Added end-to-end test for CNTK speech workloads using AN4 dataset

2015-07-29 19:39:47 -07:00 · 2015-07-29 19:39:47 -07:00 · 746c2dbd1f
--- a/Tests/Speech/Data/AN4.LICENSE.html
+++ b/Tests/Speech/Data/AN4.LICENSE.html
@ -0,0 +1,53 @@
 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN">
 <html>
 <head>
  <title>AN4 License Terms</title>
  <meta http-equiv="content-type"
 content="text/html; charset=ISO-8859-1">
 </head>
 <body>
 <h2>AN4 License Terms</h2>
 <p>This audio database is free for use for any purpose (commercial or otherwise)
 subject to the restrictions detailed below.</p>
 <pre>
 /* ====================================================================
 * Copyright (c) 1991-2005 Carnegie Mellon University.  All rights
 * reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer. 
 *
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in
 *    the documentation and/or other materials provided with the
 *    distribution.
 *
 * This work was supported in part by funding from the Defense Advanced 
 * Research Projects Agency and the National Science Foundation of the 
 * United States of America, and the CMU Sphinx Speech Consortium.
 *
 * THIS SOFTWARE IS PROVIDED BY CARNEGIE MELLON UNIVERSITY ``AS IS'' AND 
 * ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, 
 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL CARNEGIE MELLON UNIVERSITY
 * NOR ITS EMPLOYEES BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * ====================================================================
 */
 </pre>
 </body>
 </html>
--- a/Tests/Speech/Data/Features/000000000.chunk
+++ b/Tests/Speech/Data/Features/000000000.chunk
--- a/Tests/Speech/Data/README.txt
+++ b/Tests/Speech/Data/README.txt
@ -0,0 +1,3 @@
 Contents of this directory is a modified version of AN4 dataset pre-processed and optimized for CNTK end-to-end testing.
 AN4 dataset is a part of CMU audio databases located at http://www.speech.cs.cmu.edu/databases/an4
 This modified version of dataset is distributed under the terms of a AN4 license which can be found in AN4.LICENSE.html
--- a/Tests/Speech/Data/glob_0000.mlf
+++ b/Tests/Speech/Data/glob_0000.mlf
--- a/Tests/Speech/Data/glob_0000.scp
+++ b/Tests/Speech/Data/glob_0000.scp
@ -0,0 +1,948 @@
 An4/71/71/cen5-fjam-b.mfc=Features/000000000.chunk[0,367]
 An4/213/213/cen4-fsaf2-b.mfc=Features/000000000.chunk[368,805]
 An4/513/513/cen7-mgah-b.mfc=Features/000000000.chunk[806,1173]
 An4/614/614/cen7-mkdb-b.mfc=Features/000000000.chunk[1174,1421]
 An4/507/507/cen1-mgah-b.mfc=Features/000000000.chunk[1422,1669]
 An4/693/693/cen8-mmkw-b.mfc=Features/000000000.chunk[1670,2027]
 An4/918/918/cen4-mtos-b.mfc=Features/000000000.chunk[2028,2335]
 An4/477/477/an257-mewl-b.mfc=Features/000000000.chunk[2336,2943]
 An4/454/454/an70-meht-b.mfc=Features/000000000.chunk[2944,3021]
 An4/254/254/cen6-ftmj-b.mfc=Features/000000000.chunk[3022,3249]
 An4/946/946/cen6-mwhw-b.mfc=Features/000000000.chunk[3250,3467]
 An4/122/122/cen4-fkdo-b.mfc=Features/000000000.chunk[3468,3735]
 An4/181/181/an183-fnsv-b.mfc=Features/000000000.chunk[3736,4093]
 An4/93/93/cen1-fjmd-b.mfc=Features/000000000.chunk[4094,4251]
 An4/128/128/an62-flmm2-b.mfc=Features/000000000.chunk[4252,4409]
 An4/688/688/cen3-mmkw-b.mfc=Features/000000000.chunk[4410,4617]
 An4/872/872/an332-msrb-b.mfc=Features/000000000.chunk[4618,4985]
 An4/624/624/cen5-mkem-b.mfc=Features/000000000.chunk[4986,5383]
 An4/146/146/cen2-flrp-b.mfc=Features/000000000.chunk[5384,5541]
 An4/198/198/cen2-fplp-b.mfc=Features/000000000.chunk[5542,5969]
 An4/239/239/cen4-ftal-b.mfc=Features/000000000.chunk[5970,6187]
 An4/49/49/an291-ffmm-b.mfc=Features/000000000.chunk[6188,6335]
 An4/306/306/cen7-mbmg-b.mfc=Features/000000000.chunk[6336,6733]
 An4/252/252/cen4-ftmj-b.mfc=Features/000000000.chunk[6734,7171]
 An4/800/800/an359-mscg2-b.mfc=Features/000000000.chunk[7172,7509]
 An4/771/771/an236-mrjc2-b.mfc=Features/000000000.chunk[7510,7597]
 An4/880/880/cen5-msrb-b.mfc=Features/000000000.chunk[7598,7955]
 An4/795/795/cen7-mrmg-b.mfc=Features/000000000.chunk[7956,8293]
 An4/821/821/cen7-msct-b.mfc=Features/000000000.chunk[8294,8611]
 An4/255/255/cen7-ftmj-b.mfc=Features/000000000.chunk[8612,8949]
 An4/580/580/an58-mjhp-b.mfc=Features/000000000.chunk[8950,9267]
 An4/70/70/cen4-fjam-b.mfc=Features/000000000.chunk[9268,9595]
 An4/528/528/an171-mjda-b.mfc=Features/000000000.chunk[9596,9963]
 An4/901/901/an35-mtje-b.mfc=Features/000000000.chunk[9964,10371]
 An4/776/776/cen1-mrjc2-b.mfc=Features/000000000.chunk[10372,10779]
 An4/908/908/cen7-mtje-b.mfc=Features/000000000.chunk[10780,11257]
 An4/603/603/an316-mkdb-b.mfc=Features/000000000.chunk[11258,11565]
 An4/544/544/an20-mjdr-b.mfc=Features/000000000.chunk[11566,11853]
 An4/243/243/cen8-ftal-b.mfc=Features/000000000.chunk[11854,12071]
 An4/891/891/cen3-mtcv-b.mfc=Features/000000000.chunk[12072,12269]
 An4/245/245/an212-ftmj-b.mfc=Features/000000000.chunk[12270,12647]
 An4/156/156/an119-fmjc-b.mfc=Features/000000000.chunk[12648,13055]
 An4/446/446/cen5-meab-b.mfc=Features/000000000.chunk[13056,13483]
 An4/801/801/an360-mscg2-b.mfc=Features/000000000.chunk[13484,13601]
 An4/538/538/cen6-mjda-b.mfc=Features/000000000.chunk[13602,13799]
 An4/282/282/an1-mblw-b.mfc=Features/000000000.chunk[13800,13947]
 An4/589/589/cen7-mjhp-b.mfc=Features/000000000.chunk[13948,14275]
 An4/710/710/an389-mmtm-b.mfc=Features/000000000.chunk[14276,14603]
 An4/638/638/cen6-mmaf-b.mfc=Features/000000000.chunk[14604,14811]
 An4/874/874/an334-msrb-b.mfc=Features/000000000.chunk[14812,15029]
 An4/40/40/an40-fejs-b.mfc=Features/000000000.chunk[15030,15337]
 An4/176/176/cen6-fmjd-b.mfc=Features/000000000.chunk[15338,15545]
 An4/732/732/cen8-mnfe-b.mfc=Features/000000000.chunk[15546,15773]
 An4/575/575/cen6-mjgk-b.mfc=Features/000000000.chunk[15774,16191]
 An4/234/234/an329-ftal-b.mfc=Features/000000000.chunk[16192,16429]
 An4/497/497/cen4-mfaa-b.mfc=Features/000000000.chunk[16430,16687]
 An4/619/619/an189-mkem-b.mfc=Features/000000000.chunk[16688,16785]
 An4/303/303/cen4-mbmg-b.mfc=Features/000000000.chunk[16786,17093]
 An4/502/502/an196-mgah-b.mfc=Features/000000000.chunk[17094,17291]
 An4/436/436/cen8-mdxs-b.mfc=Features/000000000.chunk[17292,17619]
 An4/889/889/cen1-mtcv-b.mfc=Features/000000000.chunk[17620,18227]
 An4/697/697/an384-mmsh-b.mfc=Features/000000000.chunk[18228,18475]
 An4/413/413/an108-mdxn-b.mfc=Features/000000000.chunk[18476,18643]
 An4/165/165/cen8-fmjc-b.mfc=Features/000000000.chunk[18644,18901]
 An4/186/186/cen3-fnsv-b.mfc=Features/000000000.chunk[18902,19149]
 An4/274/274/cen1-mblb-b.mfc=Features/000000000.chunk[19150,19417]
 An4/309/309/an202-mcel-b.mfc=Features/000000000.chunk[19418,19525]
 An4/725/725/cen1-mnfe-b.mfc=Features/000000000.chunk[19526,19783]
 An4/699/699/cen1-mmsh-b.mfc=Features/000000000.chunk[19784,20051]
 An4/833/833/cen6-msjm-b.mfc=Features/000000000.chunk[20052,20299]
 An4/857/857/cen4-mskh-b.mfc=Features/000000000.chunk[20300,20687]
 An4/734/734/an82-mnjl-b.mfc=Features/000000000.chunk[20688,21025]
 An4/340/340/cen3-mcfl-b.mfc=Features/000000000.chunk[21026,21263]
 An4/36/36/an36-fejs-b.mfc=Features/000000000.chunk[21264,21641]
 An4/690/690/cen5-mmkw-b.mfc=Features/000000000.chunk[21642,22069]
 An4/545/545/cen1-mjdr-b.mfc=Features/000000000.chunk[22070,22347]
 An4/115/115/an132-fkdo-b.mfc=Features/000000000.chunk[22348,22505]
 An4/48/48/cen8-fejs-b.mfc=Features/000000000.chunk[22506,22723]
 An4/518/518/an249-mjbh-b.mfc=Features/000000000.chunk[22724,22811]
 An4/89/89/an6-fjmd-b.mfc=Features/000000000.chunk[22812,22889]
 An4/668/668/an337-mmdg-b.mfc=Features/000000000.chunk[22890,23007]
 An4/622/622/cen2-mkem-b.mfc=Features/000000000.chunk[23008,23175]
 An4/8/8/cen5-fash-b.mfc=Features/000000000.chunk[23176,23623]
 An4/601/601/cen7-mjjs2-b.mfc=Features/000000000.chunk[23624,24051]
 An4/480/480/an260-mewl-b.mfc=Features/000000000.chunk[24052,24409]
 An4/182/182/an184-fnsv-b.mfc=Features/000000000.chunk[24410,24497]
 An4/179/179/an181-fnsv-b.mfc=Features/000000000.chunk[24498,24825]
 An4/92/92/an9-fjmd-b.mfc=Features/000000000.chunk[24826,25003]
 An4/164/164/cen7-fmjc-b.mfc=Features/000000000.chunk[25004,25251]
 An4/16/16/cen2-fbbh-b.mfc=Features/000000000.chunk[25252,25549]
 An4/657/657/an49-mmap-b.mfc=Features/000000000.chunk[25550,25867]
 An4/723/723/an349-mnfe-b.mfc=Features/000000000.chunk[25868,26325]
 An4/700/700/cen2-mmsh-b.mfc=Features/000000000.chunk[26326,26453]
 An4/675/675/cen4-mmdg-b.mfc=Features/000000000.chunk[26454,26861]
 An4/386/386/an112-mdcs2-b.mfc=Features/000000000.chunk[26862,27129]
 An4/152/152/cen8-flrp-b.mfc=Features/000000000.chunk[27130,27347]
 An4/740/740/cen3-mnjl-b.mfc=Features/000000000.chunk[27348,27465]
 An4/370/370/cen7-mcsc-b.mfc=Features/000000000.chunk[27466,27783]
 An4/683/683/an364-mmkw-b.mfc=Features/000000000.chunk[27784,27861]
 An4/440/440/an139-meab-b.mfc=Features/000000000.chunk[27862,28089]
 An4/789/789/cen1-mrmg-b.mfc=Features/000000000.chunk[28090,28427]
 An4/611/611/cen4-mkdb-b.mfc=Features/000000000.chunk[28428,28685]
 An4/10/10/an86-fbbh-b.mfc=Features/000000000.chunk[28686,29013]
 An4/343/343/cen6-mcfl-b.mfc=Features/000000000.chunk[29014,29251]
 An4/438/438/an137-meab-b.mfc=Features/000000000.chunk[29252,29669]
 An4/456/456/cen2-meht-b.mfc=Features/000000000.chunk[29670,29817]
 An4/489/489/an161-mfaa-b.mfc=Features/000000000.chunk[29818,30075]
 An4/53/53/an295-ffmm-b.mfc=Features/000000000.chunk[30076,30363]
 An4/702/702/cen4-mmsh-b.mfc=Features/000000000.chunk[30364,30681]
 An4/777/777/cen2-mrjc2-b.mfc=Features/000000000.chunk[30682,30999]
 An4/873/873/an333-msrb-b.mfc=Features/000000000.chunk[31000,31097]
 An4/768/768/cen6-mrcb-b.mfc=Features/000000000.chunk[31098,31275]
 An4/552/552/cen8-mjdr-b.mfc=Features/000000000.chunk[31276,31503]
 An4/631/631/an54-mmaf-b.mfc=Features/000000000.chunk[31504,31611]
 An4/476/476/an256-mewl-b.mfc=Features/000000000.chunk[31612,31689]
 An4/151/151/cen7-flrp-b.mfc=Features/000000000.chunk[31690,31937]
 An4/920/920/cen6-mtos-b.mfc=Features/000000000.chunk[31938,32145]
 An4/358/358/cen8-mcrt-b.mfc=Features/000000000.chunk[32146,32463]
 An4/177/177/cen7-fmjd-b.mfc=Features/000000000.chunk[32464,32761]
 An4/635/635/cen3-mmaf-b.mfc=Features/000000000.chunk[32762,32929]
 An4/719/719/cen8-mmtm-b.mfc=Features/000000000.chunk[32930,33207]
 An4/750/750/cen1-mrab-b.mfc=Features/000000000.chunk[33208,33395]
 An4/755/755/cen6-mrab-b.mfc=Features/000000000.chunk[33396,33573]
 An4/721/721/an347-mnfe-b.mfc=Features/000000000.chunk[33574,33661]
 An4/380/380/cen4-mdcs-b.mfc=Features/000000000.chunk[33662,33909]
 An4/625/625/cen6-mkem-b.mfc=Features/000000000.chunk[33910,34117]
 An4/106/106/cen1-fkai-b.mfc=Features/000000000.chunk[34118,34295]
 An4/658/658/an50-mmap-b.mfc=Features/000000000.chunk[34296,34513]
 An4/402/402/an210-mdmc-b.mfc=Features/000000000.chunk[34514,35021]
 An4/192/192/an91-fplp-b.mfc=Features/000000000.chunk[35022,35469]
 An4/416/416/cen1-mdxn-b.mfc=Features/000000000.chunk[35470,35757]
 An4/161/161/cen4-fmjc-b.mfc=Features/000000000.chunk[35758,35965]
 An4/797/797/an356-mscg2-b.mfc=Features/000000000.chunk[35966,36183]
 An4/433/433/cen5-mdxs-b.mfc=Features/000000000.chunk[36184,36691]
 An4/57/57/cen4-ffmm-b.mfc=Features/000000000.chunk[36692,37119]
 An4/157/157/an120-fmjc-b.mfc=Features/000000000.chunk[37120,37347]
 An4/272/272/an374-mblb-b.mfc=Features/000000000.chunk[37348,37575]
 An4/549/549/cen5-mjdr-b.mfc=Features/000000000.chunk[37576,37903]
 An4/41/41/cen1-fejs-b.mfc=Features/000000000.chunk[37904,38341]
 An4/290/290/cen4-mblw-b.mfc=Features/000000000.chunk[38342,38549]
 An4/701/701/cen3-mmsh-b.mfc=Features/000000000.chunk[38550,38677]
 An4/398/398/an206-mdmc-b.mfc=Features/000000000.chunk[38678,39005]
 An4/640/640/cen8-mmaf-b.mfc=Features/000000000.chunk[39006,39323]
 An4/904/904/cen3-mtje-b.mfc=Features/000000000.chunk[39324,39541]
 An4/686/686/cen1-mmkw-b.mfc=Features/000000000.chunk[39542,40039]
 An4/97/97/cen5-fjmd-b.mfc=Features/000000000.chunk[40040,40397]
 An4/259/259/an223-fwxs-b.mfc=Features/000000000.chunk[40398,40495]
 An4/729/729/cen5-mnfe-b.mfc=Features/000000000.chunk[40496,41033]
 An4/709/709/an388-mmtm-b.mfc=Features/000000000.chunk[41034,41131]
 An4/692/692/cen7-mmkw-b.mfc=Features/000000000.chunk[41132,41759]
 An4/2/2/an253-fash-b.mfc=Features/000000000.chunk[41760,41827]
 An4/39/39/an39-fejs-b.mfc=Features/000000000.chunk[41828,42095]
 An4/488/488/cen8-mewl-b.mfc=Features/000000000.chunk[42096,42423]
 An4/411/411/an106-mdxn-b.mfc=Features/000000000.chunk[42424,42601]
 An4/905/905/cen4-mtje-b.mfc=Features/000000000.chunk[42602,43069]
 An4/783/783/cen8-mrjc2-b.mfc=Features/000000000.chunk[43070,43417]
 An4/205/205/an296-fsaf2-b.mfc=Features/000000000.chunk[43418,43705]
 An4/788/788/an285-mrmg-b.mfc=Features/000000000.chunk[43706,44053]
 An4/173/173/cen3-fmjd-b.mfc=Features/000000000.chunk[44054,44251]
 An4/389/389/an115-mdcs2-b.mfc=Features/000000000.chunk[44252,44579]
 An4/412/412/an107-mdxn-b.mfc=Features/000000000.chunk[44580,44867]
 An4/69/69/cen3-fjam-b.mfc=Features/000000000.chunk[44868,45045]
 An4/84/84/cen5-fjdn-b.mfc=Features/000000000.chunk[45046,45273]
 An4/826/826/an229-msjm-b.mfc=Features/000000000.chunk[45274,45361]
 An4/722/722/an348-mnfe-b.mfc=Features/000000000.chunk[45362,45589]
 An4/490/490/an162-mfaa-b.mfc=Features/000000000.chunk[45590,45897]
 An4/335/335/an263-mcfl-b.mfc=Features/000000000.chunk[45898,46275]
 An4/854/854/cen1-mskh-b.mfc=Features/000000000.chunk[46276,46503]
 An4/334/334/an262-mcfl-b.mfc=Features/000000000.chunk[46504,46851]
 An4/403/403/cen1-mdmc-b.mfc=Features/000000000.chunk[46852,47079]
 An4/46/46/cen6-fejs-b.mfc=Features/000000000.chunk[47080,47277]
 An4/154/154/an117-fmjc-b.mfc=Features/000000000.chunk[47278,47595]
 An4/565/565/cen8-mjes-b.mfc=Features/000000000.chunk[47596,47843]
 An4/251/251/cen3-ftmj-b.mfc=Features/000000000.chunk[47844,48071]
 An4/139/139/an21-flrp-b.mfc=Features/000000000.chunk[48072,48479]
 An4/6/6/cen2-fash-b.mfc=Features/000000000.chunk[48480,48607]
