30                 0.9191
+    7      679               58.68                 0.9278
+    8      776               67.06                 0.9334
+    9      873               75.44                 0.9367
+   10      970               83.83                 0.9422
+
+48 of 58 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409396551724134 +- 0.001332606941451337
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2329
+    2      194               16.77                 0.7731
+    3      291               25.15                 0.8785
+    4      388               33.53                 0.8970
+    5      485               41.91                 0.9107
+    6      582               50.30                 0.9187
+    7      679               58.68                 0.9280
+    8      776               67.06                 0.9348
+    9      873               75.44                 0.9403
+   10      970               83.83                 0.9429
+
+49 of 59 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409728813559318 +- 0.0013452780581427036
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2039
+    2      194               16.77                 0.7801
+    3      291               25.15                 0.8770
+    4      388               33.53                 0.8986
+    5      485               41.91                 0.9103
+    6      582               50.29                 0.9190
+    7      679               58.68                 0.9253
+    8      776               67.06                 0.9344
+    9      873               75.44                 0.9371
+   10      970               83.82                 0.9402
+
+50 of 60 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409599999999997 +- 0.0013376845667047253
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2142
+    2      194               16.77                 0.7742
+    3      291               25.15                 0.8754
+    4      388               33.53                 0.8977
+    5      485               41.91                 0.9109
+    6      582               50.30                 0.9202
+    7      679               58.68                 0.9287
+    8      776               67.06                 0.9338
+    9      873               75.44                 0.9381
+   10      970               83.82                 0.9401
+
+51 of 61 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409459016393439 +- 0.0013311616731867151
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1682
+    2      194               16.77                 0.7908
+    3      291               25.15                 0.8779
+    4      388               33.53                 0.8970
+    5      485               41.91                 0.9098
+    6      582               50.30                 0.9193
+    7      679               58.68                 0.9262
+    8      776               67.06                 0.9332
+    9      873               75.44                 0.9382
+   10      970               83.83                 0.9405
+
+52 of 62 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.940938709677419 +- 0.0013215771173121645
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2225
+    2      194               16.77                 0.7677
+    3      291               25.15                 0.8717
+    4      388               33.53                 0.8947
+    5      485               41.91                 0.9096
+    6      582               50.29                 0.9196
+    7      679               58.68                 0.9278
+    8      776               67.06                 0.9315
+    9      873               75.44                 0.9352
+   10      970               83.82                 0.9388
+
+52 of 63 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409047619047616 +- 0.0013380189776608198
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1942
+    2      194               16.77                 0.7633
+    3      291               25.15                 0.8802
+    4      388               33.53                 0.9019
+    5      485               41.91                 0.9142
+    6      582               50.29                 0.9212
+    7      679               58.68                 0.9282
+    8      776               67.06                 0.9341
+    9      873               75.44                 0.9392
+   10      970               83.82                 0.9405
+
+53 of 64 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408984374999997 +- 0.0013284732999175243
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1798
+    2      194               16.77                 0.7404
+    3      291               25.15                 0.8756
+    4      388               33.53                 0.8997
+    5      485               41.91                 0.9122
+    6      582               50.29                 0.9220
+    7      679               58.68                 0.9274
+    8      776               67.06                 0.9345
+    9      873               75.44                 0.9383
+   10      970               83.83                 0.9398
+
+53 of 65 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408815384615381 +- 0.001325128998967772
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1694
+    2      194               16.77                 0.7854
+    3      291               25.15                 0.8782
+    4      388               33.53                 0.8982
+    5      485               41.91                 0.9113
+    6      582               50.29                 0.9212
+    7      679               58.67                 0.9261
+    8      776               67.06                 0.9350
+    9      873               75.44                 0.9390
+   10      970               83.82                 0.9389
+
+53 of 66 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408515151515148 +- 0.0013371433168402089
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1861
+    2      194               16.77                 0.7683
+    3      291               25.15                 0.8805
+    4      388               33.53                 0.9002
+    5      485               41.91                 0.9124
+    6      582               50.29                 0.9206
+    7      679               58.68                 0.9284
+    8      776               67.06                 0.9336
+    9      873               75.44                 0.9393
+   10      970               83.82                 0.9418
+
+54 of 67 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408656716417907 +- 0.0013321010493253605
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1867