 An4/76/76/an122-fjdn-b.mfc=Features/000000000.chunk[48608,48765]
 An4/817/817/cen3-msct-b.mfc=Features/000000000.chunk[48766,48913]
 An4/328/328/cen4-mcen-b.mfc=Features/000000000.chunk[48914,49161]
 An4/293/293/cen7-mblw-b.mfc=Features/000000000.chunk[49162,49409]
 An4/214/214/cen5-fsaf2-b.mfc=Features/000000000.chunk[49410,49797]
 An4/91/91/an8-fjmd-b.mfc=Features/000000000.chunk[49798,49975]
 An4/820/820/cen6-msct-b.mfc=Features/000000000.chunk[49976,50213]
 An4/300/300/cen1-mbmg-b.mfc=Features/000000000.chunk[50214,50491]
 An4/18/18/cen4-fbbh-b.mfc=Features/000000000.chunk[50492,50829]
 An4/526/526/cen7-mjbh-b.mfc=Features/000000000.chunk[50830,51067]
 An4/408/408/cen6-mdmc-b.mfc=Features/000000000.chunk[51068,51285]
 An4/169/169/an194-fmjd-b.mfc=Features/000000000.chunk[51286,51553]
 An4/939/939/an154-mwhw-b.mfc=Features/000000000.chunk[51554,51841]
 An4/931/931/cen4-mtxj-b.mfc=Features/000000000.chunk[51842,52299]
 An4/758/758/an101-mrcb-b.mfc=Features/000000000.chunk[52300,52647]
 An4/781/781/cen6-mrjc2-b.mfc=Features/000000000.chunk[52648,52875]
 An4/321/321/an127-mcen-b.mfc=Features/000000000.chunk[52876,52973]
 An4/199/199/cen3-fplp-b.mfc=Features/000000000.chunk[52974,53271]
 An4/494/494/cen1-mfaa-b.mfc=Features/000000000.chunk[53272,53469]
 An4/560/560/cen3-mjes-b.mfc=Features/000000000.chunk[53470,53547]
 An4/713/713/cen2-mmtm-b.mfc=Features/000000000.chunk[53548,53855]
 An4/938/938/an153-mwhw-b.mfc=Features/000000000.chunk[53856,54143]
 An4/163/163/cen6-fmjc-b.mfc=Features/000000000.chunk[54144,54321]
 An4/338/338/cen1-mcfl-b.mfc=Features/000000000.chunk[54322,54569]
 An4/775/775/an240-mrjc2-b.mfc=Features/000000000.chunk[54570,54777]
 An4/264/264/cen3-fwxs-b.mfc=Features/000000000.chunk[54778,54925]
 An4/224/224/cen2-fsrb-b.mfc=Features/000000000.chunk[54926,55233]
 An4/166/166/an191-fmjd-b.mfc=Features/000000000.chunk[55234,55321]
 An4/80/80/cen1-fjdn-b.mfc=Features/000000000.chunk[55322,55469]
 An4/426/426/an28-mdxs-b.mfc=Features/000000000.chunk[55470,55577]
 An4/737/737/an85-mnjl-b.mfc=Features/000000000.chunk[55578,55965]
 An4/919/919/cen5-mtos-b.mfc=Features/000000000.chunk[55966,56363]
 An4/102/102/an312-fkai-b.mfc=Features/000000000.chunk[56364,56751]
 An4/743/743/cen7-mnjl-b.mfc=Features/000000000.chunk[56752,57129]
 An4/948/948/cen8-mwhw-b.mfc=Features/000000000.chunk[57130,57347]
 An4/17/17/cen3-fbbh-b.mfc=Features/000000000.chunk[57348,57575]
 An4/11/11/an87-fbbh-b.mfc=Features/000000000.chunk[57576,57743]
 An4/344/344/cen7-mcfl-b.mfc=Features/000000000.chunk[57744,58111]
 An4/359/359/an231-mcsc-b.mfc=Features/000000000.chunk[58112,58329]
 An4/203/203/cen7-fplp-b.mfc=Features/000000000.chunk[58330,58877]
 An4/704/704/cen6-mmsh-b.mfc=Features/000000000.chunk[58878,59035]
 An4/331/331/cen7-mcen-b.mfc=Features/000000000.chunk[59036,59323]
 An4/736/736/an84-mnjl-b.mfc=Features/000000000.chunk[59324,59511]
 An4/121/121/cen3-fkdo-b.mfc=Features/000000000.chunk[59512,59769]
 An4/574/574/cen5-mjgk-b.mfc=Features/000000000.chunk[59770,59977]
 An4/143/143/an24-flrp-b.mfc=Features/000000000.chunk[59978,60065]
 An4/209/209/an300-fsaf2-b.mfc=Features/000000000.chunk[60066,60473]
 An4/367/367/cen4-mcsc-b.mfc=Features/000000000.chunk[60474,60731]
 An4/38/38/an38-fejs-b.mfc=Features/000000000.chunk[60732,60809]
 An4/390/390/cen1-mdcs2-b.mfc=Features/000000000.chunk[60810,61057]
 An4/756/756/cen7-mrab-b.mfc=Features/000000000.chunk[61058,61275]
 An4/555/555/an158-mjes-b.mfc=Features/000000000.chunk[61276,61613]
 An4/680/680/an361-mmkw-b.mfc=Features/000000000.chunk[61614,62041]
 An4/578/578/an56-mjhp-b.mfc=Features/000000000.chunk[62042,62419]
 An4/655/655/an47-mmap-b.mfc=Features/000000000.chunk[62420,62667]
 An4/646/646/cen1-mmal-b.mfc=Features/000000000.chunk[62668,63035]
 An4/720/720/an346-mnfe-b.mfc=Features/000000000.chunk[63036,63453]
 An4/608/608/cen1-mkdb-b.mfc=Features/000000000.chunk[63454,63721]
 An4/441/441/an140-meab-b.mfc=Features/000000000.chunk[63722,64299]
 An4/356/356/cen6-mcrt-b.mfc=Features/000000000.chunk[64300,64547]
 An4/926/926/an379-mtxj-b.mfc=Features/000000000.chunk[64548,64625]
 An4/541/541/an16-mjdr-b.mfc=Features/000000000.chunk[64626,64893]
 An4/195/195/an94-fplp-b.mfc=Features/000000000.chunk[64894,65441]
 An4/591/591/an176-mjjs2-b.mfc=Features/000000000.chunk[65442,65789]
 An4/9/9/cen7-fash-b.mfc=Features/000000000.chunk[65790,66037]
 An4/484/484/cen4-mewl-b.mfc=Features/000000000.chunk[66038,66525]
 An4/537/537/cen5-mjda-b.mfc=Features/000000000.chunk[66526,66933]
 An4/242/242/cen7-ftal-b.mfc=Features/000000000.chunk[66934,67171]
 An4/848/848/cen8-msjr-b.mfc=Features/000000000.chunk[67172,67409]
 An4/220/220/an168-fsrb-b.mfc=Features/000000000.chunk[67410,67757]
 An4/906/906/cen5-mtje-b.mfc=Features/000000000.chunk[67758,68185]
 An4/444/444/cen3-meab-b.mfc=Features/000000000.chunk[68186,68373]
 An4/88/88/an10-fjmd-b.mfc=Features/000000000.chunk[68374,68531]
 An4/561/561/cen4-mjes-b.mfc=Features/000000000.chunk[68532,68919]
 An4/728/728/cen4-mnfe-b.mfc=Features/000000000.chunk[68920,69347]
 An4/784/784/an281-mrmg-b.mfc=Features/000000000.chunk[69348,69485]
 An4/55/55/cen2-ffmm-b.mfc=Features/000000000.chunk[69486,69983]
 An4/593/593/an178-mjjs2-b.mfc=Features/000000000.chunk[69984,70061]
 An4/327/327/cen3-mcen-b.mfc=Features/000000000.chunk[70062,70309]
 An4/4/4/an255-fash-b.mfc=Features/000000000.chunk[70310,70567]
 An4/922/922/cen8-mtos-b.mfc=Features/000000000.chunk[70568,70775]
 An4/229/229/cen7-fsrb-b.mfc=Features/000000000.chunk[70776,71253]
 An4/297/297/an268-mbmg-b.mfc=Features/000000000.chunk[71254,71651]
 An4/215/215/cen6-fsaf2-b.mfc=Features/000000000.chunk[71652,71839]
 An4/567/567/an217-mjgk-b.mfc=Features/000000000.chunk[71840,71987]
 An4/96/96/cen4-fjmd-b.mfc=Features/000000000.chunk[71988,72335]
 An4/846/846/cen6-msjr-b.mfc=Features/000000000.chunk[72336,72543]
 An4/850/850/an96-mskh-b.mfc=Features/000000000.chunk[72544,72621]
 An4/492/492/an164-mfaa-b.mfc=Features/000000000.chunk[72622,72859]
 An4/661/661/cen3-mmap-b.mfc=Features/000000000.chunk[72860,72987]
 An4/200/200/cen4-fplp-b.mfc=Features/000000000.chunk[72988,73485]
 An4/82/82/cen3-fjdn-b.mfc=Features/000000000.chunk[73486,73583]
 An4/936/936/an151-mwhw-b.mfc=Features/000000000.chunk[73584,73891]
 An4/60/60/cen7-ffmm-b.mfc=Features/000000000.chunk[73892,74379]
 An4/183/183/an185-fnsv-b.mfc=Features/000000000.chunk[74380,74477]
 An4/667/667/an336-mmdg-b.mfc=Features/000000000.chunk[74478,74785]
 An4/576/576/cen7-mjgk-b.mfc=Features/000000000.chunk[74786,74993]
 An4/212/212/cen3-fsaf2-b.mfc=Features/000000000.chunk[74994,75101]
 An4/779/779/cen4-mrjc2-b.mfc=Features/000000000.chunk[75102,75449]
 An4/418/418/cen3-mdxn-b.mfc=Features/000000000.chunk[75450,75637]
 An4/636/636/cen4-mmaf-b.mfc=Features/000000000.chunk[75638,75935]
 An4/257/257/an221-fwxs-b.mfc=Features/000000000.chunk[75936,76253]
 An4/59/59/cen6-ffmm-b.mfc=Features/000000000.chunk[76254,76481]
 An4/899/899/an33-mtje-b.mfc=Features/000000000.chunk[76482,76879]
 An4/886/886/an303-mtcv-b.mfc=Features/000000000.chunk[76880,77307]
 An4/932/932/cen5-mtxj-b.mfc=Features/000000000.chunk[77308,77735]
 An4/336/336/an264-mcfl-b.mfc=Features/000000000.chunk[77736,77813]
 An4/877/877/cen2-msrb-b.mfc=Features/000000000.chunk[77814,78051]
 An4/629/629/an52-mmaf-b.mfc=Features/000000000.chunk[78052,78199]
 An4/767/767/cen5-mrcb-b.mfc=Features/000000000.chunk[78200,78547]
 An4/374/374/an243-mdcs-b.mfc=Features/000000000.chunk[78548,78635]
 An4/437/437/an136-meab-b.mfc=Features/000000000.chunk[78636,79063]
 An4/202/202/cen6-fplp-b.mfc=Features/000000000.chunk[79064,79451]
 An4/29/29/cen2-fclc-b.mfc=Features/000000000.chunk[79452,79699]
 An4/669/669/an338-mmdg-b.mfc=Features/000000000.chunk[79700,80017]
 An4/216/216/cen7-fsaf2-b.mfc=Features/000000000.chunk[80018,80395]
 An4/227/227/cen5-fsrb-b.mfc=Features/000000000.chunk[80396,80903]
 An4/864/864/an278-msmn-b.mfc=Features/000000000.chunk[80904,81311]
 An4/794/794/cen6-mrmg-b.mfc=Features/000000000.chunk[81312,81549]
 An4/865/865/an279-msmn-b.mfc=Features/000000000.chunk[81550,81837]
 An4/111/111/cen6-fkai-b.mfc=Features/000000000.chunk[81838,82015]
 An4/774/774/an239-mrjc2-b.mfc=Features/000000000.chunk[82016,82293]
 An4/831/831/cen4-msjm-b.mfc=Features/000000000.chunk[82294,82481]
 An4/793/793/cen5-mrmg-b.mfc=Features/000000000.chunk[82482,83049]
 An4/301/301/cen2-mbmg-b.mfc=Features/000000000.chunk[83050,83237]
 An4/325/325/cen1-mcen-b.mfc=Features/000000000.chunk[83238,83485]
 An4/210/210/cen1-fsaf2-b.mfc=Features/000000000.chunk[83486,83863]
 An4/117/117/an134-fkdo-b.mfc=Features/000000000.chunk[83864,83991]
 An4/388/388/an114-mdcs2-b.mfc=Features/000000000.chunk[83992,84289]
 An4/718/718/cen7-mmtm-b.mfc=Features/000000000.chunk[84290,84617]
 An4/174/174/cen4-fmjd-b.mfc=Features/000000000.chunk[84618,84955]
 An4/652/652/cen7-mmal-b.mfc=Features/000000000.chunk[84956,85233]
 An4/228/228/cen6-fsrb-b.mfc=Features/000000000.chunk[85234,85451]
 An4/373/373/an242-mdcs-b.mfc=Features/000000000.chunk[85452,85729]
 An4/175/175/cen5-fmjd-b.mfc=Features/000000000.chunk[85730,86147]
 An4/184/184/cen1-fnsv-b.mfc=Features/000000000.chunk[86148,86505]
 An4/393/393/cen4-mdcs2-b.mfc=Features/000000000.chunk[86506,86723]
 An4/319/319/cen8-mcel-b.mfc=Features/000000000.chunk[86724,86961]
 An4/291/291/cen5-mblw-b.mfc=Features/000000000.chunk[86962,87079]
 An4/584/584/cen2-mjhp-b.mfc=Features/000000000.chunk[87080,87287]
 An4/827/827/an230-msjm-b.mfc=Features/000000000.chunk[87288,87385]
 An4/628/628/an51-mmaf-b.mfc=Features/000000000.chunk[87386,87733]
 An4/295/295/an266-mbmg-b.mfc=Features/000000000.chunk[87734,87901]
 An4/317/317/cen5-mcel-b.mfc=Features/000000000.chunk[87902,88269]
 An4/431/431/cen3-mdxs-b.mfc=Features/000000000.chunk[88270,88517]
 An4/52/52/an294-ffmm-b.mfc=Features/000000000.chunk[88518,88785]
 An4/491/491/an163-mfaa-b.mfc=Features/000000000.chunk[88786,89023]
 An4/844/844/cen4-msjr-b.mfc=Features/000000000.chunk[89024,89261]
 An4/116/116/an133-fkdo-b.mfc=Features/000000000.chunk[89262,89369]
 An4/61/61/cen8-ffmm-b.mfc=Features/000000000.chunk[89370,89717]
 An4/118/118/an135-fkdo-b.mfc=Features/000000000.chunk[89718,89995]
 An4/131/131/an65-flmm2-b.mfc=Features/000000000.chunk[89996,90233]
 An4/878/878/cen3-msrb-b.mfc=Features/000000000.chunk[90234,90431]
 An4/352/352/cen2-mcrt-b.mfc=Features/000000000.chunk[90432,90739]
 An4/132/132/cen1-flmm2-b.mfc=Features/000000000.chunk[90740,91127]
 An4/230/230/cen8-fsrb-b.mfc=Features/000000000.chunk[91128,91425]
 An4/933/933/cen6-mtxj-b.mfc=Features/000000000.chunk[91426,91603]
 An4/535/535/cen3-mjda-b.mfc=Features/000000000.chunk[91604,91741]
 An4/531/531/an174-mjda-b.mfc=Features/000000000.chunk[91742,91859]
 An4/525/525/cen6-mjbh-b.mfc=Features/000000000.chunk[91860,92057]
 An4/74/74/cen8-fjam-b.mfc=Features/000000000.chunk[92058,92265]
 An4/644/644/an324-mmal-b.mfc=Features/000000000.chunk[92266,92363]
 An4/240/240/cen5-ftal-b.mfc=Features/000000000.chunk[92364,92601]
 An4/726/726/cen2-mnfe-b.mfc=Features/000000000.chunk[92602,92809]
 An4/425/425/an27-mdxs-b.mfc=Features/000000000.chunk[92810,93297]
 An4/612/612/cen5-mkdb-b.mfc=Features/000000000.chunk[93298,93655]
 An4/698/698/an385-mmsh-b.mfc=Features/000000000.chunk[93656,93733]
 An4/787/787/an284-mrmg-b.mfc=Features/000000000.chunk[93734,93881]
 An4/666/666/cen8-mmap-b.mfc=Features/000000000.chunk[93882,94099]
 An4/31/31/cen4-fclc-b.mfc=Features/000000000.chunk[94100,94467]
 An4/470/470/cen3-mema-b.mfc=Features/000000000.chunk[94468,94615]
 An4/782/782/cen7-mrjc2-b.mfc=Features/000000000.chunk[94616,94943]
 An4/824/824/an227-msjm-b.mfc=Features/000000000.chunk[94944,95211]
 An4/287/287/cen1-mblw-b.mfc=Features/000000000.chunk[95212,95379]
 An4/748/748/an74-mrab-b.mfc=Features/000000000.chunk[95380,95667]
 An4/241/241/cen6-ftal-b.mfc=Features/000000000.chunk[95668,95835]
 An4/832/832/cen5-msjm-b.mfc=Features/000000000.chunk[95836,96123]
 An4/664/664/cen6-mmap-b.mfc=Features/000000000.chunk[96124,96311]
 An4/347/347/an142-mcrt-b.mfc=Features/000000000.chunk[96312,96509]
 An4/377/377/cen1-mdcs-b.mfc=Features/000000000.chunk[96510,96757]
 An4/124/124/cen6-fkdo-b.mfc=Features/000000000.chunk[96758,96995]
 An4/724/724/an350-mnfe-b.mfc=Features/000000000.chunk[96996,97133]
 An4/442/442/cen1-meab-b.mfc=Features/000000000.chunk[97134,97601]
 An4/742/742/cen6-mnjl-b.mfc=Features/000000000.chunk[97602,97809]
 An4/500/500/cen7-mfaa-b.mfc=Features/000000000.chunk[97810,98047]
 An4/909/909/cen8-mtje-b.mfc=Features/000000000.chunk[98048,98245]
 An4/626/626/cen7-mkem-b.mfc=Features/000000000.chunk[98246,98413]
 An4/627/627/cen8-mkem-b.mfc=Features/000000000.chunk[98414,98671]
 An4/401/401/an209-mdmc-b.mfc=Features/000000000.chunk[98672,98979]
 An4/838/838/an353-msjr-b.mfc=Features/000000000.chunk[98980,99057]
 An4/415/415/an110-mdxn-b.mfc=Features/000000000.chunk[99058,99265]
 An4/225/225/cen3-fsrb-b.mfc=Features/000000000.chunk[99266,99473]
 An4/595/595/an180-mjjs2-b.mfc=Features/000000000.chunk[99474,99571]
 An4/673/673/cen2-mmdg-b.mfc=Features/000000000.chunk[99572,99859]
 An4/162/162/cen5-fmjc-b.mfc=Features/000000000.chunk[99860,100147]
 An4/679/679/cen8-mmdg-b.mfc=Features/000000000.chunk[100148,100445]
 An4/590/590/cen8-mjhp-b.mfc=Features/000000000.chunk[100446,100673]
 An4/299/299/an270-mbmg-b.mfc=Features/000000000.chunk[100674,100881]
 An4/805/805/cen4-mscg2-b.mfc=Features/000000000.chunk[100882,101249]
 An4/197/197/cen1-fplp-b.mfc=Features/000000000.chunk[101250,101807]
 An4/267/267/cen6-fwxs-b.mfc=Features/000000000.chunk[101808,102115]
 An4/630/630/an53-mmaf-b.mfc=Features/000000000.chunk[102116,102463]
 An4/888/888/an305-mtcv-b.mfc=Features/000000000.chunk[102464,102691]
 An4/812/812/an343-msct-b.mfc=Features/000000000.chunk[102692,102879]
 An4/233/233/an328-ftal-b.mfc=Features/000000000.chunk[102880,103207]
 An4/529/529/an172-mjda-b.mfc=Features/000000000.chunk[103208,103425]
 An4/707/707/an386-mmtm-b.mfc=Features/000000000.chunk[103426,103663]
 An4/592/592/an177-mjjs2-b.mfc=Features/000000000.chunk[103664,103791]
 An4/130/130/an64-flmm2-b.mfc=Features/000000000.chunk[103792,103919]
 An4/310/310/an203-mcel-b.mfc=Features/000000000.chunk[103920,104167]
 An4/170/170/an195-fmjd-b.mfc=Features/000000000.chunk[104168,104485]
 An4/119/119/cen1-fkdo-b.mfc=Features/000000000.chunk[104486,104753]
 An4/345/345/cen8-mcfl-b.mfc=Features/000000000.chunk[104754,105011]
 An4/365/365/cen2-mcsc-b.mfc=Features/000000000.chunk[105012,105319]
 An4/735/735/an83-mnjl-b.mfc=Features/000000000.chunk[105320,105467]
 An4/633/633/cen1-mmaf-b.mfc=Features/000000000.chunk[105468,105795]
 An4/654/654/an46-mmap-b.mfc=Features/000000000.chunk[105796,105893]
 An4/149/149/cen5-flrp-b.mfc=Features/000000000.chunk[105894,106341]
 An4/751/751/cen2-mrab-b.mfc=Features/000000000.chunk[106342,106519]
 An4/238/238/cen3-ftal-b.mfc=Features/000000000.chunk[106520,106657]
 An4/360/360/an232-mcsc-b.mfc=Features/000000000.chunk[106658,106745]
 An4/881/881/cen6-msrb-b.mfc=Features/000000000.chunk[106746,106973]
 An4/757/757/cen8-mrab-b.mfc=Features/000000000.chunk[106974,107171]
 An4/400/400/an208-mdmc-b.mfc=Features/000000000.chunk[107172,107549]
 An4/168/168/an193-fmjd-b.mfc=Features/000000000.chunk[107550,107977]
 An4/897/897/an31-mtje-b.mfc=Features/000000000.chunk[107978,108345]
 An4/530/530/an173-mjda-b.mfc=Features/000000000.chunk[108346,108653]
 An4/566/566/an216-mjgk-b.mfc=Features/000000000.chunk[108654,108911]
 An4/95/95/cen3-fjmd-b.mfc=Features/000000000.chunk[108912,109009]
 An4/43/43/cen3-fejs-b.mfc=Features/000000000.chunk[109010,109287]
 An4/753/753/cen4-mrab-b.mfc=Features/000000000.chunk[109288,109615]
 An4/405/405/cen3-mdmc-b.mfc=Features/000000000.chunk[109616,109743]
 An4/66/66/an80-fjam-b.mfc=Features/000000000.chunk[109744,109831]
 An4/858/858/cen5-mskh-b.mfc=Features/000000000.chunk[109832,110219]
 An4/852/852/an98-mskh-b.mfc=Features/000000000.chunk[110220,110307]
 An4/237/237/cen2-ftal-b.mfc=Features/000000000.chunk[110308,110485]
 An4/602/602/cen8-mjjs2-b.mfc=Features/000000000.chunk[110486,110903]
 An4/842/842/cen2-msjr-b.mfc=Features/000000000.chunk[110904,111131]
 An4/13/13/an89-fbbh-b.mfc=Features/000000000.chunk[111132,111769]
 An4/283/283/an2-mblw-b.mfc=Features/000000000.chunk[111770,111917]
 An4/460/460/cen6-meht-b.mfc=Features/000000000.chunk[111918,112105]
 An4/20/20/cen6-fbbh-b.mfc=Features/000000000.chunk[112106,112343]
 An4/308/308/an201-mcel-b.mfc=Features/000000000.chunk[112344,112451]
 An4/471/471/cen4-mema-b.mfc=Features/000000000.chunk[112452,112819]
 An4/546/546/cen2-mjdr-b.mfc=Features/000000000.chunk[112820,113017]
 An4/468/468/cen1-mema-b.mfc=Features/000000000.chunk[113018,113255]
 An4/236/236/cen1-ftal-b.mfc=Features/000000000.chunk[113256,113453]
 An4/372/372/an241-mdcs-b.mfc=Features/000000000.chunk[113454,113691]
 An4/395/395/cen6-mdcs2-b.mfc=Features/000000000.chunk[113692,113869]
 An4/945/945/cen5-mwhw-b.mfc=Features/000000000.chunk[113870,114177]
 An4/754/754/cen5-mrab-b.mfc=Features/000000000.chunk[114178,114445]
 An4/509/509/cen3-mgah-b.mfc=Features/000000000.chunk[114446,114643]
 An4/556/556/an159-mjes-b.mfc=Features/000000000.chunk[114644,114721]
 An4/594/594/an179-mjjs2-b.mfc=Features/000000000.chunk[114722,115229]
 An4/487/487/cen7-mewl-b.mfc=Features/000000000.chunk[115230,115487]
 An4/684/684/an365-mmkw-b.mfc=Features/000000000.chunk[115488,115915]
 An4/855/855/cen2-mskh-b.mfc=Features/000000000.chunk[115916,116113]
 An4/439/439/an138-meab-b.mfc=Features/000000000.chunk[116114,116401]
 An4/354/354/cen4-mcrt-b.mfc=Features/000000000.chunk[116402,116879]
 An4/26/26/an149-fclc-b.mfc=Features/000000000.chunk[116880,117107]
 An4/588/588/cen6-mjhp-b.mfc=Features/000000000.chunk[117108,117275]
 An4/823/823/an226-msjm-b.mfc=Features/000000000.chunk[117276,117363]
 An4/68/68/cen2-fjam-b.mfc=Features/000000000.chunk[117364,117511]
 An4/101/101/an311-fkai-b.mfc=Features/000000000.chunk[117512,117819]
 An4/98/98/cen6-fjmd-b.mfc=Features/000000000.chunk[117820,118017]
 An4/450/450/an66-meht-b.mfc=Features/000000000.chunk[118018,118285]
 An4/662/662/cen4-mmap-b.mfc=Features/000000000.chunk[118286,118753]
 An4/452/452/an68-meht-b.mfc=Features/000000000.chunk[118754,118961]
 An4/687/687/cen2-mmkw-b.mfc=Features/000000000.chunk[118962,119419]
 An4/218/218/an166-fsrb-b.mfc=Features/000000000.chunk[119420,119667]
 An4/314/314/cen2-mcel-b.mfc=Features/000000000.chunk[119668,119955]
 An4/33/33/cen6-fclc-b.mfc=Features/000000000.chunk[119956,120163]
 An4/424/424/an26-mdxs-b.mfc=Features/000000000.chunk[120164,120281]
 An4/615/615/cen8-mkdb-b.mfc=Features/000000000.chunk[120282,120529]
 An4/298/298/an269-mbmg-b.mfc=Features/000000000.chunk[120530,120927]
 An4/527/527/cen8-mjbh-b.mfc=Features/000000000.chunk[120928,121145]
 An4/15/15/cen1-fbbh-b.mfc=Features/000000000.chunk[121146,121383]
 An4/910/910/an366-mtos-b.mfc=Features/000000000.chunk[121384,121661]
 An4/158/158/cen1-fmjc-b.mfc=Features/000000000.chunk[121662,121949]
 An4/246/246/an213-ftmj-b.mfc=Features/000000000.chunk[121950,122347]
 An4/849/849/an100-mskh-b.mfc=Features/000000000.chunk[122348,122425]
 An4/56/56/cen3-ffmm-b.mfc=Features/000000000.chunk[122426,122623]
 An4/404/404/cen2-mdmc-b.mfc=Features/000000000.chunk[122624,122871]
 An4/351/351/cen1-mcrt-b.mfc=Features/000000000.chunk[122872,123289]
 An4/863/863/an277-msmn-b.mfc=Features/000000000.chunk[123290,123377]
 An4/322/322/an128-mcen-b.mfc=Features/000000000.chunk[123378,123775]
 An4/419/419/cen4-mdxn-b.mfc=Features/000000000.chunk[123776,124043]
 An4/86/86/cen7-fjdn-b.mfc=Features/000000000.chunk[124044,124301]
 An4/311/311/an204-mcel-b.mfc=Features/000000000.chunk[124302,124399]
 An4/142/142/an23-flrp-b.mfc=Features/000000000.chunk[124400,124807]
 An4/25/25/an148-fclc-b.mfc=Features/000000000.chunk[124808,125205]
 An4/947/947/cen7-mwhw-b.mfc=Features/000000000.chunk[125206,125493]
 An4/250/250/cen2-ftmj-b.mfc=Features/000000000.chunk[125494,125781]
 An4/381/381/cen5-mdcs-b.mfc=Features/000000000.chunk[125782,126059]
 An4/927/927/an380-mtxj-b.mfc=Features/000000000.chunk[126060,126247]
 An4/941/941/cen1-mwhw-b.mfc=Features/000000000.chunk[126248,126425]
 An4/769/769/cen7-mrcb-b.mfc=Features/000000000.chunk[126426,126863]
 An4/685/685/an59-mmkw-b.mfc=Features/000000000.chunk[126864,127411]
 An4/72/72/cen6-fjam-b.mfc=Features/000000000.chunk[127412,127699]
 An4/420/420/cen5-mdxn-b.mfc=Features/000000000.chunk[127700,127907]
 An4/457/457/cen3-meht-b.mfc=Features/000000000.chunk[127908,128095]
 An4/279/279/cen6-mblb-b.mfc=Features/000000000.chunk[128096,128263]
 An4/656/656/an48-mmap-b.mfc=Features/000000000.chunk[128264,128481]
 An4/773/773/an238-mrjc2-b.mfc=Features/000000000.chunk[128482,128559]
 An4/562/562/cen5-mjes-b.mfc=Features/000000000.chunk[128560,128797]
 An4/811/811/an342-msct-b.mfc=Features/000000000.chunk[128798,129205]
 An4/23/23/an146-fclc-b.mfc=Features/000000000.chunk[129206,129543]
 An4/391/391/cen2-mdcs2-b.mfc=Features/000000000.chunk[129544,129761]
 An4/172/172/cen2-fmjd-b.mfc=Features/000000000.chunk[129762,130099]
 An4/185/185/cen2-fnsv-b.mfc=Features/000000000.chunk[130100,130377]
 An4/78/78/an124-fjdn-b.mfc=Features/000000000.chunk[130378,130645]
 An4/148/148/cen4-flrp-b.mfc=Features/000000000.chunk[130646,130943]
 An4/253/253/cen5-ftmj-b.mfc=Features/000000000.chunk[130944,131361]
 An4/445/445/cen4-meab-b.mfc=Features/000000000.chunk[131362,131749]
 An4/523/523/cen4-mjbh-b.mfc=Features/000000000.chunk[131750,131907]
 An4/524/524/cen5-mjbh-b.mfc=Features/000000000.chunk[131908,132325]
 An4/428/428/an30-mdxs-b.mfc=Features/000000000.chunk[132326,132423]
 An4/315/315/cen3-mcel-b.mfc=Features/000000000.chunk[132424,132611]
 An4/281/281/cen8-mblb-b.mfc=Features/000000000.chunk[132612,132819]
 An4/570/570/an220-mjgk-b.mfc=Features/000000000.chunk[132820,132987]
 An4/727/727/cen3-mnfe-b.mfc=Features/000000000.chunk[132988,133155]
 An4/231/231/an326-ftal-b.mfc=Features/000000000.chunk[133156,133273]
 An4/193/193/an92-fplp-b.mfc=Features/000000000.chunk[133274,133871]
 An4/892/892/cen4-mtcv-b.mfc=Features/000000000.chunk[133872,134479]
 An4/834/834/cen7-msjm-b.mfc=Features/000000000.chunk[134480,134747]
 An4/144/144/an25-flrp-b.mfc=Features/000000000.chunk[134748,135185]
 An4/828/828/cen1-msjm-b.mfc=Features/000000000.chunk[135186,135413]
 An4/934/934/cen7-mtxj-b.mfc=Features/000000000.chunk[135414,135661]
 An4/387/387/an113-mdcs2-b.mfc=Features/000000000.chunk[135662,136139]
 An4/434/434/cen6-mdxs-b.mfc=Features/000000000.chunk[136140,136377]
 An4/469/469/cen2-mema-b.mfc=Features/000000000.chunk[136378,136515]
 An4/232/232/an327-ftal-b.mfc=Features/000000000.chunk[136516,136673]
 An4/378/378/cen2-mdcs-b.mfc=Features/000000000.chunk[136674,136831]
 An4/275/275/cen2-mblb-b.mfc=Features/000000000.chunk[136832,137079]
 An4/837/837/an352-msjr-b.mfc=Features/000000000.chunk[137080,137207]
 An4/447/447/cen6-meab-b.mfc=Features/000000000.chunk[137208,137425]
 An4/521/521/cen2-mjbh-b.mfc=Features/000000000.chunk[137426,137573]
 An4/733/733/an81-mnjl-b.mfc=Features/000000000.chunk[137574,137791]
 An4/510/510/cen4-mgah-b.mfc=Features/000000000.chunk[137792,138119]
 An4/276/276/cen3-mblb-b.mfc=Features/000000000.chunk[138120,138227]
 An4/894/894/cen6-mtcv-b.mfc=Features/000000000.chunk[138228,138465]
 An4/741/741/cen5-mnjl-b.mfc=Features/000000000.chunk[138466,138783]
 An4/898/898/an32-mtje-b.mfc=Features/000000000.chunk[138784,138881]
 An4/532/532/an175-mjda-b.mfc=Features/000000000.chunk[138882,138959]
 An4/150/150/cen6-flrp-b.mfc=Features/000000000.chunk[138960,139177]
 An4/280/280/cen7-mblb-b.mfc=Features/000000000.chunk[139178,139555]
 An4/902/902/cen1-mtje-b.mfc=Features/000000000.chunk[139556,139813]
 An4/896/896/cen8-mtcv-b.mfc=Features/000000000.chunk[139814,140101]
 An4/900/900/an34-mtje-b.mfc=Features/000000000.chunk[140102,140209]
 An4/323/323/an129-mcen-b.mfc=Features/000000000.chunk[140210,140357]
 An4/579/579/an57-mjhp-b.mfc=Features/000000000.chunk[140358,140595]
 An4/451/451/an67-meht-b.mfc=Features/000000000.chunk[140596,140673]
 An4/830/830/cen3-msjm-b.mfc=Features/000000000.chunk[140674,140851]
 An4/75/75/an121-fjdn-b.mfc=Features/000000000.chunk[140852,140919]
 An4/194/194/an93-fplp-b.mfc=Features/000000000.chunk[140920,141027]
 An4/620/620/an190-mkem-b.mfc=Features/000000000.chunk[141028,141295]
 An4/266/266/cen5-fwxs-b.mfc=Features/000000000.chunk[141296,141713]
 An4/659/659/cen1-mmap-b.mfc=Features/000000000.chunk[141714,141931]
 An4/903/903/cen2-mtje-b.mfc=Features/000000000.chunk[141932,142359]
 An4/189/189/cen6-fnsv-b.mfc=Features/000000000.chunk[142360,142617]
 An4/271/271/an373-mblb-b.mfc=Features/000000000.chunk[142618,143015]
 An4/67/67/cen1-fjam-b.mfc=Features/000000000.chunk[143016,143253]
 An4/219/219/an167-fsrb-b.mfc=Features/000000000.chunk[143254,143511]
 An4/778/778/cen3-mrjc2-b.mfc=Features/000000000.chunk[143512,143719]
 An4/814/814/an345-msct-b.mfc=Features/000000000.chunk[143720,144287]
 An4/829/829/cen2-msjm-b.mfc=Features/000000000.chunk[144288,144525]
 An4/47/47/cen7-fejs-b.mfc=Features/000000000.chunk[144526,144823]
 An4/799/799/an358-mscg2-b.mfc=Features/000000000.chunk[144824,145241]
 An4/804/804/cen3-mscg2-b.mfc=Features/000000000.chunk[145242,145469]
 An4/329/329/cen5-mcen-b.mfc=Features/000000000.chunk[145470,145887]
 An4/600/600/cen6-mjjs2-b.mfc=Features/000000000.chunk[145888,146225]
 An4/876/876/cen1-msrb-b.mfc=Features/000000000.chunk[146226,146493]
 An4/708/708/an387-mmtm-b.mfc=Features/000000000.chunk[146494,146781]
 An4/24/24/an147-fclc-b.mfc=Features/000000000.chunk[146782,147069]
 An4/808/808/cen7-mscg2-b.mfc=Features/000000000.chunk[147070,147457]
 An4/313/313/cen1-mcel-b.mfc=Features/000000000.chunk[147458,147765]
 An4/482/482/cen2-mewl-b.mfc=Features/000000000.chunk[147766,147923]
 An4/51/51/an293-ffmm-b.mfc=Features/000000000.chunk[147924,148261]
 An4/935/935/cen8-mtxj-b.mfc=Features/000000000.chunk[148262,148529]
 An4/244/244/an211-ftmj-b.mfc=Features/000000000.chunk[148530,148737]
 An4/396/396/cen7-mdcs2-b.mfc=Features/000000000.chunk[148738,148945]
 An4/745/745/an71-mrab-b.mfc=Features/000000000.chunk[148946,149253]
 An4/569/569/an219-mjgk-b.mfc=Features/000000000.chunk[149254,149331]
 An4/277/277/cen4-mblb-b.mfc=Features/000000000.chunk[149332,149599]
 An4/371/371/cen8-mcsc-b.mfc=Features/000000000.chunk[149600,149847]
 An4/650/650/cen5-mmal-b.mfc=Features/000000000.chunk[149848,150175]
 An4/135/135/cen4-flmm2-b.mfc=Features/000000000.chunk[150176,150793]
 An4/206/206/an297-fsaf2-b.mfc=Features/000000000.chunk[150794,151221]
 An4/294/294/cen8-mblw-b.mfc=Features/000000000.chunk[151222,151419]
 An4/85/85/cen6-fjdn-b.mfc=Features/000000000.chunk[151420,151567]
 An4/785/785/an282-mrmg-b.mfc=Features/000000000.chunk[151568,151695]
 An4/406/406/cen4-mdmc-b.mfc=Features/000000000.chunk[151696,152323]
 An4/474/474/cen7-mema-b.mfc=Features/000000000.chunk[152324,152561]
 An4/790/790/cen2-mrmg-b.mfc=Features/000000000.chunk[152562,152919]
 An4/463/463/an286-mema-b.mfc=Features/000000000.chunk[152920,153027]
 An4/559/559/cen2-mjes-b.mfc=Features/000000000.chunk[153028,153255]
 An4/353/353/cen3-mcrt-b.mfc=Features/000000000.chunk[153256,153523]
 An4/435/435/cen7-mdxs-b.mfc=Features/000000000.chunk[153524,153941]
 An4/145/145/cen1-flrp-b.mfc=Features/000000000.chunk[153942,154229]
 An4/278/278/cen5-mblb-b.mfc=Features/000000000.chunk[154230,154557]
 An4/517/517/an248-mjbh-b.mfc=Features/000000000.chunk[154558,154765]
 An4/65/65/an79-fjam-b.mfc=Features/000000000.chunk[154766,155073]
 An4/341/341/cen4-mcfl-b.mfc=Features/000000000.chunk[155074,155411]
 An4/520/520/cen1-mjbh-b.mfc=Features/000000000.chunk[155412,155679]
 An4/137/137/cen7-flmm2-b.mfc=Features/000000000.chunk[155680,155967]
 An4/806/806/cen5-mscg2-b.mfc=Features/000000000.chunk[155968,156415]
 An4/429/429/cen1-mdxs-b.mfc=Features/000000000.chunk[156416,156933]
 An4/610/610/cen3-mkdb-b.mfc=Features/000000000.chunk[156934,157151]
 An4/869/869/cen7-msmn-b.mfc=Features/000000000.chunk[157152,157539]