+    2      194               16.77                 0.7632
+    3      291               25.15                 0.8763
+    4      388               33.53                 0.8988
+    5      485               41.91                 0.9131
+    6      582               50.29                 0.9209
+    7      679               58.67                 0.9279
+    8      776               67.06                 0.9331
+    9      873               75.44                 0.9381
+   10      970               83.82                 0.9404
+
+55 of 68 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408588235294113 +- 0.0013234575143870496
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1635
+    2      194               16.77                 0.7892
+    3      291               25.15                 0.8791
+    4      388               33.53                 0.8993
+    5      485               41.91                 0.9117
+    6      582               50.29                 0.9209
+    7      679               58.68                 0.9276
+    8      776               67.06                 0.9363
+    9      873               75.44                 0.9387
+   10      970               83.82                 0.9427
+
+56 of 69 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408855072463763 +- 0.0013321308377138024
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1834
+    2      194               16.77                 0.7837
+    3      291               25.15                 0.8775
+    4      388               33.53                 0.8984
+    5      485               41.91                 0.9134
+    6      582               50.29                 0.9219
+    7      679               58.67                 0.9288
+    8      776               67.06                 0.9330
+    9      873               75.44                 0.9364
+   10      970               83.82                 0.9389
+
+56 of 70 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408571428571424 +- 0.0013434041439111153
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1898
+    2      194               16.77                 0.7501
+    3      291               25.15                 0.8768
+    4      388               33.53                 0.9008
+    5      485               41.91                 0.9119
+    6      582               50.29                 0.9205
+    7      679               58.68                 0.9258
+    8      776               67.06                 0.9325
+    9      873               75.44                 0.9382
+   10      970               83.82                 0.9402
+
+57 of 71 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408478873239433 +- 0.0013361558434585158
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2717
+    2      194               16.77                 0.7899
+    3      291               25.15                 0.8754
+    4      388               33.53                 0.8981
+    5      485               41.91                 0.9138
+    6      582               50.30                 0.9220
+    7      679               58.68                 0.9278
+    8      776               67.06                 0.9339
+    9      873               75.44                 0.9384
+   10      970               83.82                 0.9414
+
+58 of 72 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408555555555552 +- 0.0013284168615403436
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2068
+    2      194               16.77                 0.7802
+    3      291               25.15                 0.8781
+    4      388               33.53                 0.8996
+    5      485               41.91                 0.9140
+    6      582               50.30                 0.9216
+    7      679               58.68                 0.9280
+    8      776               67.06                 0.9335
+    9      873               75.44                 0.9387
+   10      970               83.83                 0.9428
+
+59 of 73 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408821917808217 +- 0.0013385068828918626
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1775
+    2      194               16.77                 0.7809
+    3      291               25.15                 0.8761
+    4      388               33.53                 0.8968
+    5      485               41.91                 0.9111
+    6      582               50.29                 0.9209
+    7      679               58.67                 0.9280
+    8      776               67.06                 0.9331
+    9      873               75.44                 0.9385
+   10      970               83.82                 0.9424
+
+60 of 74 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9409027027027026 +- 0.0013409328109589289
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1966
+    2      194               16.77                 0.7746
+    3      291               25.15                 0.8780
+    4      388               33.53                 0.8998
+    5      485               41.91                 0.9114
+    6      582               50.30                 0.9220
+    7      679               58.68                 0.9281
+    8      776               67.06                 0.9344
+    9      873               75.44                 0.9375
+   10      970               83.83                 0.9400
+
+61 of 75 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408906666666664 +- 0.0013359813704622633
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1727
+    2      194               16.77                 0.7745
+    3      291               25.15                 0.8747
+    4      388               33.53                 0.9003
+    5      485               41.91                 0.9128
+    6      582               50.29                 0.9226
+    7      679               58.67                 0.9278
+    8      776               67.06                 0.9336
+    9      873               75.44                 0.9369
+   10      970               83.82                 0.9407
+
+62 of 76 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9434
+Mean accuracy: 0.9408881578947368 +- 0.0013273407432733571
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1865
+    2      194               16.77                 0.7655
+    3      291               25.15                 0.8717
+    4      388               33.53                 0.8978
+    5      485               41.91                 0.9140
+    6      582               50.30                 0.9232
+    7      679               58.68                 0.9304