 An4/141/141/an22-flrp-b.mfc=Features/000000000.chunk[157540,157637]
 An4/791/791/cen3-mrmg-b.mfc=Features/000000000.chunk[157638,157805]
 An4/289/289/cen3-mblw-b.mfc=Features/000000000.chunk[157806,157963]
 An4/711/711/an390-mmtm-b.mfc=Features/000000000.chunk[157964,158061]
 An4/432/432/cen4-mdxs-b.mfc=Features/000000000.chunk[158062,158369]
 An4/350/350/an145-mcrt-b.mfc=Features/000000000.chunk[158370,158557]
 An4/670/670/an339-mmdg-b.mfc=Features/000000000.chunk[158558,158905]
 An4/581/581/an59-mjhp-b.mfc=Features/000000000.chunk[158906,159203]
 An4/461/461/cen7-meht-b.mfc=Features/000000000.chunk[159204,159531]
 An4/103/103/an313-fkai-b.mfc=Features/000000000.chunk[159532,160009]
 An4/263/263/cen2-fwxs-b.mfc=Features/000000000.chunk[160010,160317]
 An4/362/362/an234-mcsc-b.mfc=Features/000000000.chunk[160318,160415]
 An4/478/478/an258-mewl-b.mfc=Features/000000000.chunk[160416,160633]
 An4/786/786/an283-mrmg-b.mfc=Features/000000000.chunk[160634,161131]
 An4/512/512/cen6-mgah-b.mfc=Features/000000000.chunk[161132,161369]
 An4/847/847/cen7-msjr-b.mfc=Features/000000000.chunk[161370,161647]
 An4/498/498/cen5-mfaa-b.mfc=Features/000000000.chunk[161648,161975]
 An4/916/916/cen2-mtos-b.mfc=Features/000000000.chunk[161976,162233]
 An4/410/410/cen8-mdmc-b.mfc=Features/000000000.chunk[162234,162561]
 An4/459/459/cen5-meht-b.mfc=Features/000000000.chunk[162562,162859]
 An4/223/223/cen1-fsrb-b.mfc=Features/000000000.chunk[162860,163137]
 An4/764/764/cen2-mrcb-b.mfc=Features/000000000.chunk[163138,163325]
 An4/564/564/cen7-mjes-b.mfc=Features/000000000.chunk[163326,163663]
 An4/5/5/cen1-fash-b.mfc=Features/000000000.chunk[163664,164011]
 An4/129/129/an63-flmm2-b.mfc=Features/000000000.chunk[164012,164249]
 An4/369/369/cen6-mcsc-b.mfc=Features/000000000.chunk[164250,164457]
 An4/87/87/cen8-fjdn-b.mfc=Features/000000000.chunk[164458,164625]
 An4/167/167/an192-fmjd-b.mfc=Features/000000000.chunk[164626,165043]
 An4/598/598/cen4-mjjs2-b.mfc=Features/000000000.chunk[165044,165511]
 An4/188/188/cen5-fnsv-b.mfc=Features/000000000.chunk[165512,166029]
 An4/749/749/an75-mrab-b.mfc=Features/000000000.chunk[166030,166347]
 An4/582/582/an60-mjhp-b.mfc=Features/000000000.chunk[166348,166435]
 An4/160/160/cen3-fmjc-b.mfc=Features/000000000.chunk[166436,166633]
 An4/180/180/an182-fnsv-b.mfc=Features/000000000.chunk[166634,166761]
 An4/682/682/an363-mmkw-b.mfc=Features/000000000.chunk[166762,167379]
 An4/339/339/cen2-mcfl-b.mfc=Features/000000000.chunk[167380,167907]
 An4/921/921/cen7-mtos-b.mfc=Features/000000000.chunk[167908,168075]
 An4/421/421/cen6-mdxn-b.mfc=Features/000000000.chunk[168076,168253]
 An4/247/247/an214-ftmj-b.mfc=Features/000000000.chunk[168254,168391]
 An4/815/815/cen1-msct-b.mfc=Features/000000000.chunk[168392,168659]
 An4/671/671/an340-mmdg-b.mfc=Features/000000000.chunk[168660,168787]
 An4/616/616/an186-mkem-b.mfc=Features/000000000.chunk[168788,169105]
 An4/196/196/an95-fplp-b.mfc=Features/000000000.chunk[169106,169733]
 An4/235/235/an330-ftal-b.mfc=Features/000000000.chunk[169734,169901]
 An4/268/268/cen7-fwxs-b.mfc=Features/000000000.chunk[169902,170319]
 An4/506/506/an200-mgah-b.mfc=Features/000000000.chunk[170320,170417]
 An4/647/647/cen2-mmal-b.mfc=Features/000000000.chunk[170418,170615]
 An4/127/127/an61-flmm2-b.mfc=Features/000000000.chunk[170616,170873]
 An4/803/803/cen2-mscg2-b.mfc=Features/000000000.chunk[170874,171151]
 An4/475/475/cen8-mema-b.mfc=Features/000000000.chunk[171152,171499]
 An4/472/472/cen5-mema-b.mfc=Features/000000000.chunk[171500,171617]
 An4/599/599/cen5-mjjs2-b.mfc=Features/000000000.chunk[171618,172245]
 An4/108/108/cen3-fkai-b.mfc=Features/000000000.chunk[172246,172353]
 An4/357/357/cen7-mcrt-b.mfc=Features/000000000.chunk[172354,172711]
 An4/342/342/cen5-mcfl-b.mfc=Features/000000000.chunk[172712,173029]
 An4/714/714/cen3-mmtm-b.mfc=Features/000000000.chunk[173030,173177]
 An4/747/747/an73-mrab-b.mfc=Features/000000000.chunk[173178,173325]
 An4/643/643/an323-mmal-b.mfc=Features/000000000.chunk[173326,173523]
 An4/99/99/cen7-fjmd-b.mfc=Features/000000000.chunk[173524,173741]
 An4/503/503/an197-mgah-b.mfc=Features/000000000.chunk[173742,173949]
 An4/533/533/cen1-mjda-b.mfc=Features/000000000.chunk[173950,174167]
 An4/691/691/cen6-mmkw-b.mfc=Features/000000000.chunk[174168,174525]
 An4/305/305/cen6-mbmg-b.mfc=Features/000000000.chunk[174526,174763]
 An4/191/191/cen8-fnsv-b.mfc=Features/000000000.chunk[174764,175101]
 An4/772/772/an237-mrjc2-b.mfc=Features/000000000.chunk[175102,175239]
 An4/273/273/an375-mblb-b.mfc=Features/000000000.chunk[175240,175467]
 An4/660/660/cen2-mmap-b.mfc=Features/000000000.chunk[175468,175615]
 An4/540/540/cen8-mjda-b.mfc=Features/000000000.chunk[175616,175903]
 An4/930/930/cen3-mtxj-b.mfc=Features/000000000.chunk[175904,176001]
 An4/346/346/an141-mcrt-b.mfc=Features/000000000.chunk[176002,176089]
 An4/125/125/cen7-fkdo-b.mfc=Features/000000000.chunk[176090,176447]
 An4/107/107/cen2-fkai-b.mfc=Features/000000000.chunk[176448,176605]
 An4/504/504/an198-mgah-b.mfc=Features/000000000.chunk[176606,176843]
 An4/316/316/cen4-mcel-b.mfc=Features/000000000.chunk[176844,177161]
 An4/840/840/an355-msjr-b.mfc=Features/000000000.chunk[177162,177359]
 An4/689/689/cen4-mmkw-b.mfc=Features/000000000.chunk[177360,177877]
 An4/113/113/cen8-fkai-b.mfc=Features/000000000.chunk[177878,178095]
 An4/548/548/cen4-mjdr-b.mfc=Features/000000000.chunk[178096,178533]
 An4/915/915/cen1-mtos-b.mfc=Features/000000000.chunk[178534,178741]
 An4/326/326/cen2-mcen-b.mfc=Features/000000000.chunk[178742,178989]
 An4/770/770/cen8-mrcb-b.mfc=Features/000000000.chunk[178990,179267]
 An4/618/618/an188-mkem-b.mfc=Features/000000000.chunk[179268,179515]
 An4/543/543/an19-mjdr-b.mfc=Features/000000000.chunk[179516,179923]
 An4/597/597/cen2-mjjs2-b.mfc=Features/000000000.chunk[179924,180391]
 An4/304/304/cen5-mbmg-b.mfc=Features/000000000.chunk[180392,180779]
 An4/613/613/cen6-mkdb-b.mfc=Features/000000000.chunk[180780,180947]
 An4/551/551/cen7-mjdr-b.mfc=Features/000000000.chunk[180948,181235]
 An4/260/260/an224-fwxs-b.mfc=Features/000000000.chunk[181236,181553]
 An4/810/810/an341-msct-b.mfc=Features/000000000.chunk[181554,181651]
 An4/211/211/cen2-fsaf2-b.mfc=Features/000000000.chunk[181652,181959]
 An4/3/3/an254-fash-b.mfc=Features/000000000.chunk[181960,182047]
 An4/285/285/an4-mblw-b.mfc=Features/000000000.chunk[182048,182185]
 An4/651/651/cen6-mmal-b.mfc=Features/000000000.chunk[182186,182343]
 An4/862/862/an276-msmn-b.mfc=Features/000000000.chunk[182344,182631]
 An4/296/296/an267-mbmg-b.mfc=Features/000000000.chunk[182632,182969]
 An4/261/261/an225-fwxs-b.mfc=Features/000000000.chunk[182970,183067]
 An4/705/705/cen7-mmsh-b.mfc=Features/000000000.chunk[183068,183285]
 An4/28/28/cen1-fclc-b.mfc=Features/000000000.chunk[183286,183713]
 An4/382/382/cen6-mdcs-b.mfc=Features/000000000.chunk[183714,183891]
 An4/765/765/cen3-mrcb-b.mfc=Features/000000000.chunk[183892,184069]
 An4/499/499/cen6-mfaa-b.mfc=Features/000000000.chunk[184070,184247]
 An4/7/7/cen4-fash-b.mfc=Features/000000000.chunk[184248,184605]
 An4/110/110/cen5-fkai-b.mfc=Features/000000000.chunk[184606,184913]
 An4/893/893/cen5-mtcv-b.mfc=Features/000000000.chunk[184914,185201]
 An4/677/677/cen6-mmdg-b.mfc=Features/000000000.chunk[185202,185509]
 An4/204/204/cen8-fplp-b.mfc=Features/000000000.chunk[185510,185897]
 An4/427/427/an29-mdxs-b.mfc=Features/000000000.chunk[185898,186005]
 An4/284/284/an3-mblw-b.mfc=Features/000000000.chunk[186006,186133]
 An4/348/348/an143-mcrt-b.mfc=Features/000000000.chunk[186134,186241]
 An4/449/449/cen8-meab-b.mfc=Features/000000000.chunk[186242,186549]
 An4/423/423/cen8-mdxn-b.mfc=Features/000000000.chunk[186550,186817]
 An4/466/466/an289-mema-b.mfc=Features/000000000.chunk[186818,186985]
 An4/508/508/cen2-mgah-b.mfc=Features/000000000.chunk[186986,187243]
 An4/1/1/an251-fash-b.mfc=Features/000000000.chunk[187244,187341]
 An4/399/399/an207-mdmc-b.mfc=Features/000000000.chunk[187342,187799]
 An4/21/21/cen7-fbbh-b.mfc=Features/000000000.chunk[187800,188107]
 An4/573/573/cen4-mjgk-b.mfc=Features/000000000.chunk[188108,188525]
 An4/706/706/cen8-mmsh-b.mfc=Features/000000000.chunk[188526,188743]
 An4/609/609/cen2-mkdb-b.mfc=Features/000000000.chunk[188744,188911]
 An4/839/839/an354-msjr-b.mfc=Features/000000000.chunk[188912,189299]
 An4/312/312/an205-mcel-b.mfc=Features/000000000.chunk[189300,189657]
 An4/63/63/an77-fjam-b.mfc=Features/000000000.chunk[189658,189965]
 An4/678/678/cen7-mmdg-b.mfc=Features/000000000.chunk[189966,190503]
 An4/761/761/an104-mrcb-b.mfc=Features/000000000.chunk[190504,190601]
 An4/642/642/an322-mmal-b.mfc=Features/000000000.chunk[190602,190929]
 An4/262/262/cen1-fwxs-b.mfc=Features/000000000.chunk[190930,191347]
 An4/171/171/cen1-fmjd-b.mfc=Features/000000000.chunk[191348,191665]
 An4/114/114/an131-fkdo-b.mfc=Features/000000000.chunk[191666,192263]
 An4/853/853/an99-mskh-b.mfc=Features/000000000.chunk[192264,192521]
 An4/333/333/an261-mcfl-b.mfc=Features/000000000.chunk[192522,192619]
 An4/112/112/cen7-fkai-b.mfc=Features/000000000.chunk[192620,192777]
 An4/265/265/cen4-fwxs-b.mfc=Features/000000000.chunk[192778,193325]
 An4/813/813/an344-msct-b.mfc=Features/000000000.chunk[193326,193723]
 An4/496/496/cen3-mfaa-b.mfc=Features/000000000.chunk[193724,193881]
 An4/178/178/cen8-fmjd-b.mfc=Features/000000000.chunk[193882,194129]
 An4/54/54/cen1-ffmm-b.mfc=Features/000000000.chunk[194130,194457]
 An4/596/596/cen1-mjjs2-b.mfc=Features/000000000.chunk[194458,194895]
 An4/875/875/an335-msrb-b.mfc=Features/000000000.chunk[194896,195033]
 An4/430/430/cen2-mdxs-b.mfc=Features/000000000.chunk[195034,195281]
 An4/928/928/cen1-mtxj-b.mfc=Features/000000000.chunk[195282,195529]
 An4/493/493/an165-mfaa-b.mfc=Features/000000000.chunk[195530,195807]
 An4/887/887/an304-mtcv-b.mfc=Features/000000000.chunk[195808,195905]
 An4/859/859/cen6-mskh-b.mfc=Features/000000000.chunk[195906,196083]
 An4/883/883/cen8-msrb-b.mfc=Features/000000000.chunk[196084,196301]
 An4/871/871/an331-msrb-b.mfc=Features/000000000.chunk[196302,196479]
 An4/836/836/an351-msjr-b.mfc=Features/000000000.chunk[196480,196757]
 An4/553/553/an156-mjes-b.mfc=Features/000000000.chunk[196758,197075]
 An4/861/861/cen8-mskh-b.mfc=Features/000000000.chunk[197076,197333]
 An4/248/248/an215-ftmj-b.mfc=Features/000000000.chunk[197334,197441]
 An4/77/77/an123-fjdn-b.mfc=Features/000000000.chunk[197442,197569]
 An4/417/417/cen2-mdxn-b.mfc=Features/000000000.chunk[197570,197787]
 An4/856/856/cen3-mskh-b.mfc=Features/000000000.chunk[197788,197965]
 An4/44/44/cen4-fejs-b.mfc=Features/000000000.chunk[197966,198493]
 An4/645/645/an325-mmal-b.mfc=Features/000000000.chunk[198494,198581]
 An4/453/453/an69-meht-b.mfc=Features/000000000.chunk[198582,198669]
 An4/639/639/cen7-mmaf-b.mfc=Features/000000000.chunk[198670,198957]
 An4/907/907/cen6-mtje-b.mfc=Features/000000000.chunk[198958,199175]
 An4/330/330/cen6-mcen-b.mfc=Features/000000000.chunk[199176,199373]
 An4/258/258/an222-fwxs-b.mfc=Features/000000000.chunk[199374,199461]
 An4/746/746/an72-mrab-b.mfc=Features/000000000.chunk[199462,199549]
 An4/752/752/cen3-mrab-b.mfc=Features/000000000.chunk[199550,199677]
 An4/663/663/cen5-mmap-b.mfc=Features/000000000.chunk[199678,199945]
 An4/843/843/cen3-msjr-b.mfc=Features/000000000.chunk[199946,200083]
 An4/674/674/cen3-mmdg-b.mfc=Features/000000000.chunk[200084,200321]
 An4/105/105/an315-fkai-b.mfc=Features/000000000.chunk[200322,200519]
 An4/604/604/an317-mkdb-b.mfc=Features/000000000.chunk[200520,200657]
 An4/557/557/an160-mjes-b.mfc=Features/000000000.chunk[200658,201035]
 An4/320/320/an126-mcen-b.mfc=Features/000000000.chunk[201036,201123]
 An4/605/605/an318-mkdb-b.mfc=Features/000000000.chunk[201124,201251]
 An4/731/731/cen7-mnfe-b.mfc=Features/000000000.chunk[201252,201649]
 An4/467/467/an290-mema-b.mfc=Features/000000000.chunk[201650,201907]
 An4/368/368/cen5-mcsc-b.mfc=Features/000000000.chunk[201908,202045]
 An4/672/672/cen1-mmdg-b.mfc=Features/000000000.chunk[202046,202463]
 An4/744/744/cen8-mnjl-b.mfc=Features/000000000.chunk[202464,202721]
 An4/37/37/an37-fejs-b.mfc=Features/000000000.chunk[202722,203189]
 An4/376/376/an245-mdcs-b.mfc=Features/000000000.chunk[203190,203447]
 An4/64/64/an78-fjam-b.mfc=Features/000000000.chunk[203448,203645]
 An4/27/27/an150-fclc-b.mfc=Features/000000000.chunk[203646,204073]
 An4/917/917/cen3-mtos-b.mfc=Features/000000000.chunk[204074,204311]
 An4/637/637/cen5-mmaf-b.mfc=Features/000000000.chunk[204312,204659]
 An4/464/464/an287-mema-b.mfc=Features/000000000.chunk[204660,204907]
 An4/187/187/cen4-fnsv-b.mfc=Features/000000000.chunk[204908,205375]
 An4/385/385/an111-mdcs2-b.mfc=Features/000000000.chunk[205376,205573]
 An4/885/885/an302-mtcv-b.mfc=Features/000000000.chunk[205574,205981]
 An4/914/914/an370-mtos-b.mfc=Features/000000000.chunk[205982,206439]
 An4/153/153/an116-fmjc-b.mfc=Features/000000000.chunk[206440,206667]
 An4/375/375/an244-mdcs-b.mfc=Features/000000000.chunk[206668,206915]
 An4/868/868/cen6-msmn-b.mfc=Features/000000000.chunk[206916,207093]
 An4/495/495/cen2-mfaa-b.mfc=Features/000000000.chunk[207094,207221]
 An4/712/712/cen1-mmtm-b.mfc=Features/000000000.chunk[207222,207509]
 An4/364/364/cen1-mcsc-b.mfc=Features/000000000.chunk[207510,207987]
 An4/739/739/cen2-mnjl-b.mfc=Features/000000000.chunk[207988,208185]
 An4/256/256/cen8-ftmj-b.mfc=Features/000000000.chunk[208186,208473]
 An4/222/222/an170-fsrb-b.mfc=Features/000000000.chunk[208474,208721]
 An4/332/332/cen8-mcen-b.mfc=Features/000000000.chunk[208722,208979]
 An4/759/759/an102-mrcb-b.mfc=Features/000000000.chunk[208980,209057]
 An4/571/571/cen1-mjgk-b.mfc=Features/000000000.chunk[209058,209245]
 An4/585/585/cen3-mjhp-b.mfc=Features/000000000.chunk[209246,209413]
 An4/422/422/cen7-mdxn-b.mfc=Features/000000000.chunk[209414,209681]
 An4/50/50/an292-ffmm-b.mfc=Features/000000000.chunk[209682,209849]
 An4/483/483/cen3-mewl-b.mfc=Features/000000000.chunk[209850,210167]
 An4/104/104/an314-fkai-b.mfc=Features/000000000.chunk[210168,210535]
 An4/641/641/an321-mmal-b.mfc=Features/000000000.chunk[210536,210833]
 An4/798/798/an357-mscg2-b.mfc=Features/000000000.chunk[210834,210931]
 An4/42/42/cen2-fejs-b.mfc=Features/000000000.chunk[210932,211159]
 An4/632/632/an55-mmaf-b.mfc=Features/000000000.chunk[211160,211347]
 An4/716/716/cen5-mmtm-b.mfc=Features/000000000.chunk[211348,211585]
 An4/19/19/cen5-fbbh-b.mfc=Features/000000000.chunk[211586,211873]
 An4/923/923/an376-mtxj-b.mfc=Features/000000000.chunk[211874,212041]
 An4/890/890/cen2-mtcv-b.mfc=Features/000000000.chunk[212042,212179]
 An4/825/825/an228-msjm-b.mfc=Features/000000000.chunk[212180,212537]
 An4/379/379/cen3-mdcs-b.mfc=Features/000000000.chunk[212538,212745]
 An4/870/870/cen8-msmn-b.mfc=Features/000000000.chunk[212746,213333]
 An4/623/623/cen4-mkem-b.mfc=Features/000000000.chunk[213334,213461]
 An4/703/703/cen5-mmsh-b.mfc=Features/000000000.chunk[213462,213679]
 An4/126/126/cen8-fkdo-b.mfc=Features/000000000.chunk[213680,213937]
 An4/79/79/an125-fjdn-b.mfc=Features/000000000.chunk[213938,214045]
 An4/384/384/cen8-mdcs-b.mfc=Features/000000000.chunk[214046,214303]
 An4/681/681/an362-mmkw-b.mfc=Features/000000000.chunk[214304,214741]
 An4/913/913/an369-mtos-b.mfc=Features/000000000.chunk[214742,214839]
 An4/392/392/cen3-mdcs2-b.mfc=Features/000000000.chunk[214840,215007]
 An4/217/217/cen8-fsaf2-b.mfc=Features/000000000.chunk[215008,215205]
 An4/409/409/cen7-mdmc-b.mfc=Features/000000000.chunk[215206,215533]
 An4/515/515/an246-mjbh-b.mfc=Features/000000000.chunk[215534,215621]
 An4/90/90/an7-fjmd-b.mfc=Features/000000000.chunk[215622,215839]
 An4/760/760/an103-mrcb-b.mfc=Features/000000000.chunk[215840,216247]
 An4/62/62/an76-fjam-b.mfc=Features/000000000.chunk[216248,216335]
 An4/822/822/cen8-msct-b.mfc=Features/000000000.chunk[216336,216563]
 An4/462/462/cen8-meht-b.mfc=Features/000000000.chunk[216564,216761]
 An4/292/292/cen6-mblw-b.mfc=Features/000000000.chunk[216762,216929]
 An4/676/676/cen5-mmdg-b.mfc=Features/000000000.chunk[216930,217477]
 An4/572/572/cen2-mjgk-b.mfc=Features/000000000.chunk[217478,217695]
 An4/363/363/an235-mcsc-b.mfc=Features/000000000.chunk[217696,217773]
 An4/522/522/cen3-mjbh-b.mfc=Features/000000000.chunk[217774,217931]
 An4/924/924/an377-mtxj-b.mfc=Features/000000000.chunk[217932,218299]
 An4/816/816/cen2-msct-b.mfc=Features/000000000.chunk[218300,218547]
 An4/485/485/cen5-mewl-b.mfc=Features/000000000.chunk[218548,218915]
 An4/621/621/cen1-mkem-b.mfc=Features/000000000.chunk[218916,219183]
 An4/577/577/cen8-mjgk-b.mfc=Features/000000000.chunk[219184,219391]
 An4/318/318/cen6-mcel-b.mfc=Features/000000000.chunk[219392,219619]
 An4/792/792/cen4-mrmg-b.mfc=Features/000000000.chunk[219620,219977]
 An4/841/841/cen1-msjr-b.mfc=Features/000000000.chunk[219978,220345]
 An4/763/763/cen1-mrcb-b.mfc=Features/000000000.chunk[220346,220553]
 An4/458/458/cen4-meht-b.mfc=Features/000000000.chunk[220554,220751]
 An4/715/715/cen4-mmtm-b.mfc=Features/000000000.chunk[220752,221289]
 An4/607/607/an320-mkdb-b.mfc=Features/000000000.chunk[221290,221527]
 An4/208/208/an299-fsaf2-b.mfc=Features/000000000.chunk[221528,221925]
 An4/134/134/cen3-flmm2-b.mfc=Features/000000000.chunk[221926,222103]
 An4/649/649/cen4-mmal-b.mfc=Features/000000000.chunk[222104,222391]
 An4/911/911/an367-mtos-b.mfc=Features/000000000.chunk[222392,222799]
 An4/730/730/cen6-mnfe-b.mfc=Features/000000000.chunk[222800,223087]
 An4/349/349/an144-mcrt-b.mfc=Features/000000000.chunk[223088,223485]
 An4/324/324/an130-mcen-b.mfc=Features/000000000.chunk[223486,223553]
 An4/501/501/cen8-mfaa-b.mfc=Features/000000000.chunk[223554,223751]
 An4/226/226/cen4-fsrb-b.mfc=Features/000000000.chunk[223752,224189]
 An4/547/547/cen3-mjdr-b.mfc=Features/000000000.chunk[224190,224357]
 An4/414/414/an109-mdxn-b.mfc=Features/000000000.chunk[224358,224625]
 An4/201/201/cen5-fplp-b.mfc=Features/000000000.chunk[224626,225233]
 An4/221/221/an169-fsrb-b.mfc=Features/000000000.chunk[225234,225391]
 An4/12/12/an88-fbbh-b.mfc=Features/000000000.chunk[225392,225859]
 An4/879/879/cen4-msrb-b.mfc=Features/000000000.chunk[225860,226267]
 An4/563/563/cen6-mjes-b.mfc=Features/000000000.chunk[226268,226415]
 An4/123/123/cen5-fkdo-b.mfc=Features/000000000.chunk[226416,226813]
 An4/207/207/an298-fsaf2-b.mfc=Features/000000000.chunk[226814,226911]
 An4/617/617/an187-mkem-b.mfc=Features/000000000.chunk[226912,227079]
 An4/866/866/cen4-msmn-b.mfc=Features/000000000.chunk[227080,227517]
 An4/943/943/cen3-mwhw-b.mfc=Features/000000000.chunk[227518,227625]
 An4/542/542/an18-mjdr-b.mfc=Features/000000000.chunk[227626,227783]
 An4/762/762/an105-mrcb-b.mfc=Features/000000000.chunk[227784,227931]
 An4/465/465/an288-mema-b.mfc=Features/000000000.chunk[227932,228019]
 An4/249/249/cen1-ftmj-b.mfc=Features/000000000.chunk[228020,228257]
 An4/307/307/cen8-mbmg-b.mfc=Features/000000000.chunk[228258,228585]
 An4/802/802/cen1-mscg2-b.mfc=Features/000000000.chunk[228586,228823]
 An4/73/73/cen7-fjam-b.mfc=Features/000000000.chunk[228824,229061]
 An4/554/554/an157-mjes-b.mfc=Features/000000000.chunk[229062,229189]
 An4/539/539/cen7-mjda-b.mfc=Features/000000000.chunk[229190,229517]
 An4/505/505/an199-mgah-b.mfc=Features/000000000.chunk[229518,229615]
 An4/780/780/cen5-mrjc2-b.mfc=Features/000000000.chunk[229616,229993]
 An4/100/100/cen8-fjmd-b.mfc=Features/000000000.chunk[229994,230211]
 An4/696/696/an383-mmsh-b.mfc=Features/000000000.chunk[230212,230349]
 An4/486/486/cen6-mewl-b.mfc=Features/000000000.chunk[230350,230547]
 An4/32/32/cen5-fclc-b.mfc=Features/000000000.chunk[230548,230975]
 An4/835/835/cen8-msjm-b.mfc=Features/000000000.chunk[230976,231193]
 An4/514/514/cen8-mgah-b.mfc=Features/000000000.chunk[231194,231541]
 An4/694/694/an381-mmsh-b.mfc=Features/000000000.chunk[231542,231779]
 An4/867/867/cen5-msmn-b.mfc=Features/000000000.chunk[231780,232107]
 An4/366/366/cen3-mcsc-b.mfc=Features/000000000.chunk[232108,232335]
 An4/912/912/an368-mtos-b.mfc=Features/000000000.chunk[232336,232753]
 An4/738/738/cen1-mnjl-b.mfc=Features/000000000.chunk[232754,233161]
 An4/270/270/an372-mblb-b.mfc=Features/000000000.chunk[233162,233459]
 An4/155/155/an118-fmjc-b.mfc=Features/000000000.chunk[233460,233707]
 An4/558/558/cen1-mjes-b.mfc=Features/000000000.chunk[233708,233925]
 An4/606/606/an319-mkdb-b.mfc=Features/000000000.chunk[233926,234273]
 An4/819/819/cen5-msct-b.mfc=Features/000000000.chunk[234274,234671]
 An4/288/288/cen2-mblw-b.mfc=Features/000000000.chunk[234672,234829]
 An4/120/120/cen2-fkdo-b.mfc=Features/000000000.chunk[234830,235117]
 An4/536/536/cen4-mjda-b.mfc=Features/000000000.chunk[235118,235695]
 An4/302/302/cen3-mbmg-b.mfc=Features/000000000.chunk[235696,235843]
 An4/860/860/cen7-mskh-b.mfc=Features/000000000.chunk[235844,236141]
 An4/269/269/an371-mblb-b.mfc=Features/000000000.chunk[236142,236509]
 An4/455/455/cen1-meht-b.mfc=Features/000000000.chunk[236510,236717]
 An4/286/286/an5-mblw-b.mfc=Features/000000000.chunk[236718,236815]
 An4/136/136/cen6-flmm2-b.mfc=Features/000000000.chunk[236816,237043]
 An4/481/481/cen1-mewl-b.mfc=Features/000000000.chunk[237044,237471]
 An4/58/58/cen5-ffmm-b.mfc=Features/000000000.chunk[237472,237959]
 An4/583/583/cen1-mjhp-b.mfc=Features/000000000.chunk[237960,238337]
 An4/534/534/cen2-mjda-b.mfc=Features/000000000.chunk[238338,238555]
 An4/940/940/an155-mwhw-b.mfc=Features/000000000.chunk[238556,238693]
 An4/882/882/cen7-msrb-b.mfc=Features/000000000.chunk[238694,239061]
 An4/473/473/cen6-mema-b.mfc=Features/000000000.chunk[239062,239239]
 An4/937/937/an152-mwhw-b.mfc=Features/000000000.chunk[239240,239337]
 An4/94/94/cen2-fjmd-b.mfc=Features/000000000.chunk[239338,239615]
 An4/83/83/cen4-fjdn-b.mfc=Features/000000000.chunk[239616,239763]
 An4/568/568/an218-mjgk-b.mfc=Features/000000000.chunk[239764,239851]
 An4/45/45/cen5-fejs-b.mfc=Features/000000000.chunk[239852,240259]
 An4/766/766/cen4-mrcb-b.mfc=Features/000000000.chunk[240260,240537]
 An4/929/929/cen2-mtxj-b.mfc=Features/000000000.chunk[240538,240695]
 An4/634/634/cen2-mmaf-b.mfc=Features/000000000.chunk[240696,240953]
 An4/337/337/an265-mcfl-b.mfc=Features/000000000.chunk[240954,241051]
 An4/884/884/an301-mtcv-b.mfc=Features/000000000.chunk[241052,241429]
 An4/516/516/an247-mjbh-b.mfc=Features/000000000.chunk[241430,241507]
 An4/796/796/cen8-mrmg-b.mfc=Features/000000000.chunk[241508,241725]
 An4/397/397/cen8-mdcs2-b.mfc=Features/000000000.chunk[241726,241973]
 An4/648/648/cen3-mmal-b.mfc=Features/000000000.chunk[241974,242151]
 An4/81/81/cen2-fjdn-b.mfc=Features/000000000.chunk[242152,242329]
 An4/807/807/cen6-mscg2-b.mfc=Features/000000000.chunk[242330,242617]
 An4/717/717/cen6-mmtm-b.mfc=Features/000000000.chunk[242618,242845]
 An4/394/394/cen5-mdcs2-b.mfc=Features/000000000.chunk[242846,243113]
 An4/895/895/cen7-mtcv-b.mfc=Features/000000000.chunk[243114,243461]
 An4/140/140/an2121-flrp-b.mfc=Features/000000000.chunk[243462,243779]
 An4/653/653/cen8-mmal-b.mfc=Features/000000000.chunk[243780,243957]
 An4/355/355/cen5-mcrt-b.mfc=Features/000000000.chunk[243958,244555]
 An4/159/159/cen2-fmjc-b.mfc=Features/000000000.chunk[244556,244803]
 An4/443/443/cen2-meab-b.mfc=Features/000000000.chunk[244804,245111]
 An4/942/942/cen2-mwhw-b.mfc=Features/000000000.chunk[245112,245329]
 An4/809/809/cen8-mscg2-b.mfc=Features/000000000.chunk[245330,245637]
 An4/519/519/an250-mjbh-b.mfc=Features/000000000.chunk[245638,245725]
 An4/944/944/cen4-mwhw-b.mfc=Features/000000000.chunk[245726,246083]
 An4/190/190/cen7-fnsv-b.mfc=Features/000000000.chunk[246084,246471]
 An4/925/925/an378-mtxj-b.mfc=Features/000000000.chunk[246472,246619]
 An4/665/665/cen7-mmap-b.mfc=Features/000000000.chunk[246620,246907]
 An4/448/448/cen7-meab-b.mfc=Features/000000000.chunk[246908,247345]
 An4/845/845/cen5-msjr-b.mfc=Features/000000000.chunk[247346,247563]
 An4/818/818/cen4-msct-b.mfc=Features/000000000.chunk[247564,247821]
 An4/695/695/an382-mmsh-b.mfc=Features/000000000.chunk[247822,248089]
 An4/511/511/cen5-mgah-b.mfc=Features/000000000.chunk[248090,248567]
 An4/479/479/an259-mewl-b.mfc=Features/000000000.chunk[248568,248705]
 An4/35/35/cen8-fclc-b.mfc=Features/000000000.chunk[248706,248973]
 An4/109/109/cen4-fkai-b.mfc=Features/000000000.chunk[248974,249221]
 An4/14/14/an90-fbbh-b.mfc=Features/000000000.chunk[249222,249319]
 An4/586/586/cen4-mjhp-b.mfc=Features/000000000.chunk[249320,249647]
 An4/133/133/cen2-flmm2-b.mfc=Features/000000000.chunk[249648,249845]
 An4/30/30/cen3-fclc-b.mfc=Features/000000000.chunk[249846,250033]
 An4/383/383/cen7-mdcs-b.mfc=Features/000000000.chunk[250034,250381]
 An4/34/34/cen7-fclc-b.mfc=Features/000000000.chunk[250382,250679]
 An4/851/851/an97-mskh-b.mfc=Features/000000000.chunk[250680,250817]
 An4/147/147/cen3-flrp-b.mfc=Features/000000000.chunk[250818,250975]
 An4/550/550/cen6-mjdr-b.mfc=Features/000000000.chunk[250976,251143]
 An4/407/407/cen5-mdmc-b.mfc=Features/000000000.chunk[251144,251521]
 An4/587/587/cen5-mjhp-b.mfc=Features/000000000.chunk[251522,251799]
 An4/22/22/cen8-fbbh-b.mfc=Features/000000000.chunk[251800,252077]
 An4/138/138/cen8-flmm2-b.mfc=Features/000000000.chunk[252078,252655]
 An4/361/361/an233-mcsc-b.mfc=Features/000000000.chunk[252656,252733]
--- a/Tests/Speech/Data/state.list
+++ b/Tests/Speech/Data/state.list
@ -0,0 +1,132 @@
 _ah_[2]
 _ah_[3]
 _ah_[4]
 _hmm_[2]
 _hmm_[3]
 _hmm_[4]
 _noise_[2]
 _noise_[3]
 _noise_[4]
 aa_s2_1
 aa_s3_1
 aa_s4_1
 ae_s2_1
 ae_s3_1
 ae_s4_1
 ah_s2_1
 ah_s3_1
 ah_s4_1
 ao_s2_1
 ao_s3_1
 ao_s4_1
 aw_s2_1
 aw_s3_1
 aw_s4_1
 ax_s2_1
 ax_s3_1
 ax_s4_1
 ay_s2_1
 ay_s3_1
 ay_s4_1
 b_s2_1
 b_s3_1
 b_s4_1
 ch_s2_1
 ch_s3_1
 ch_s4_1
 d_s2_1
 d_s3_1
 d_s4_1
 dh_s2_1
 dh_s3_1
 dh_s4_1
 eh_s2_1
 eh_s3_1
 eh_s4_1
 er_s2_1
 er_s3_1
 er_s4_1
 ey_s2_1
 ey_s3_1
 ey_s4_1
 f_s2_1
 f_s3_1
 f_s4_1
 g_s2_1
 g_s3_1
 g_s4_1
 hh_s2_1
 hh_s3_1
 hh_s4_1
 ih_s2_1
 ih_s3_1
 ih_s4_1
 iy_s2_1
 iy_s3_1
 iy_s4_1
 jh_s2_1
 jh_s3_1
 jh_s4_1
 k_s2_1
 k_s3_1
 k_s4_1
 l_s2_1
 l_s3_1
 l_s4_1
 m_s2_1
 m_s3_1
 m_s4_1
 n_s2_1
 n_s3_1
 n_s4_1
 ng_s2_1
 ng_s3_1
 ng_s4_1
 ow_s2_1
 ow_s3_1
 ow_s4_1
 oy_s2_1
 oy_s3_1
 oy_s4_1
 p_s2_1
 p_s3_1
 p_s4_1
 r_s2_1
 r_s3_1
 r_s4_1
 s_s2_1
 s_s3_1
 s_s4_1
 sh_s2_1
 sh_s3_1
 sh_s4_1
 sil[2]
 sil[3]
 sil[4]
 t_s2_1
 t_s3_1
 t_s4_1
 th_s2_1
 th_s3_1
 th_s4_1
 uh_s2_1
 uh_s3_1
 uh_s4_1
 uw_s2_1
 uw_s3_1
 uw_s4_1
 v_s2_1
 v_s3_1
 v_s4_1
 w_s2_1
 w_s3_1
 w_s4_1
 y_s2_1
 y_s3_1
 y_s4_1
 z_s2_1
 z_s3_1
 z_s4_1
 zh_s2_1
 zh_s3_1
 zh_s4_1
--- a/Tests/Speech/QuickE2E/baseline.cpu.txt
+++ b/Tests/Speech/QuickE2E/baseline.cpu.txt
@ -0,0 +1,758 @@
 === Running /home/vlivan/cntk/bin/x86_64.gpu.release.acml/cntk configFile=/home/vlivan/cntk/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu DataDir=/home/vlivan/cntk/Tests/Speech/Data DeviceId=Auto
 running on localhost at 2015/07/29 19:11:01
 command line options: 
 configFile=/home/vlivan/cntk/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu DataDir=/home/vlivan/cntk/Tests/Speech/Data DeviceId=Auto 
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=$DeviceId$
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=$RunDir$/models/cntkSpeech.dnn
    deviceId=$DeviceId$
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=$DataDir$/glob_0000.mlf
          labelMappingFile=$DataDir$/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=Auto
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 configparameters: cntk.config:command=speechTrain
 configparameters: cntk.config:DataDir=/home/vlivan/cntk/Tests/Speech/Data
 configparameters: cntk.config:deviceId=Auto
 configparameters: cntk.config:parallelTrain=false
 configparameters: cntk.config:precision=float
 configparameters: cntk.config:RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu
 configparameters: cntk.config:speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 <<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 command: speechTrain 
 precision = float
 lsof: WARNING: can't stat() ext4 file system /var/lib/docker/aufs
      Output information may be incomplete.