+    8      776               67.06                 0.9364
+    9      873               75.44                 0.9405
+   10      970               83.82                 0.9435
+
+63 of 77 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9409220779220778 +- 0.001351442075059024
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2233
+    2      194               16.77                 0.7653
+    3      291               25.15                 0.8776
+    4      388               33.53                 0.9003
+    5      485               41.91                 0.9148
+    6      582               50.30                 0.9229
+    7      679               58.68                 0.9290
+    8      776               67.06                 0.9338
+    9      873               75.44                 0.9377
+   10      970               83.82                 0.9398
+
+63 of 78 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9409076923076923 +- 0.0013486716380856857
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1690
+    2      194               16.77                 0.7789
+    3      291               25.15                 0.8756
+    4      388               33.53                 0.8999
+    5      485               41.91                 0.9114
+    6      582               50.30                 0.9200
+    7      679               58.68                 0.9282
+    8      776               67.06                 0.9351
+    9      873               75.44                 0.9386
+   10      970               83.82                 0.9411
+
+64 of 79 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9409101265822786 +- 0.0013402809961101032
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1467
+    2      194               16.77                 0.7810
+    3      291               25.15                 0.8784
+    4      388               33.53                 0.8990
+    5      485               41.91                 0.9114
+    6      582               50.30                 0.9204
+    7      679               58.68                 0.9272
+    8      776               67.06                 0.9346
+    9      873               75.44                 0.9383
+   10      970               83.83                 0.9406
+
+65 of 80 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.94090625 +- 0.0013323235108261063
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1904
+    2      194               16.77                 0.7659
+    3      291               25.15                 0.8766
+    4      388               33.53                 0.9017
+    5      485               41.91                 0.9155
+    6      582               50.29                 0.9233
+    7      679               58.68                 0.9298
+    8      776               67.06                 0.9355
+    9      873               75.44                 0.9393
+   10      970               83.83                 0.9418
+
+66 of 81 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9409172839506174 +- 0.0013277466405885844
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2094
+    2      194               16.77                 0.7563
+    3      291               25.15                 0.8759
+    4      388               33.53                 0.8991
+    5      485               41.91                 0.9136
+    6      582               50.29                 0.9182
+    7      679               58.67                 0.9278
+    8      776               67.06                 0.9332
+    9      873               75.44                 0.9372
+   10      970               83.82                 0.9405
+
+67 of 82 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9409121951219513 +- 0.0013204203177723237
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1654
+    2      194               16.77                 0.7718
+    3      291               25.15                 0.8765
+    4      388               33.53                 0.8998
+    5      485               41.91                 0.9105
+    6      582               50.29                 0.9196
+    7      679               58.68                 0.9261
+    8      776               67.06                 0.9321
+    9      873               75.44                 0.9361
+   10      970               83.83                 0.9378
+
+67 of 83 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9408746987951808 +- 0.0013556524569278267
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1831
+    2      194               16.77                 0.7600
+    3      291               25.15                 0.8758
+    4      388               33.53                 0.9003
+    5      485               41.91                 0.9126
+    6      582               50.30                 0.9209
+    7      679               58.68                 0.9286
+    8      776               67.06                 0.9345
+    9      873               75.44                 0.9400
+   10      970               83.83                 0.9416
+
+68 of 84 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9435
+Mean accuracy: 0.9408833333333333 +- 0.0013498530196296094
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1735
+    2      194               16.77                 0.7721
+    3      291               25.15                 0.8775
+    4      388               33.53                 0.9012
+    5      485               41.91                 0.9110
+    6      582               50.30                 0.9210
+    7      679               58.68                 0.9287
+    8      776               67.06                 0.9360
+    9      873               75.44                 0.9410
+   10      970               83.82                 0.9436
+
+69 of 85 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409152941176471 +- 0.0013734889778656161
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2085
+    2      194               16.77                 0.7808
+    3      291               25.15                 0.8788
+    4      388               33.53                 0.9009
+    5      485               41.91                 0.9117
+    6      582               50.29                 0.9229
+    7      679               58.68                 0.9280
+    8      776               67.06                 0.9341
+    9      873               75.44                 0.9383
+   10      970               83.82                 0.9418
+
+70 of 86 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409255813953489 +- 0.001368770119020587
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2021
+    2      194               16.77                 0.7647
+    3      291               25.15                 0.8786
+    4      388               33.53                 0.8994
+    5      485               41.91                 0.9118
+    6      582               50.29                 0.9191
+    7      679               58.68                 0.9276
+    8      776               67.06                 0.9348
+    9      873               75.44                 0.9390
+   10      970               83.82                 0.9415
+
+71 of 87 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409321839080461 +- 0.0013622576133970725
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1871
+    2      194               16.77                 0.7743
+    3      291               25.15                 0.8779
+    4      388               33.53                 0.9008
+    5      485               41.91                 0.9143
+    6      582               50.29                 0.9210
+    7      679               58.68                 0.9317
+    8      776               67.06                 0.9375
+    9      873               75.44                 0.9412
+   10      970               83.82                 0.9422
+
+72 of 88 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409465909090909 +- 0.0013611449778867949
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1602
+    2      194               16.77                 0.7761
+    3      291               25.15                 0.8784
+    4      388               33.53                 0.8988
+    5      485               41.91                 0.9115
+    6      582               50.29                 0.9213
+    7      679               58.67                 0.9277
+    8      776               67.06                 0.9336
+    9      873               75.44                 0.9376
+   10      970               83.82                 0.9413
+
+73 of 89 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409505617977528 +- 0.001353988995980844
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1528
+    2      194               16.77                 0.7758
+    3      291               25.15                 0.8769
+    4      388               33.53                 0.9015
+    5      485               41.91                 0.9150
+    6      582               50.30                 0.9225
+    7      679               58.68                 0.9279
+    8      776               67.06                 0.9353
+    9      873               75.45                 0.9391
+   10      970               83.83                 0.9410
+
+74 of 90 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409511111111112 +- 0.0013464557957380333
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1469
+    2      194               16.77                 0.7698
+    3      291               25.15                 0.8761
+    4      388               33.53                 0.8977
+    5      485               41.91                 0.9116
+    6      582               50.30                 0.9201
+    7      679               58.68                 0.9282
+    8      776               67.06                 0.9343
+    9      873               75.45                 0.9381
+   10      970               83.83                 0.9412
+
+75 of 91 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409538461538461 +- 0.0013392886165085479
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1856
+    2      194               16.77                 0.7583
+    3      291               25.15                 0.8759
+    4      388               33.53                 0.9009
+    5      485               41.91                 0.9100
+    6      582               50.29                 0.9209
+    7      679               58.67                 0.9257
+    8      776               67.06                 0.9337
+    9      873               75.44                 0.9380
+   10      970               83.82                 0.9426
+
+76 of 92 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409717391304347 +- 0.001342881885362232
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1997
+    2      194               16.77                 0.7691
+    3      291               25.15                 0.8740
+    4      388               33.53                 0.8997
+    5      485               41.91                 0.9119
+    6      582               50.29                 0.9213
+    7      679               58.68                 0.9298
+    8      776               67.06                 0.9357
+    9      873               75.44                 0.9395
+   10      970               83.82                 0.9419
+
+77 of 93 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409817204301076 +- 0.0013390693475468685
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1701
+    2      194               16.77                 0.7776
+    3      291               25.15                 0.8770
+    4      388               33.53                 0.9011
+    5      485               41.91                 0.9122
+    6      582               50.29                 0.9210
+    7      679               58.67                 0.9281
+    8      776               67.06                 0.9331
+    9      873               75.44                 0.9377
+   10      970               83.82                 0.9402
+
+78 of 94 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9436
+Mean accuracy: 0.9409734042553193 +- 0.001334339866424902
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2069
+    2      194               16.77                 0.7783
+    3      291               25.15                 0.8775
+    4      388               33.53                 0.9025
+    5      485               41.91                 0.9138
+    6      582               50.30                 0.9222
+    7      679               58.68                 0.9309
+    8      776               67.06                 0.9368
+    9      873               75.44                 0.9413
+   10      970               83.83                 0.9444
+
+79 of 95 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9444
+Mean accuracy: 0.9410094736842106 +- 0.001372594445736019
+
+Files already downloaded and verified
+Preprocessing: 1.31 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1579
+    2      194               16.77                 0.7725
+    3      291               25.15                 0.8786
+    4      388               33.53                 0.8998
+    5      485               41.91                 0.9115
+    6      582               50.29                 0.9201
+    7      679               58.68                 0.9275