 LockDevice: Capture device 0 and lock it for exclusive use
 LockDevice: Capture device 0 and lock it for exclusive use
 SimpleNetworkBuilder Using GPU 0
 reading script file glob_0000.scp ... 948 entries
 total 132 state names in state list /home/vlivan/cntk/Tests/Speech/Data/state.list
 htkmlfreader: reading MLF file /home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf ...parse the line 55130
 total 948 entries
 ...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
 label set 0: 129 classes
 minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
 GetTrainCriterionNodes  ...
 GetEvalCriterionNodes  ...
 Validating node CrossEntropyWithSoftmax 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 3])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 3])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 3], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 3])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 3], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 3])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 3])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 3], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 3])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 3])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 3], B2[132, 1])
 Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax(labels[132, 3], HLast[132, 3])
 Found 3 PreCompute nodes
 	NodeName: InvStdOfFeatures
 	NodeName: MeanOfFeatures
 	NodeName: Prior
 minibatchiterator: epoch 0: frames [0..252734] (first utterance at frame 0) with 1 datapasses
 requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms
 Validating node InvStdOfFeatures 
 Validating --> features = InputValue
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 64])
 Validating node MeanOfFeatures 
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 64])
 Validating node Prior 
 Validating --> labels = InputValue
 Validating --> Prior = Mean(labels[132, 64])
 Set Max Temp Mem Size For Convolution Nodes to 0 samples.
 Starting Epoch 1: learning rate per sample = 0.015625  momentum = 0.900000 
 minibatchiterator: epoch 0: frames [0..20480] (first utterance at frame 0) with 1 datapasses
 Validating node EvalErrorPrediction 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 64])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 64])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 64], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 64])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 64], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 64])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 64])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 64], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 64])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 64])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 64], B2[132, 1])
 Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 64], HLast[132, 64])
 Epoch[1 of 3]-Minibatch[1-10 of 320]: SamplesSeen = 640; TrainLossPerSample = 4.3213539; EvalErr[0]PerSample = 0.89999998; TotalTime=0.064177; TotalTimePerSample=0.00010027656, SamplesPerSecond=9972
 Epoch[1 of 3]-Minibatch[11-20 of 320]: SamplesSeen = 640; TrainLossPerSample = 4.1507101; EvalErr[0]PerSample = 0.8671875; TotalTime=0.060664; TotalTimePerSample=9.47875e-05, SamplesPerSecond=10549
 Epoch[1 of 3]-Minibatch[21-30 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.9990096; EvalErr[0]PerSample = 0.87656248; TotalTime=0.062395; TotalTimePerSample=9.7492187e-05, SamplesPerSecond=10257
 Epoch[1 of 3]-Minibatch[31-40 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.8694596; EvalErr[0]PerSample = 0.87656248; TotalTime=0.058102; TotalTimePerSample=9.0784375e-05, SamplesPerSecond=11015
 Epoch[1 of 3]-Minibatch[41-50 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.8021927; EvalErr[0]PerSample = 0.87812501; TotalTime=0.058272; TotalTimePerSample=9.105e-05, SamplesPerSecond=10982
 Epoch[1 of 3]-Minibatch[51-60 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.7289093; EvalErr[0]PerSample = 0.86874998; TotalTime=0.056752; TotalTimePerSample=8.8675e-05, SamplesPerSecond=11277
 Epoch[1 of 3]-Minibatch[61-70 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.5618699; EvalErr[0]PerSample = 0.82343751; TotalTime=0.06015; TotalTimePerSample=9.3984375e-05, SamplesPerSecond=10640
 Epoch[1 of 3]-Minibatch[71-80 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.4279053; EvalErr[0]PerSample = 0.80781251; TotalTime=0.061573; TotalTimePerSample=9.6207812e-05, SamplesPerSecond=10394
 Epoch[1 of 3]-Minibatch[81-90 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.3392854; EvalErr[0]PerSample = 0.7734375; TotalTime=0.057831; TotalTimePerSample=9.0360938e-05, SamplesPerSecond=11066
 Epoch[1 of 3]-Minibatch[91-100 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.3639894; EvalErr[0]PerSample = 0.84375; TotalTime=0.05709; TotalTimePerSample=8.9203125e-05, SamplesPerSecond=11210
 WARNING: The same matrix with dim [1, 1] has been transferred between different devices for 20 times.
 Epoch[1 of 3]-Minibatch[101-110 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.2122345; EvalErr[0]PerSample = 0.75312501; TotalTime=0.061065; TotalTimePerSample=9.5414062e-05, SamplesPerSecond=10480
 Epoch[1 of 3]-Minibatch[111-120 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.3126526; EvalErr[0]PerSample = 0.78750002; TotalTime=0.058543; TotalTimePerSample=9.1473437e-05, SamplesPerSecond=10932
 Epoch[1 of 3]-Minibatch[121-130 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.1408203; EvalErr[0]PerSample = 0.74687499; TotalTime=0.0605; TotalTimePerSample=9.453125e-05, SamplesPerSecond=10578
 Epoch[1 of 3]-Minibatch[131-140 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.006897; EvalErr[0]PerSample = 0.69687498; TotalTime=0.054623; TotalTimePerSample=8.5348438e-05, SamplesPerSecond=11716
 Epoch[1 of 3]-Minibatch[141-150 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.0049591; EvalErr[0]PerSample = 0.72343749; TotalTime=0.059955; TotalTimePerSample=9.3679687e-05, SamplesPerSecond=10674
 Epoch[1 of 3]-Minibatch[151-160 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.9785829; EvalErr[0]PerSample = 0.73906249; TotalTime=0.060773; TotalTimePerSample=9.4957812e-05, SamplesPerSecond=10530
 Epoch[1 of 3]-Minibatch[161-170 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.8568604; EvalErr[0]PerSample = 0.70781249; TotalTime=0.060235; TotalTimePerSample=9.4117187e-05, SamplesPerSecond=10625
 Epoch[1 of 3]-Minibatch[171-180 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.6905334; EvalErr[0]PerSample = 0.671875; TotalTime=0.064974; TotalTimePerSample=0.00010152188, SamplesPerSecond=9850
 Epoch[1 of 3]-Minibatch[181-190 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.7865357; EvalErr[0]PerSample = 0.70468748; TotalTime=0.05438; TotalTimePerSample=8.496875e-05, SamplesPerSecond=11769
 Epoch[1 of 3]-Minibatch[191-200 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.5770202; EvalErr[0]PerSample = 0.6484375; TotalTime=0.063006; TotalTimePerSample=9.8446875e-05, SamplesPerSecond=10157
 Epoch[1 of 3]-Minibatch[201-210 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.6157165; EvalErr[0]PerSample = 0.6640625; TotalTime=0.058268; TotalTimePerSample=9.104375e-05, SamplesPerSecond=10983
 Epoch[1 of 3]-Minibatch[211-220 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.552362; EvalErr[0]PerSample = 0.65781248; TotalTime=0.059349; TotalTimePerSample=9.2732812e-05, SamplesPerSecond=10783
 Epoch[1 of 3]-Minibatch[221-230 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.4821167; EvalErr[0]PerSample = 0.625; TotalTime=0.061069; TotalTimePerSample=9.5420313e-05, SamplesPerSecond=10479
 Epoch[1 of 3]-Minibatch[231-240 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.3877869; EvalErr[0]PerSample = 0.62812501; TotalTime=0.055723; TotalTimePerSample=8.7067188e-05, SamplesPerSecond=11485
 Epoch[1 of 3]-Minibatch[241-250 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.3690064; EvalErr[0]PerSample = 0.6484375; TotalTime=0.061959; TotalTimePerSample=9.6810937e-05, SamplesPerSecond=10329
 Epoch[1 of 3]-Minibatch[251-260 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.4396729; EvalErr[0]PerSample = 0.6328125; TotalTime=0.062976; TotalTimePerSample=9.84e-05, SamplesPerSecond=10162
 Epoch[1 of 3]-Minibatch[261-270 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.3028197; EvalErr[0]PerSample = 0.61250001; TotalTime=0.060925; TotalTimePerSample=9.5195312e-05, SamplesPerSecond=10504
 Epoch[1 of 3]-Minibatch[271-280 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.1966858; EvalErr[0]PerSample = 0.55937499; TotalTime=0.060799; TotalTimePerSample=9.4998438e-05, SamplesPerSecond=10526
 Epoch[1 of 3]-Minibatch[281-290 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.2898011; EvalErr[0]PerSample = 0.60468751; TotalTime=0.055702; TotalTimePerSample=8.7034375e-05, SamplesPerSecond=11489
 Epoch[1 of 3]-Minibatch[291-300 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.1775086; EvalErr[0]PerSample = 0.62187499; TotalTime=0.061515; TotalTimePerSample=9.6117187e-05, SamplesPerSecond=10403
 Epoch[1 of 3]-Minibatch[301-310 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.2626343; EvalErr[0]PerSample = 0.59687501; TotalTime=0.059247; TotalTimePerSample=9.2573438e-05, SamplesPerSecond=10802
 Epoch[1 of 3]-Minibatch[311-320 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.1507263; EvalErr[0]PerSample = 0.5625; TotalTime=0.059464; TotalTimePerSample=9.29125e-05, SamplesPerSecond=10762
 Finished Epoch[1]: [Training Set] TrainLossPerSample = 2.9799569; EvalErrPerSample = 0.72216797; Ave LearnRatePerSample = 0.015625; EpochTime=1.913549
 Starting Epoch 2: learning rate per sample = 0.001953  momentum = 0.656119 
 minibatchiterator: epoch 1: frames [20480..40960] (first utterance at frame 20480) with 1 datapasses
 Epoch[2 of 3]-Minibatch[1-10 of 80]: SamplesSeen = 2560; TrainLossPerSample = 2.0159853; EvalErr[0]PerSample = 0.54140627; TotalTime=0.100302; TotalTimePerSample=3.9180469e-05, SamplesPerSecond=25522
 Epoch[2 of 3]-Minibatch[11-20 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9881856; EvalErr[0]PerSample = 0.54296875; TotalTime=0.093995; TotalTimePerSample=3.6716797e-05, SamplesPerSecond=27235
 Epoch[2 of 3]-Minibatch[21-30 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9869812; EvalErr[0]PerSample = 0.54140627; TotalTime=0.09237; TotalTimePerSample=3.6082031e-05, SamplesPerSecond=27714
 Epoch[2 of 3]-Minibatch[31-40 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9312614; EvalErr[0]PerSample = 0.5277344; TotalTime=0.092894; TotalTimePerSample=3.6286719e-05, SamplesPerSecond=27558
 Epoch[2 of 3]-Minibatch[41-50 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9006774; EvalErr[0]PerSample = 0.52656251; TotalTime=0.08927; TotalTimePerSample=3.4871094e-05, SamplesPerSecond=28677
 Epoch[2 of 3]-Minibatch[51-60 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9711578; EvalErr[0]PerSample = 0.54140627; TotalTime=0.091869; TotalTimePerSample=3.5886328e-05, SamplesPerSecond=27865
 Epoch[2 of 3]-Minibatch[61-70 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.8951813; EvalErr[0]PerSample = 0.52031249; TotalTime=0.092242; TotalTimePerSample=3.6032031e-05, SamplesPerSecond=27753
 Epoch[2 of 3]-Minibatch[71-80 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.904506; EvalErr[0]PerSample = 0.53164065; TotalTime=0.094062; TotalTimePerSample=3.6742969e-05, SamplesPerSecond=27216
 Finished Epoch[2]: [Training Set] TrainLossPerSample = 1.949242; EvalErrPerSample = 0.53417969; Ave LearnRatePerSample = 0.001953125; EpochTime=0.747962
 Starting Epoch 3: learning rate per sample = 0.000098  momentum = 0.656119 
 minibatchiterator: epoch 2: frames [40960..61440] (first utterance at frame 40960) with 1 datapasses
 Epoch[3 of 3]-Minibatch[1-10 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8735985; EvalErr[0]PerSample = 0.51933593; TotalTime=0.27124; TotalTimePerSample=2.6488281e-05, SamplesPerSecond=37752
 Epoch[3 of 3]-Minibatch[11-20 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8665626; EvalErr[0]PerSample = 0.51748049; TotalTime=0.266098; TotalTimePerSample=2.5986133e-05, SamplesPerSecond=38482
 Finished Epoch[3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; Ave LearnRatePerSample = 9.765625146e-05; EpochTime=0.539342
 COMPLETED
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 Linked to libnvidia-ml library at wrong path : /usr/src/gdk/nvml/lib/libnvidia-ml.so.1
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 === Deleting last epoch data
 ==== Re-running from checkpoint
 running on localhost at 2015/07/29 19:11:07
 command line options: 
 configFile=/home/vlivan/cntk/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu DataDir=/home/vlivan/cntk/Tests/Speech/Data DeviceId=Auto 
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=$DeviceId$
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=$RunDir$/models/cntkSpeech.dnn
    deviceId=$DeviceId$
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=$DataDir$/glob_0000.mlf
          labelMappingFile=$DataDir$/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=Auto
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 configparameters: cntk.config:command=speechTrain
 configparameters: cntk.config:DataDir=/home/vlivan/cntk/Tests/Speech/Data
 configparameters: cntk.config:deviceId=Auto
 configparameters: cntk.config:parallelTrain=false
 configparameters: cntk.config:precision=float
 configparameters: cntk.config:RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu
 configparameters: cntk.config:speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 <<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 command: speechTrain 
 precision = float
 lsof: WARNING: can't stat() ext4 file system /var/lib/docker/aufs
      Output information may be incomplete.
 LockDevice: Capture device 0 and lock it for exclusive use
 LockDevice: Capture device 0 and lock it for exclusive use
 SimpleNetworkBuilder Using GPU 0
 reading script file glob_0000.scp ... 948 entries
 total 132 state names in state list /home/vlivan/cntk/Tests/Speech/Data/state.list
 htkmlfreader: reading MLF file /home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf ...parse the line 55130
 total 948 entries
 ...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
 label set 0: 129 classes
 minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
 Starting from checkpoint. Load Network From File /tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_cpu/models/cntkSpeech.dnn.2.
 Printing Gradient Computation Node Order ... 
 CrossEntropyWithSoftmax[0, 0] = CrossEntropyWithSoftmax(labels[132, 256], HLast[0, 0])
 HLast[0, 0] = Plus(W2*H1[0, 0], B2[132, 1])
 B2[132, 1] = LearnableParameter
 W2*H1[0, 0] = Times(W2[132, 512], H2[0, 0])
 H2[0, 0] = Sigmoid(W1*H1+B1[0, 0])
 W1*H1+B1[0, 0] = Plus(W1*H1[0, 0], B1[512, 1])
 B1[512, 1] = LearnableParameter
 W1*H1[0, 0] = Times(W1[512, 512], H1[0, 0])
 H1[0, 0] = Sigmoid(W0*features+B0[0, 0])
 W0*features+B0[0, 0] = Plus(W0*features[0, 0], B0[512, 1])
 B0[512, 1] = LearnableParameter
 W0*features[0, 0] = Times(W0[512, 363], MVNormalizedFeatures[0, 0])
 MVNormalizedFeatures[0, 0] = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 InvStdOfFeatures[363, 1] = InvStdDev(features[363, 256])
 MeanOfFeatures[363, 1] = Mean(features[363, 256])
 features[363, 256] = InputValue
 W0[512, 363] = LearnableParameter
 W1[512, 512] = LearnableParameter
 W2[132, 512] = LearnableParameter
 labels[132, 256] = InputValue
 Validating node CrossEntropyWithSoftmax 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax(labels[132, 256], HLast[132, 256])
 Validating node ScaledLogLikelihood 
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> labels = InputValue
 Validating --> Prior = Mean(labels[132, 256])
 Validating --> LogOfPrior = Log(Prior[132, 1])
 Validating --> ScaledLogLikelihood = Minus(HLast[132, 256], LogOfPrior[132, 1])
 Validating node EvalErrorPrediction 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 256], HLast[132, 256])
 GetTrainCriterionNodes  ...
 GetEvalCriterionNodes  ...
 Validating node CrossEntropyWithSoftmax 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax(labels[132, 256], HLast[132, 256])
 No PreCompute nodes found, skipping PreCompute step
 Set Max Temp Mem Size For Convolution Nodes to 0 samples.
 Starting Epoch 3: learning rate per sample = 0.000098  momentum = 0.656119 
 minibatchiterator: epoch 2: frames [40960..61440] (first utterance at frame 40960) with 1 datapasses
 requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms
 Validating node EvalErrorPrediction 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 1024])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 1024])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 1024], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 1024])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 1024], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 1024])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 1024])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 1024], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 1024])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 1024])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 1024], B2[132, 1])
 Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 1024], HLast[132, 1024])
 Epoch[3 of 3]-Minibatch[1-10 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8735985; EvalErr[0]PerSample = 0.51933593; TotalTime=0.390092; TotalTimePerSample=3.8094922e-05, SamplesPerSecond=26250
 Epoch[3 of 3]-Minibatch[11-20 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8665626; EvalErr[0]PerSample = 0.51748049; TotalTime=0.261875; TotalTimePerSample=2.557373e-05, SamplesPerSecond=39102
 Finished Epoch[3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; Ave LearnRatePerSample = 9.765625146e-05; EpochTime=0.770276
 COMPLETED
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 Linked to libnvidia-ml library at wrong path : /usr/src/gdk/nvml/lib/libnvidia-ml.so.1
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
--- a/Tests/Speech/QuickE2E/baseline.gpu.txt
+++ b/Tests/Speech/QuickE2E/baseline.gpu.txt
@ -0,0 +1,758 @@
 === Running /home/vlivan/cntk/bin/x86_64.gpu.release.acml/cntk configFile=/home/vlivan/cntk/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu DataDir=/home/vlivan/cntk/Tests/Speech/Data DeviceId=Auto
 running on localhost at 2015/07/29 19:11:08
 command line options: 
 configFile=/home/vlivan/cntk/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu DataDir=/home/vlivan/cntk/Tests/Speech/Data DeviceId=Auto 
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=$DeviceId$
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=$RunDir$/models/cntkSpeech.dnn
    deviceId=$DeviceId$
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=$DataDir$/glob_0000.mlf
          labelMappingFile=$DataDir$/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=Auto
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 configparameters: cntk.config:command=speechTrain
 configparameters: cntk.config:DataDir=/home/vlivan/cntk/Tests/Speech/Data
 configparameters: cntk.config:deviceId=Auto
 configparameters: cntk.config:parallelTrain=false
 configparameters: cntk.config:precision=float
 configparameters: cntk.config:RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu
 configparameters: cntk.config:speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 <<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 command: speechTrain 
 precision = float
 lsof: WARNING: can't stat() ext4 file system /var/lib/docker/aufs
      Output information may be incomplete.
 LockDevice: Capture device 0 and lock it for exclusive use
 LockDevice: Capture device 0 and lock it for exclusive use
 SimpleNetworkBuilder Using GPU 0
 reading script file glob_0000.scp ... 948 entries
 total 132 state names in state list /home/vlivan/cntk/Tests/Speech/Data/state.list
 htkmlfreader: reading MLF file /home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf ...parse the line 55130
 total 948 entries
 ...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
 label set 0: 129 classes
 minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
 GetTrainCriterionNodes  ...