+    8      776               67.06                 0.9333
+    9      873               75.44                 0.9366
+   10      970               83.83                 0.9401
+
+80 of 96 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9444
+Mean accuracy: 0.9410000000000002 +- 0.001368545456558408
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.2063
+    2      194               16.77                 0.7808
+    3      291               25.15                 0.8783
+    4      388               33.53                 0.8991
+    5      485               41.91                 0.9125
+    6      582               50.30                 0.9227
+    7      679               58.68                 0.9253
+    8      776               67.06                 0.9315
+    9      873               75.44                 0.9353
+   10      970               83.83                 0.9386
+
+80 of 97 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9444
+Mean accuracy: 0.9409752577319588 +- 0.001382887381855016
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1519
+    2      194               16.77                 0.7667
+    3      291               25.15                 0.8728
+    4      388               33.53                 0.9021
+    5      485               41.91                 0.9114
+    6      582               50.30                 0.9203
+    7      679               58.68                 0.9277
+    8      776               67.06                 0.9333
+    9      873               75.44                 0.9405
+   10      970               83.83                 0.9427
+
+81 of 98 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9444
+Mean accuracy: 0.9409928571428572 +- 0.0013866896420126984
+
+Files already downloaded and verified
+Preprocessing: 1.32 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1389
+    2      194               16.77                 0.7433
+    3      291               25.15                 0.8699
+    4      388               33.53                 0.8942
+    5      485               41.91                 0.9110
+    6      582               50.29                 0.9202
+    7      679               58.68                 0.9262
+    8      776               67.06                 0.9320
+    9      873               75.44                 0.9358
+   10      970               83.82                 0.9402
+
+82 of 99 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9444
+Mean accuracy: 0.9409848484848486 +- 0.0013819444386763934
+
+Files already downloaded and verified
+Preprocessing: 1.33 seconds
+
+epoch    batch    train time [sec]    validation accuracy
+    1       97                8.38                 0.1503
+    2      194               16.77                 0.7679
+    3      291               25.15                 0.8743
+    4      388               33.53                 0.8981
+    5      485               41.91                 0.9114
+    6      582               50.30                 0.9235
+    7      679               58.68                 0.9301
+    8      776               67.06                 0.9363
+    9      873               75.44                 0.9392
+   10      970               83.83                 0.9443
+
+83 of 100 runs >= 94.0 % accuracy
+Min  accuracy: 0.9366000000000001
+Max  accuracy: 0.9444
+Mean accuracy: 0.9410180000000001 +- 0.0014140282882601739
+
diff --git a/cifar10-fast-simple/model.py b/cifar10-fast-simple/model.py
new file mode 100644
index 0000000..fe33495
--- /dev/null
+++ b/cifar10-fast-simple/model.py
@@ -0,0 +1,141 @@
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+
+
+def label_smoothing_loss(inputs, targets, alpha):
+    log_probs = torch.nn.functional.log_softmax(inputs, dim=1, _stacklevel=5)
+    kl = -log_probs.mean(dim=1)
+    xent = torch.nn.functional.nll_loss(log_probs, targets, reduction="none")
+    loss = (1 - alpha) * xent + alpha * kl
+    return loss
+
+
+class GhostBatchNorm(nn.BatchNorm2d):
+    def __init__(self, num_features, num_splits, **kw):
+        super().__init__(num_features, **kw)
+
+        running_mean = torch.zeros(num_features * num_splits)
+        running_var = torch.ones(num_features * num_splits)
+
+        self.weight.requires_grad = False
+        self.num_splits = num_splits
+        self.register_buffer("running_mean", running_mean)
+        self.register_buffer("running_var", running_var)
+
+    def train(self, mode=True):
+        if (self.training is True) and (mode is False):
+            # lazily collate stats when we are going to use them
+            self.running_mean = torch.mean(
+                self.running_mean.view(self.num_splits, self.num_features), dim=0
+            ).repeat(self.num_splits)
+            self.running_var = torch.mean(
+                self.running_var.view(self.num_splits, self.num_features), dim=0
+            ).repeat(self.num_splits)
+        return super().train(mode)
+
+    def forward(self, input):
+        n, c, h, w = input.shape
+        if self.training or not self.track_running_stats:
+            assert n % self.num_splits == 0, f"Batch size ({n}) must be divisible by num_splits ({self.num_splits}) of GhostBatchNorm"
+            return F.batch_norm(
+                input.view(-1, c * self.num_splits, h, w),
+                self.running_mean,
+                self.running_var,
+                self.weight.repeat(self.num_splits),
+                self.bias.repeat(self.num_splits),
+                True,
+                self.momentum,
+                self.eps,
+            ).view(n, c, h, w)
+        else:
+            return F.batch_norm(
+                input,
+                self.running_mean[: self.num_features],
+                self.running_var[: self.num_features],
+                self.weight,
+                self.bias,
+                False,
+                self.momentum,
+                self.eps,
+            )
+
+
+def conv_bn_relu(c_in, c_out, kernel_size=(3, 3), padding=(1, 1)):
+    return nn.Sequential(
+        nn.Conv2d(c_in, c_out, kernel_size=kernel_size, padding=padding, bias=False),
+        GhostBatchNorm(c_out, num_splits=16),
+        nn.CELU(alpha=0.3),
+    )
+
+
+def conv_pool_norm_act(c_in, c_out):
+    return nn.Sequential(
+        nn.Conv2d(c_in, c_out, kernel_size=(3, 3), padding=(1, 1), bias=False),
+        nn.MaxPool2d(kernel_size=2, stride=2),
+        GhostBatchNorm(c_out, num_splits=16),
+        nn.CELU(alpha=0.3),
+    )
+
+
+def patch_whitening(data, patch_size=(3, 3)):
+    # Compute weights from data such that
+    # torch.std(F.conv2d(data, weights), dim=(2, 3))
+    # is close to 1.
+    h, w = patch_size
+    c = data.size(1)
+    patches = data.unfold(2, h, 1).unfold(3, w, 1)
+    patches = patches.transpose(1, 3).reshape(-1, c, h, w).to(torch.float32)
+
+    n, c, h, w = patches.shape
+    X = patches.reshape(n, c * h * w)
+    X = X / (X.size(0) - 1) ** 0.5
+    covariance = X.t() @ X
+
+    eigenvalues, eigenvectors = torch.linalg.eigh(covariance)
+
+    eigenvalues = eigenvalues.flip(0)
+
+    eigenvectors = eigenvectors.t().reshape(c * h * w, c, h, w).flip(0)
+
+    return eigenvectors / torch.sqrt(eigenvalues + 1e-2).view(-1, 1, 1, 1)
+
+
+class ResNetBagOfTricks(nn.Module):
+    def __init__(self, first_layer_weights, c_in, c_out, scale_out):
+        super().__init__()
+
+        c = first_layer_weights.size(0)
+
+        conv1 = nn.Conv2d(c_in, c, kernel_size=(3, 3), padding=(1, 1), bias=False)
+        conv1.weight.data = first_layer_weights
+        conv1.weight.requires_grad = False
+
+        self.conv1 = conv1
+        self.conv2 = conv_bn_relu(c, 64, kernel_size=(1, 1), padding=0)
+        self.conv3 = conv_pool_norm_act(64, 128)
+        self.conv4 = conv_bn_relu(128, 128)
+        self.conv5 = conv_bn_relu(128, 128)
+        self.conv6 = conv_pool_norm_act(128, 256)
+        self.conv7 = conv_pool_norm_act(256, 512)
+        self.conv8 = conv_bn_relu(512, 512)
+        self.conv9 = conv_bn_relu(512, 512)
+        self.pool10 = nn.MaxPool2d(kernel_size=4, stride=4)
+        self.linear11 = nn.Linear(512, c_out, bias=False)
+        self.scale_out = scale_out
+
+    def forward(self, x):
+        x = self.conv1(x)
+        x = self.conv2(x)
+        x = self.conv3(x)
+        x = x + self.conv5(self.conv4(x))
+        x = self.conv6(x)
+        x = self.conv7(x)
+        x = x + self.conv9(self.conv8(x))
+        x = self.pool10(x)
+        x = x.reshape(x.size(0), x.size(1))
+        x = self.linear11(x)
+        x = self.scale_out * x
+        return x
+
+Model = ResNetBagOfTricks
diff --git a/cifar10-fast-simple/train.py b/cifar10-fast-simple/train.py
new file mode 100644
index 0000000..0dba02a
--- /dev/null
+++ b/cifar10-fast-simple/train.py
@@ -0,0 +1,278 @@
+import time
+import copy
+import torch
+import torch.nn as nn
+import torchvision
+import model
+
+
+def train(seed=0):
+    # Configurable parameters
+    epochs = 10
+    batch_size = 512
+    momentum = 0.9
+    weight_decay = 0.256
+    weight_decay_bias = 0.004
+    ema_update_freq = 5
+    ema_rho = 0.99 ** ema_update_freq
+    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+    dtype = torch.float16 if device.type != "cpu" else torch.float32
+
+    # First, the learning rate rises from 0 to 0.002 for the first 194 batches.
+    # Next, the learning rate shrinks down to 0.0002 over the next 582 batches.