 GetEvalCriterionNodes  ...
 Validating node CrossEntropyWithSoftmax 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 3])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 3])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 3], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 3])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 3], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 3])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 3])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 3], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 3])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 3])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 3], B2[132, 1])
 Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax(labels[132, 3], HLast[132, 3])
 Found 3 PreCompute nodes
 	NodeName: InvStdOfFeatures
 	NodeName: MeanOfFeatures
 	NodeName: Prior
 minibatchiterator: epoch 0: frames [0..252734] (first utterance at frame 0) with 1 datapasses
 requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms
 Validating node InvStdOfFeatures 
 Validating --> features = InputValue
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 64])
 Validating node MeanOfFeatures 
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 64])
 Validating node Prior 
 Validating --> labels = InputValue
 Validating --> Prior = Mean(labels[132, 64])
 Set Max Temp Mem Size For Convolution Nodes to 0 samples.
 Starting Epoch 1: learning rate per sample = 0.015625  momentum = 0.900000 
 minibatchiterator: epoch 0: frames [0..20480] (first utterance at frame 0) with 1 datapasses
 Validating node EvalErrorPrediction 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 64])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 64])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 64], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 64])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 64], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 64])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 64])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 64], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 64])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 64])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 64], B2[132, 1])
 Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 64], HLast[132, 64])
 Epoch[1 of 3]-Minibatch[1-10 of 320]: SamplesSeen = 640; TrainLossPerSample = 4.3213539; EvalErr[0]PerSample = 0.89999998; TotalTime=0.064294; TotalTimePerSample=0.00010045938, SamplesPerSecond=9954
 Epoch[1 of 3]-Minibatch[11-20 of 320]: SamplesSeen = 640; TrainLossPerSample = 4.1507101; EvalErr[0]PerSample = 0.8671875; TotalTime=0.055813; TotalTimePerSample=8.7207812e-05, SamplesPerSecond=11466
 Epoch[1 of 3]-Minibatch[21-30 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.9990096; EvalErr[0]PerSample = 0.87656248; TotalTime=0.062703; TotalTimePerSample=9.7973437e-05, SamplesPerSecond=10206
 Epoch[1 of 3]-Minibatch[31-40 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.8694596; EvalErr[0]PerSample = 0.87656248; TotalTime=0.059923; TotalTimePerSample=9.3629687e-05, SamplesPerSecond=10680
 Epoch[1 of 3]-Minibatch[41-50 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.8021927; EvalErr[0]PerSample = 0.87812501; TotalTime=0.061061; TotalTimePerSample=9.5407812e-05, SamplesPerSecond=10481
 Epoch[1 of 3]-Minibatch[51-60 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.7289093; EvalErr[0]PerSample = 0.86874998; TotalTime=0.062101; TotalTimePerSample=9.7032813e-05, SamplesPerSecond=10305
 Epoch[1 of 3]-Minibatch[61-70 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.5618699; EvalErr[0]PerSample = 0.82343751; TotalTime=0.056094; TotalTimePerSample=8.7646875e-05, SamplesPerSecond=11409
 Epoch[1 of 3]-Minibatch[71-80 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.4279053; EvalErr[0]PerSample = 0.80781251; TotalTime=0.063459; TotalTimePerSample=9.9154687e-05, SamplesPerSecond=10085
 Epoch[1 of 3]-Minibatch[81-90 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.3392854; EvalErr[0]PerSample = 0.7734375; TotalTime=0.062265; TotalTimePerSample=9.7289063e-05, SamplesPerSecond=10278
 Epoch[1 of 3]-Minibatch[91-100 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.3639894; EvalErr[0]PerSample = 0.84375; TotalTime=0.059843; TotalTimePerSample=9.3504687e-05, SamplesPerSecond=10694
 WARNING: The same matrix with dim [1, 1] has been transferred between different devices for 20 times.
 Epoch[1 of 3]-Minibatch[101-110 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.2122345; EvalErr[0]PerSample = 0.75312501; TotalTime=0.062375; TotalTimePerSample=9.7460937e-05, SamplesPerSecond=10260
 Epoch[1 of 3]-Minibatch[111-120 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.3126526; EvalErr[0]PerSample = 0.78750002; TotalTime=0.061085; TotalTimePerSample=9.5445313e-05, SamplesPerSecond=10477
 Epoch[1 of 3]-Minibatch[121-130 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.1408203; EvalErr[0]PerSample = 0.74687499; TotalTime=0.064562; TotalTimePerSample=0.00010087812, SamplesPerSecond=9912
 Epoch[1 of 3]-Minibatch[131-140 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.006897; EvalErr[0]PerSample = 0.69687498; TotalTime=0.0575; TotalTimePerSample=8.984375e-05, SamplesPerSecond=11130
 Epoch[1 of 3]-Minibatch[141-150 of 320]: SamplesSeen = 640; TrainLossPerSample = 3.0049591; EvalErr[0]PerSample = 0.72343749; TotalTime=0.058338; TotalTimePerSample=9.1153125e-05, SamplesPerSecond=10970
 Epoch[1 of 3]-Minibatch[151-160 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.9785829; EvalErr[0]PerSample = 0.73906249; TotalTime=0.064603; TotalTimePerSample=0.00010094219, SamplesPerSecond=9906
 Epoch[1 of 3]-Minibatch[161-170 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.8568604; EvalErr[0]PerSample = 0.70781249; TotalTime=0.060368; TotalTimePerSample=9.4325e-05, SamplesPerSecond=10601
 Epoch[1 of 3]-Minibatch[171-180 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.6905334; EvalErr[0]PerSample = 0.671875; TotalTime=0.059125; TotalTimePerSample=9.2382812e-05, SamplesPerSecond=10824
 Epoch[1 of 3]-Minibatch[181-190 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.7865357; EvalErr[0]PerSample = 0.70468748; TotalTime=0.056113; TotalTimePerSample=8.7676563e-05, SamplesPerSecond=11405
 Epoch[1 of 3]-Minibatch[191-200 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.5770202; EvalErr[0]PerSample = 0.6484375; TotalTime=0.060745; TotalTimePerSample=9.4914062e-05, SamplesPerSecond=10535
 Epoch[1 of 3]-Minibatch[201-210 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.6157165; EvalErr[0]PerSample = 0.6640625; TotalTime=0.059709; TotalTimePerSample=9.3295312e-05, SamplesPerSecond=10718
 Epoch[1 of 3]-Minibatch[211-220 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.552362; EvalErr[0]PerSample = 0.65781248; TotalTime=0.061917; TotalTimePerSample=9.6745313e-05, SamplesPerSecond=10336
 Epoch[1 of 3]-Minibatch[221-230 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.4821167; EvalErr[0]PerSample = 0.625; TotalTime=0.053813; TotalTimePerSample=8.4082813e-05, SamplesPerSecond=11893
 Epoch[1 of 3]-Minibatch[231-240 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.3877869; EvalErr[0]PerSample = 0.62812501; TotalTime=0.061932; TotalTimePerSample=9.676875e-05, SamplesPerSecond=10333
 Epoch[1 of 3]-Minibatch[241-250 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.3690064; EvalErr[0]PerSample = 0.6484375; TotalTime=0.059294; TotalTimePerSample=9.2646875e-05, SamplesPerSecond=10793
 Epoch[1 of 3]-Minibatch[251-260 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.4396729; EvalErr[0]PerSample = 0.6328125; TotalTime=0.060513; TotalTimePerSample=9.4551562e-05, SamplesPerSecond=10576
 Epoch[1 of 3]-Minibatch[261-270 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.3028197; EvalErr[0]PerSample = 0.61250001; TotalTime=0.06037; TotalTimePerSample=9.4328125e-05, SamplesPerSecond=10601
 Epoch[1 of 3]-Minibatch[271-280 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.1966858; EvalErr[0]PerSample = 0.55937499; TotalTime=0.056485; TotalTimePerSample=8.8257812e-05, SamplesPerSecond=11330
 Epoch[1 of 3]-Minibatch[281-290 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.2898011; EvalErr[0]PerSample = 0.60468751; TotalTime=0.059356; TotalTimePerSample=9.274375e-05, SamplesPerSecond=10782
 Epoch[1 of 3]-Minibatch[291-300 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.1775086; EvalErr[0]PerSample = 0.62187499; TotalTime=0.059501; TotalTimePerSample=9.2970312e-05, SamplesPerSecond=10756
 Epoch[1 of 3]-Minibatch[301-310 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.2626343; EvalErr[0]PerSample = 0.59687501; TotalTime=0.064342; TotalTimePerSample=0.00010053437, SamplesPerSecond=9946
 Epoch[1 of 3]-Minibatch[311-320 of 320]: SamplesSeen = 640; TrainLossPerSample = 2.1507263; EvalErr[0]PerSample = 0.5625; TotalTime=0.064522; TotalTimePerSample=0.00010081563, SamplesPerSecond=9919
 Finished Epoch[1]: [Training Set] TrainLossPerSample = 2.9799569; EvalErrPerSample = 0.72216797; Ave LearnRatePerSample = 0.015625; EpochTime=1.935613
 Starting Epoch 2: learning rate per sample = 0.001953  momentum = 0.656119 
 minibatchiterator: epoch 1: frames [20480..40960] (first utterance at frame 20480) with 1 datapasses
 Epoch[2 of 3]-Minibatch[1-10 of 80]: SamplesSeen = 2560; TrainLossPerSample = 2.0159853; EvalErr[0]PerSample = 0.54140627; TotalTime=0.102487; TotalTimePerSample=4.0033984e-05, SamplesPerSecond=24978
 Epoch[2 of 3]-Minibatch[11-20 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9881856; EvalErr[0]PerSample = 0.54296875; TotalTime=0.09473; TotalTimePerSample=3.7003906e-05, SamplesPerSecond=27024
 Epoch[2 of 3]-Minibatch[21-30 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9869812; EvalErr[0]PerSample = 0.54140627; TotalTime=0.091318; TotalTimePerSample=3.5671094e-05, SamplesPerSecond=28033
 Epoch[2 of 3]-Minibatch[31-40 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9312614; EvalErr[0]PerSample = 0.5277344; TotalTime=0.092408; TotalTimePerSample=3.6096875e-05, SamplesPerSecond=27703
 Epoch[2 of 3]-Minibatch[41-50 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9006774; EvalErr[0]PerSample = 0.52656251; TotalTime=0.098698; TotalTimePerSample=3.8553906e-05, SamplesPerSecond=25937
 Epoch[2 of 3]-Minibatch[51-60 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.9711578; EvalErr[0]PerSample = 0.54140627; TotalTime=0.0896; TotalTimePerSample=3.5e-05, SamplesPerSecond=28571
 Epoch[2 of 3]-Minibatch[61-70 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.8951813; EvalErr[0]PerSample = 0.52031249; TotalTime=0.092477; TotalTimePerSample=3.6123828e-05, SamplesPerSecond=27682
 Epoch[2 of 3]-Minibatch[71-80 of 80]: SamplesSeen = 2560; TrainLossPerSample = 1.904506; EvalErr[0]PerSample = 0.53164065; TotalTime=0.091179; TotalTimePerSample=3.5616797e-05, SamplesPerSecond=28076
 Finished Epoch[2]: [Training Set] TrainLossPerSample = 1.949242; EvalErrPerSample = 0.53417969; Ave LearnRatePerSample = 0.001953125; EpochTime=0.753703
 Starting Epoch 3: learning rate per sample = 0.000098  momentum = 0.656119 
 minibatchiterator: epoch 2: frames [40960..61440] (first utterance at frame 40960) with 1 datapasses
 Epoch[3 of 3]-Minibatch[1-10 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8735985; EvalErr[0]PerSample = 0.51933593; TotalTime=0.27395; TotalTimePerSample=2.675293e-05, SamplesPerSecond=37379
 Epoch[3 of 3]-Minibatch[11-20 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8665626; EvalErr[0]PerSample = 0.51748049; TotalTime=0.261453; TotalTimePerSample=2.553252e-05, SamplesPerSecond=39165
 Finished Epoch[3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; Ave LearnRatePerSample = 9.765625146e-05; EpochTime=0.537273
 COMPLETED
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 Linked to libnvidia-ml library at wrong path : /usr/src/gdk/nvml/lib/libnvidia-ml.so.1
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 === Deleting last epoch data
 ==== Re-running from checkpoint
 running on localhost at 2015/07/29 19:11:14
 command line options: 
 configFile=/home/vlivan/cntk/Tests/Speech/QuickE2E/cntk.config RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu DataDir=/home/vlivan/cntk/Tests/Speech/Data DeviceId=Auto 
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG (VARIABLES NOT RESOLVED) >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=$DeviceId$
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=$RunDir$/models/cntkSpeech.dnn
    deviceId=$DeviceId$
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=$DataDir$/glob_0000.mlf
          labelMappingFile=$DataDir$/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG (VARIABLES NOT RESOLVED)  <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> RAW CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 precision=float
 command=speechTrain
 deviceId=Auto
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu
 DataDir=/home/vlivan/cntk/Tests/Speech/Data
 DeviceId=Auto
 <<<<<<<<<<<<<<<<<<<< RAW CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 >>>>>>>>>>>>>>>>>>>> PROCESSED CONFIG WITH ALL VARIABLES RESOLVED >>>>>>>>>>>>>>>>>>>>
 configparameters: cntk.config:command=speechTrain
 configparameters: cntk.config:DataDir=/home/vlivan/cntk/Tests/Speech/Data
 configparameters: cntk.config:deviceId=Auto
 configparameters: cntk.config:parallelTrain=false
 configparameters: cntk.config:precision=float
 configparameters: cntk.config:RunDir=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu
 configparameters: cntk.config:speechTrain=[
    action=train
    modelPath=/tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu/models/cntkSpeech.dnn
    deviceId=Auto
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=/home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf
          labelMappingFile=/home/vlivan/cntk/Tests/Speech/Data/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
 <<<<<<<<<<<<<<<<<<<< PROCESSED CONFIG WITH ALL VARIABLES RESOLVED <<<<<<<<<<<<<<<<<<<<
 command: speechTrain 
 precision = float
 lsof: WARNING: can't stat() ext4 file system /var/lib/docker/aufs
      Output information may be incomplete.
 LockDevice: Capture device 0 and lock it for exclusive use
 LockDevice: Capture device 0 and lock it for exclusive use
 SimpleNetworkBuilder Using GPU 0
 reading script file glob_0000.scp ... 948 entries
 total 132 state names in state list /home/vlivan/cntk/Tests/Speech/Data/state.list
 htkmlfreader: reading MLF file /home/vlivan/cntk/Tests/Speech/Data/glob_0000.mlf ...parse the line 55130
 total 948 entries
 ...............................................................................................feature set 0: 252734 frames in 948 out of 948 utterances
 label set 0: 129 classes
 minibatchutterancesource: 948 utterances grouped into 3 chunks, av. chunk size: 316.0 utterances, 84244.7 frames
 Starting from checkpoint. Load Network From File /tmp/cntk-test-20150729191101.973007/Speech_QuickE2E@release_gpu/models/cntkSpeech.dnn.2.
 Printing Gradient Computation Node Order ... 
 CrossEntropyWithSoftmax[0, 0] = CrossEntropyWithSoftmax(labels[132, 256], HLast[0, 0])
 HLast[0, 0] = Plus(W2*H1[0, 0], B2[132, 1])
 B2[132, 1] = LearnableParameter
 W2*H1[0, 0] = Times(W2[132, 512], H2[0, 0])
 H2[0, 0] = Sigmoid(W1*H1+B1[0, 0])
 W1*H1+B1[0, 0] = Plus(W1*H1[0, 0], B1[512, 1])
 B1[512, 1] = LearnableParameter
 W1*H1[0, 0] = Times(W1[512, 512], H1[0, 0])
 H1[0, 0] = Sigmoid(W0*features+B0[0, 0])
 W0*features+B0[0, 0] = Plus(W0*features[0, 0], B0[512, 1])
 B0[512, 1] = LearnableParameter
 W0*features[0, 0] = Times(W0[512, 363], MVNormalizedFeatures[0, 0])
 MVNormalizedFeatures[0, 0] = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 InvStdOfFeatures[363, 1] = InvStdDev(features[363, 256])
 MeanOfFeatures[363, 1] = Mean(features[363, 256])
 features[363, 256] = InputValue
 W0[512, 363] = LearnableParameter
 W1[512, 512] = LearnableParameter
 W2[132, 512] = LearnableParameter
 labels[132, 256] = InputValue
 Validating node CrossEntropyWithSoftmax 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax(labels[132, 256], HLast[132, 256])
 Validating node ScaledLogLikelihood 
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> labels = InputValue
 Validating --> Prior = Mean(labels[132, 256])
 Validating --> LogOfPrior = Log(Prior[132, 1])
 Validating --> ScaledLogLikelihood = Minus(HLast[132, 256], LogOfPrior[132, 1])
 Validating node EvalErrorPrediction 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 256], HLast[132, 256])
 GetTrainCriterionNodes  ...
 GetEvalCriterionNodes  ...
 Validating node CrossEntropyWithSoftmax 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 256])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 256])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 256], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 256])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 256], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 256])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 256])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 256], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 256])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 256])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 256], B2[132, 1])
 Validating --> CrossEntropyWithSoftmax = CrossEntropyWithSoftmax(labels[132, 256], HLast[132, 256])
 No PreCompute nodes found, skipping PreCompute step
 Set Max Temp Mem Size For Convolution Nodes to 0 samples.
 Starting Epoch 3: learning rate per sample = 0.000098  momentum = 0.656119 
 minibatchiterator: epoch 2: frames [40960..61440] (first utterance at frame 40960) with 1 datapasses
 requiredata: determined feature kind as 33-dimensional 'USER' with frame shift 10.0 ms
 Validating node EvalErrorPrediction 
 Validating --> labels = InputValue
 Validating --> W2 = LearnableParameter
 Validating --> W1 = LearnableParameter
 Validating --> W0 = LearnableParameter
 Validating --> features = InputValue
 Validating --> MeanOfFeatures = Mean(features[363, 1024])
 Validating --> InvStdOfFeatures = InvStdDev(features[363, 1024])
 Validating --> MVNormalizedFeatures = PerDimMeanVarNormalization(features[363, 1024], MeanOfFeatures[363, 1], InvStdOfFeatures[363, 1])
 Validating --> W0*features = Times(W0[512, 363], MVNormalizedFeatures[363, 1024])
 Validating --> B0 = LearnableParameter
 Validating --> W0*features+B0 = Plus(W0*features[512, 1024], B0[512, 1])
 Validating --> H1 = Sigmoid(W0*features+B0[512, 1024])
 Validating --> W1*H1 = Times(W1[512, 512], H1[512, 1024])
 Validating --> B1 = LearnableParameter
 Validating --> W1*H1+B1 = Plus(W1*H1[512, 1024], B1[512, 1])
 Validating --> H2 = Sigmoid(W1*H1+B1[512, 1024])
 Validating --> W2*H1 = Times(W2[132, 512], H2[512, 1024])
 Validating --> B2 = LearnableParameter
 Validating --> HLast = Plus(W2*H1[132, 1024], B2[132, 1])
 Validating --> EvalErrorPrediction = ErrorPrediction(labels[132, 1024], HLast[132, 1024])
 Epoch[3 of 3]-Minibatch[1-10 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8735985; EvalErr[0]PerSample = 0.51933593; TotalTime=0.430752; TotalTimePerSample=4.2065625e-05, SamplesPerSecond=23772
 Epoch[3 of 3]-Minibatch[11-20 of 20]: SamplesSeen = 10240; TrainLossPerSample = 1.8665626; EvalErr[0]PerSample = 0.51748049; TotalTime=0.2702; TotalTimePerSample=2.6386719e-05, SamplesPerSecond=37897
 Finished Epoch[3]: [Training Set] TrainLossPerSample = 1.8700806; EvalErrPerSample = 0.51840824; Ave LearnRatePerSample = 9.765625146e-05; EpochTime=0.868162
 COMPLETED
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 Linked to libnvidia-ml library at wrong path : /usr/src/gdk/nvml/lib/libnvidia-ml.so.1
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
 WARNING:
 You should always run with libnvidia-ml.so that is installed with your
 NVIDIA Display Driver. By default it's installed in /usr/lib and /usr/lib64.
 libnvidia-ml.so in GDK package is a stub library that is attached only for
 build purposes (e.g. machine that you build your application doesn't have
 to have Display Driver installed).