+    lr_schedule = torch.cat([
+        torch.linspace(0e+0, 2e-3, 194),
+        torch.linspace(2e-3, 2e-4, 582),
+    ])
+
+    lr_schedule_bias = 64.0 * lr_schedule
+
+    # Print information about hardware on first run
+    if seed == 0:
+        if device.type == "cuda":
+            print("Device :", torch.cuda.get_device_name(device.index))
+
+        print("Dtype  :", dtype)
+        print()
+
+    # Start measuring time
+    start_time = time.perf_counter()
+
+    # Set random seed to increase chance of reproducability
+    torch.manual_seed(seed)
+
+    # Setting cudnn.benchmark to True hampers reproducability, but is faster
+    torch.backends.cudnn.benchmark = True
+
+    # Load dataset
+    train_data, train_targets, valid_data, valid_targets = load_cifar10(device, dtype)
+
+    # Compute special weights for first layer
+    weights = model.patch_whitening(train_data[:10000, :, 4:-4, 4:-4])
+
+    # Construct the neural network
+    train_model = model.Model(weights, c_in=3, c_out=10, scale_out=0.125)
+
+    # Convert model weights to half precision
+    train_model.to(dtype)
+
+    # Convert BatchNorm back to single precision for better accuracy
+    for module in train_model.modules():
+        if isinstance(module, nn.BatchNorm2d):
+            module.float()
+
+    # Upload model to GPU
+    train_model.to(device)
+
+    # Collect weights and biases and create nesterov velocity values
+    weights = [
+        (w, torch.zeros_like(w))
+        for w in train_model.parameters()
+        if w.requires_grad and len(w.shape) > 1
+    ]
+    biases = [
+        (w, torch.zeros_like(w))
+        for w in train_model.parameters()
+        if w.requires_grad and len(w.shape) <= 1
+    ]
+
+    # Copy the model for validation
+    valid_model = copy.deepcopy(train_model)
+
+    print(f"Preprocessing: {time.perf_counter() - start_time:.2f} seconds")
+
+    # Train and validate
+    print("\nepoch    batch    train time [sec]    validation accuracy")
+    train_time = 0.0
+    batch_count = 0
+    for epoch in range(1, epochs + 1):
+        # Flush CUDA pipeline for more accurate time measurement
+        if torch.cuda.is_available():
+            torch.cuda.synchronize()
+
+        start_time = time.perf_counter()
+
+        # Randomly shuffle training data
+        indices = torch.randperm(len(train_data), device=device)
+        data = train_data[indices]
+        targets = train_targets[indices]
+
+        # Crop random 32x32 patches from 40x40 training data
+        data = [
+            random_crop(data[i : i + batch_size], crop_size=(32, 32))
+            for i in range(0, len(data), batch_size)
+        ]
+        data = torch.cat(data)
+
+        # Randomly flip half the training data
+        data[: len(data) // 2] = torch.flip(data[: len(data) // 2], [-1])
+
+        for i in range(0, len(data), batch_size):
+            # discard partial batches
+            if i + batch_size > len(data):
+                break
+
+            # Slice batch from data
+            inputs = data[i : i + batch_size]
+            target = targets[i : i + batch_size]
+            batch_count += 1
+
+            # Compute new gradients
+            train_model.zero_grad()
+            train_model.train(True)
+
+            logits = train_model(inputs)
+
+            loss = model.label_smoothing_loss(logits, target, alpha=0.2)
+
+            loss.sum().backward()
+
+            lr_index = min(batch_count, len(lr_schedule) - 1)
+            lr = lr_schedule[lr_index]
+            lr_bias = lr_schedule_bias[lr_index]
+
+            # Update weights and biases of training model
+            update_nesterov(weights, lr, weight_decay, momentum)
+            update_nesterov(biases, lr_bias, weight_decay_bias, momentum)
+
+            # Update validation model with exponential moving averages
+            if (i // batch_size % ema_update_freq) == 0:
+                update_ema(train_model, valid_model, ema_rho)
+
+        if torch.cuda.is_available():
+            torch.cuda.synchronize()
+
+        # Add training time
+        train_time += time.perf_counter() - start_time
+
+        valid_correct = []
+        for i in range(0, len(valid_data), batch_size):
+            valid_model.train(False)
+
+            # Test time agumentation: Test model on regular and flipped data
+            regular_inputs = valid_data[i : i + batch_size]
+            flipped_inputs = torch.flip(regular_inputs, [-1])
+
+            logits1 = valid_model(regular_inputs).detach()
+            logits2 = valid_model(flipped_inputs).detach()
+
+            # Final logits are average of augmented logits
+            logits = torch.mean(torch.stack([logits1, logits2], dim=0), dim=0)
+
+            # Compute correct predictions
+            correct = logits.max(dim=1)[1] == valid_targets[i : i + batch_size]
+
+            valid_correct.append(correct.detach().type(torch.float64))
+
+        # Accuracy is average number of correct predictions
+        valid_acc = torch.mean(torch.cat(valid_correct)).item()
+
+        print(f"{epoch:5} {batch_count:8d} {train_time:19.2f} {valid_acc:22.4f}")
+
+    return valid_acc
+
+def preprocess_data(data, device, dtype):
+    # Convert to torch float16 tensor
+    data = torch.tensor(data, device=device).to(dtype)
+
+    # Normalize
+    mean = torch.tensor([125.31, 122.95, 113.87], device=device).to(dtype)
+    std = torch.tensor([62.99, 62.09, 66.70], device=device).to(dtype)
+    data = (data - mean) / std
+
+    # Permute data from NHWC to NCHW format
+    data = data.permute(0, 3, 1, 2)
+
+    return data
+
+
+def load_cifar10(device, dtype, data_dir="~/data"):
+    train = torchvision.datasets.CIFAR10(root=data_dir, download=True)
+    valid = torchvision.datasets.CIFAR10(root=data_dir, train=False)
+
+    train_data = preprocess_data(train.data, device, dtype)
+    valid_data = preprocess_data(valid.data, device, dtype)
+
+    train_targets = torch.tensor(train.targets).to(device)
+    valid_targets = torch.tensor(valid.targets).to(device)
+
+    # Pad 32x32 to 40x40
+    train_data = nn.ReflectionPad2d(4)(train_data)
+
+    return train_data, train_targets, valid_data, valid_targets
+
+
+def update_ema(train_model, valid_model, rho):
+    # The trained model is not used for validation directly. Instead, the
+    # validation model weights are updated with exponential moving averages.
+    train_weights = train_model.state_dict().values()
+    valid_weights = valid_model.state_dict().values()
+    for train_weight, valid_weight in zip(train_weights, valid_weights):
+        if valid_weight.dtype in [torch.float16, torch.float32]:
+            valid_weight *= rho
+            valid_weight += (1 - rho) * train_weight
+
+
+def update_nesterov(weights, lr, weight_decay, momentum):
+    for weight, velocity in weights:
+        if weight.requires_grad:
+            gradient = weight.grad.data
+            weight = weight.data
+
+            gradient.add_(weight, alpha=weight_decay).mul_(-lr)
+            velocity.mul_(momentum).add_(gradient)
+            weight.add_(gradient.add_(velocity, alpha=momentum))
+
+
+def random_crop(data, crop_size):
+    crop_h, crop_w = crop_size
+    h = data.size(2)
+    w = data.size(3)
+    x = torch.randint(w - crop_w, size=(1,))[0]
+    y = torch.randint(h - crop_h, size=(1,))[0]
+    return data[:, :, y : y + crop_h, x : x + crop_w]
+
+
+def sha256(path):
+    import hashlib
+    with open(path, "rb") as f:
+        return hashlib.sha256(f.read()).hexdigest()
+
+
+def getrelpath(abspath):
+    import os
+    return os.path.relpath(abspath, os.getcwd())
+
+
+def print_info():
+    # Knowing this information might improve chance of reproducability
+    print("File   :", getrelpath(__file__), sha256(__file__))
+    print("Model  :", getrelpath(model.__file__), sha256(model.__file__))
+    print("PyTorch:", torch.__version__)
+
+
+def main():
+    print_info()
+
+    accuracies = []
+    threshold = 0.94
+    for run in range(100):
+        valid_acc = train(seed=run)
+        accuracies.append(valid_acc)
+
+        # Print accumulated results
+        within_threshold = sum(acc >= threshold for acc in accuracies)
+        acc = threshold * 100.0
+        print()
+        print(f"{within_threshold} of {run + 1} runs >= {acc} % accuracy")
+        mean = sum(accuracies) / len(accuracies)
+        variance = sum((acc - mean)**2 for acc in accuracies) / len(accuracies)
+        std = variance**0.5
+        print(f"Min  accuracy: {min(accuracies)}")
+        print(f"Max  accuracy: {max(accuracies)}")
+        print(f"Mean accuracy: {mean} +- {std}")
+        print()
+
+
+if __name__ == "__main__":
+    main()