 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
--- a/Tests/Speech/QuickE2E/cntk.config
+++ b/Tests/Speech/QuickE2E/cntk.config
@ -0,0 +1,63 @@
 precision=float
 command=speechTrain
 deviceId=$DeviceId$
 parallelTrain=false
 speechTrain=[
    action=train
    modelPath=$RunDir$/models/cntkSpeech.dnn
    deviceId=$DeviceId$
    traceLevel=1
    SimpleNetworkBuilder=[
        layerSizes=363:512:512:132
        trainingCriterion=CrossEntropyWithSoftmax
        evalCriterion=ErrorPrediction
        layerTypes=Sigmoid
        initValueScale=1.0
        applyMeanVarNorm=true
        uniformInit=true
        needPrior=true
    ]
    SGD=[
        epochSize=20480
        minibatchSize=64:256:1024:
        learningRatesPerMB=1.0:0.5:0.1
        numMBsToShowResult=10
        momentumPerMB=0.9:0.656119
        dropoutRate=0.0
        maxEpochs=3
        keepCheckPointFiles=true       
        AutoAdjust=[
            reduceLearnRateIfImproveLessThan=0
            loadBestModel=true
            increaseLearnRateIfImproveMoreThan=1000000000
            learnRateDecreaseFactor=0.5
            learnRateIncreaseFactor=1.382
            autoAdjustLR=AdjustAfterEpoch
        ]
        clippingThresholdPerSample=1#INF
    ]
    reader=[
      readerType=HTKMLFReader
      readMethod=blockRandomize
      miniBatchMode=Partial
      randomize=Auto
      verbosity=0
      features=[
          dim=363
          type=Real
          scpFile=glob_0000.scp
      ]
      labels=[
          mlfFile=$DataDir$/glob_0000.mlf
          labelMappingFile=$DataDir$/state.list
          labelDim=132
          labelType=Category
      ]
    ]
 ]
--- a/Tests/Speech/QuickE2E/run-test
+++ b/Tests/Speech/QuickE2E/run-test
@ -0,0 +1,17 @@
 #!/bin/bash
 CNTK_BINARY=$TEST_BUILD_LOCATION/x86_64.gpu.$TEST_FLAVOR.acml/cntk
 if [ "$TEST_DEVICE" == "CPU" ]; then
   CNTK_DEVICE_ID=-1
 else
   CNTK_DEVICE_ID=Auto
 fi
 CNTK_ARGS="configFile=$TEST_DIR/cntk.config RunDir=$TEST_RUN_DIR DataDir=$TEST_DATA_DIR DeviceId=$CNTK_DEVICE_ID"
 MODELS_DIR=$TEST_RUN_DIR/models
 [ -d $MODELS_DIR ] && rm -rf $MODELS_DIR
 mkdir -p $MODELS_DIR || exit $?
 echo === Running $CNTK_BINARY $CNTK_ARGS
 $CNTK_BINARY $CNTK_ARGS || exit $?
 echo === Deleting last epoch data
 rm $TEST_RUN_DIR/models/*.dnn
 echo ==== Re-running from checkpoint
 $CNTK_BINARY $CNTK_ARGS || exit $?
--- a/Tests/Speech/QuickE2E/testcases.yml
+++ b/Tests/Speech/QuickE2E/testcases.yml
@ -0,0 +1,27 @@
 dataDir: ../Data
 testCases:
  CNTK Run must be completed:
    patterns:
      - ^COMPLETED
  Must train epochs in exactly same order and parameters:
    patterns:
      - ^Starting Epoch {{integer}}
      - learning rate per sample = {{float}}
      - momentum = {{float}}
  Epochs must be finished with expected results:
    patterns:
      - ^Finished Epoch[{{integer}}]
      - TrainLossPerSample = {{float,tolerance=1%}}
      - EvalErrPerSample = {{float,tolerance=1%}}
      - Ave LearnRatePerSample = {{float,tolerance=1%}}
  Per-minibatch training results must match:
    patterns:
      - ^ Epoch[{{integer}} of {{integer}}]-Minibatch[{{integer}}-{{integer}} of {{integer}}]
      - SamplesSeen = {{integer}}
      - TrainLossPerSample = {{float,tolerance=1%}}
      - EvalErr[0]PerSample = {{float,tolerance=1%}}
--- a/Tests/TestDriver.py
+++ b/Tests/TestDriver.py
@ -0,0 +1,551 @@
 #!/usr/bin/env python
 # ----------------------------------------------------------
 # Copyright (c) Microsoft Corporation.  All rights reserved.
 # ---------------------------------------------------------
 # This is a test driver for running end-to-end CNTK tests
 #
 # ----- Running a test and/or updating baselines ------
 # For instructions see:
 # ./TestDriver.py --help
 #
 # ---- Adding the tests: -------
 # File system organization:
 #   Each test suite (e.g. Speech) has its own directory inside Tests
 #   Each test (e.g. QuickE2E) has its own directory within test suite
 #
 # Each test directory has a following components:
 #    - testcases.yml - main test confuguration file, whcih defines all test cases
 #    - run-test  - (run-test) script
 #    - baseline*.txt - baseline files whith a captured expected output of run-test script
 #
 # ----- testcases.yml format -------
 # dataDir: <path> #<relative-path-to the data directory
 #
 # testCases:
 #   <name of the testcase 1>:  
 #     patterns:
 #       - <pattern 1> # see pattern language
 #       - <pattern 2>
 #       - .....
 #
 #   <name of the testcase 2>:
 #     patterns:
 #       - <pattern 1>
 #       - <pattern 2>
 #       - .....
 #   .....
 #
 # ----- pattern language --------
 # Multpile patterns of the same testcase are matching a *single* line of text
 # Pattern is essentiually a substring which has to be found in a line
 # if pattern starts with ^ then matching is constrained to look only at the beginning of the line
 #
 # pattern can have one or multiple placelohders wrapped with double-curly braces:  {{...}}
 # this placeholders can match any text conforming to the type constraint. Available placeholders
 #  {{integer}} - matches any (positive or negative integer) value
 #  {{float}} - matches any float value
 #  {{float,tolerance=0.00001}} - matches float value with given absolute tolerance: 0.00001 in this example
 #  {{float,tolerance=2%}} - matches float value with relative tolerance, 2% in this example
 #
 # At runtime patterns are compiled by TestDriver.py to regular expressions
 #
 # ---- Baseline files ----
 # Order of searching baseline files, depends on the current mode for a given test:
 #
 #   1. baseline.<flavor>.<device>.txt
 #   2. baseline.<flavor>.txt
 #   3. baseline.<device>.txt
 #   4. baseline.txt
 #        where <flavor> = { debug | release }
 #              <device> = { cpu | gpu }
 # 
 # ----- Algorithm ------
 # Baseline verification:
 #   For each testcase 
 #     - filter all lines which matches
 #       - if no lines found then abord with an error - since either baseline and/or pattern are invalid
 # Running test:
 #    Run test script (run-test) and capture output:
 #
 #    For each testcase
 #      - filter all matching lines from baseline 
 #      - filter all matching lines from test output
 #      - compare filtered lines one by one, ensuring that substrings defined by patterns are matching
 #
 # In practice, TestDriver performs 1 pass through the output of run-test performing a real-time 
 # matching against all test-cases/pattern simulteneously
 #
 import sys, os, argparse, traceback, yaml, subprocess, random, re, time
 thisDir = os.path.dirname(os.path.realpath(__file__))
 # This class encapsulates an instance of the test
 class Test:
  # "Suite/TestName" => instance of Test
  allTestsIndexedByFullName = {} 
  # suite - name of the test suite
  # name - name of the test
  # path to the testcases.yml file
  def __init__(self, suite, name, pathToYmlFile):
    self.suite = suite
    self.name = name
    self.fullName = suite + "/" + name
    # computing location of test directory (yml file directory)
    self.testDir = os.path.dirname(pathToYmlFile)
    # parsing yml file with testcases 
    with open(pathToYmlFile, "r") as f:
      self.rawYamlData = yaml.safe_load(f.read())
    # finding location of data directory
    if self.rawYamlData["dataDir"]:
      self.dataDir = os.path.realpath(os.path.join(self.testDir, self.rawYamlData["dataDir"]))
    else:
      self.dataDir = self.testDir
    testCasesYaml = self.rawYamlData["testCases"]
    self.testCases = []
    for name in testCasesYaml.keys():
      try:
        self.testCases.append(TestCase(name, testCasesYaml[name]))
      except Exception as e:
        print >>sys.stderr, "ERROR registering test case: " + name
        raise
  # Populates Tests.allTestsIndexedByFullName by scanning directory tree
  # and finding all testcases.yml files
  @staticmethod
  def discoverAllTests():
    for dirName, subdirList, fileList in os.walk(thisDir):
      if 'testcases.yml' in fileList:
        testDir = dirName
        testName = os.path.basename(dirName)
        suiteDir = os.path.dirname(dirName)
        # sute name will be derived from the path components
        suiteName = os.path.relpath(suiteDir, thisDir).replace('\\', '/')
        try:
          test = Test(suiteName,  testName, dirName + "/testcases.yml")
          Test.allTestsIndexedByFullName[test.fullName.lower()] = test
        except Exception as e:
          print >>sys.stderr, "ERROR registering test: " + dirName
          traceback.print_exc()
          sys.exit(1)
  # Runs this test
  #   flavor - "debug" or "release"
  #   device - "cpu" or "gpu"
  #   args - command line arguments from argparse
  # returns an instance of TestRunResult
  def run(self, flavor, device, args):
    # Locating and reading baseline file
    baselineFile = self.findBaselineFile(flavor, device)
    if baselineFile == None:
      return TestRunResult.fatalError("Baseline file sanity check", "Can't find baseline file")
    with open(baselineFile, "r") as f:
      baseline = f.read().split("\n")
      if args.verbose:
         print "Baseline:", baselineFile
    # Before running the test, pre-creating TestCaseRunResult object for each test case
    # and compute filtered lines from baseline file.
    # Note: some test cases might fail at this time if baseline and/or patterns are inconsistant
    result = TestRunResult()
    result.succeeded = True
    if not args.update_baseline:
      for testCase in self.testCases:
        testCaseRunResult = testCase.processBaseline(baseline)
        if not testCaseRunResult.succeeded:
           result.succeeded = False
        result.testCaseRunResults.append(testCaseRunResult)
    # preparing run directory
    runDir = os.path.join(args.run_dir, "{0}_{1}@{2}_{3}".format(self.suite, self.name, flavor, device))
    if not os.path.isdir(runDir):
      os.makedirs(runDir)
    # preparing environment for the test script
    os.environ["TEST_FLAVOR"] = flavor
    os.environ["TEST_DEVICE"] = device
    os.environ["TEST_BUILD_LOCATION"] = args.build_location
    os.environ["TEST_DIR"] = self.testDir
    os.environ["TEST_DATA_DIR"] = self.dataDir
    os.environ["TEST_RUN_DIR"] = runDir
    # WORKAROUND: changing current dir to the dataDir so relative paths in SCP files work as expected
    os.chdir(self.dataDir)
    # Running test script
    #TODO:port this properly to windows
    # Writing standard output to the file and to the console (if --verbose)
    logFile = os.path.join(runDir, "output.txt")
    allLines = []
    if args.verbose:
      print self.fullName + ":>" + logFile
    with open(logFile, "w") as output:
      cmdLine = ["bash", "-c", self.testDir + "/run-test 2>&1"]
      process = subprocess.Popen(cmdLine, stdout=subprocess.PIPE)
      while True:
        line = process.stdout.readline() 
        if not line:
           break
        if len(line)>0 and line[-1]=='\n':
          line=line[:len(line)-1]
        if args.verbose:
          print self.fullName + ": " + line
        print >>output, line
        allLines.append(line)
        output.flush()
        for testCaseRunResult in result.testCaseRunResults:
          testCaseRunResult.testCase.processLine(line, testCaseRunResult, args.verbose)
    exitCode = process.wait()
    success = True
    # checking exit code
    if exitCode != 0:
      return TestRunResult.fatalError("Exit code must be 0", "==> got exit code {0} when running: {1}".format(exitCode, " ".join(cmdLine)), logFile = logFile)
    # saving log file path, so it can be reported later
    result.logFile = logFile
    # finalizing verification - need to check whether we have any unmatched lines
    for testCaseRunResult in result.testCaseRunResults:
      testCaseRunResult.testCase.finalize(testCaseRunResult)
      if not testCaseRunResult.succeeded:
        result.succeeded = False
    if args.update_baseline and result.succeeded:
      # When running in --update-baseline mode 
      # verifying that new output is succesfully matching every pattern in the testcases.yml
      # If this is not the case then baseline update will be rejected
      for testCase in self.testCases:
        testCaseRunResult = testCase.processBaseline(allLines)
        if not testCaseRunResult.succeeded:
           result.succeeded = False
        result.testCaseRunResults.append(testCaseRunResult)
      if result.succeeded:
       if args.verbose:
         print "Updating baseline file", baselineFile
       with open(baselineFile, "w") as f:
         f.write("\n".join(allLines))
    return result
  # Finds a location of a baseline file by probing different names in the following order:
  #   baseline.$flavor.$device.txt
  #   baseline.$flavor.txt
  #   baseline.$device.txt
  #   baseline.txt
  def findBaselineFile(self, flavor, device):
    for f in ["." + flavor.lower(), ""]:
      for d in ["." + device.lower(), ""]:
        candidateName = "baseline" + f + d + ".txt";
        fullPath = os.path.join(self.testDir, candidateName)
        if os.path.isfile(fullPath):
           return fullPath
    return None
 # This class encapsulates one testcase (in testcases.yml file)
 class TestCase:
  def __init__(self, name, yamlNode):
    self.name = name
    self.patterns = []
    if "patterns" in yamlNode:
      for pattern in yamlNode["patterns"]:
        try:
          self.patterns.append(TestPattern(pattern))
        except Exception as e:
          print >>sys.stderr, "ERROR registering pattern: " + pattern
          raise
  # Processes the baseline file and return an instance of TestCaseRunResult
  # which is ready to be passed into processLine
  def processBaseline(self, baseline):
    result = TestCaseRunResult(self.name, True)
    result.diagnostics = ""
    result.testCase = self
    # filter all lines of baseline file leaving only those which match ALL the patterns   
    filteredLines = []
    for line in baseline:
      if all([p.match(line) for p in self.patterns]):
        filteredLines.append(line)
    if len(filteredLines) == 0:
       result.succeeded = False
       result.diagnostics+="Baseline file doesn't have any lines matching all patterns defined in the test case.\n"\
                           "Possible cause: patterns are wrong and/or baseline file doesn't have required line"
    result.expectedLines = filteredLines
    return result
  # Runs this test case and report result into TestCaseRunResult
  def processLine(self, line, result, verbose):
    if all([p.match(line) for p in self.patterns]):
      if len(result.expectedLines) > 0:
        # we have mathed line in the output and at leat one remaining unmatched in a baseline
        expected = result.expectedLines[0]
        # running comparison logic for each pattern
        failedPatterns = []
        for p in self.patterns:
          if not p.compare(expected, line):
            result.succeeded = False
            failedPatterns.append(p)
        # in the case of failure - reporting mismatched lines
        if len(failedPatterns)>0:
          result.diagnostics+=("Baseline: {0}\n"+
                               "Output:   {1}\n"
                              ).format(expected, line)
          if verbose:
            print "[FAILED]: Testcase", self.name
            print "Baseline:", expected
          # also show all failed patterns
          for p in failedPatterns:
            msg = "Failed pattern: " + p.patternText
            if verbose:
              print msg
            result.diagnostics+=msg+"\n"
        # removing this line, since we already matched it (whether succesfully or not - doesn't matter)
        del result.expectedLines[0]
      else:
        # we have matched line in the output - but don't have any remaining unmatched in a baseline
        result.succeeded = False
        result.diagnostics+=("Unexpected (extra) line in the output which matches the pattern, but doesn't appear in baseline file.\n"+
                             "Extra line: {0}"
                            ).format(line)
  # called once for each TestCaseRunResult at the end to check for unmatched patterns
  def finalize(self, result):
    if len(result.expectedLines) > 0:
       result.succeeded = False
       result.diagnostics+=("{0} expected lines weren't observed in the output.\n"+
                            "First unmatched: {1}"
                           ).format(len(result.expectedLines), result.expectedLines[0])
 # This encapsulates parsing and evaluation of a test patterns occurring in testcases.yml file
 class TestPattern:
  # maps a type (specified in {{...}} expressions) to a regular expression
  typeTable = {
     "integer" : r"\s*-?[0-9]+",
     "float"   : r"\s*-?([0-9]*\.[0-9]+|[0-9]+)(e[+-]?[0-9]+)?"
   }
  def __init__(self, patternText):
    self.patternText = str(patternText)
    if len(patternText) == 0: 
      raise Exception("Empty pattern")
    if patternText[0]=='^':
      patternText = patternText[1:]
      prefix = "^"
    else:
      prefix = ".*?"
    # After parsing this will be a list of tuples (dataType, tolerance) for each {{...}} section from left to right
    self.groupInfo = []
    # Transforming our pattern into a sigle regular expression
    # processing {{...}} fragments and escaping all regex special characters
    self.regexText = prefix + re.sub(r"(\{\{[^}]+\}\}|[\[\]\.\*\+\{\}\(\)\$\^\\\|\?])", self.patternParse, patternText)
    # Compiling it to perform a check (fail-fast) and for faster matching later
    self.regex = re.compile(self.regexText)
  # this is a callback method passed to re.sub call above - it performs the core parsing logic
  def patternParse(self, match):
    fragment = match.group(1)
    if len(fragment) == 1:
      # this is a spexcial character of regex
      return "\\" + fragment;
    else:
      # parsing {{...}} expressions
      m = re.match(r"{{(integer|float)(,tolerance=([-0-9\.e]*)(%?))?}}", fragment)
      dataType = m.group(1)
      if m.group(3):
        tolerance = float(m.group(3))
        if m.group(4) == "%":
          # using minus sign to indicate that it is a relative value
          tolerance = - tolerance/100.0;
      else:
        tolerance = 0.0
      # saving information about data type and tolerance
      self.groupInfo.append((dataType, tolerance))
      # converting this to regex which mathes specific type
      # All {{...}} sections are converted to regex groups named as G0, G1, G2...
      return "(?P<G{0}>{1})".format(len(self.groupInfo)-1, TestPattern.typeTable[dataType])
  # Checks wether given line matches this pattern
  # returns True or False
  def match(self, line):
    return self.regex.match(line) != None
  # Compares a line from baseline log and a line from real output against this pattern
  # return true or false
  def compare(self, expected, actual):
    em = self.regex.match(expected)
    am = self.regex.match(actual)
    if em == None and am == None:
       return True
    if em == None or am == None:
       return False
    for i in range(0, len(self.groupInfo)):
      dataType, tolerance = self.groupInfo[i]
      groupId = "G"+str(i)
      expectedText = em.group(groupId).strip()
      actualText = am.group(groupId).strip()
      if dataType=="integer":
        return int(expectedText) == int(actualText)
      elif dataType=="float":
        epsilon = tolerance if tolerance > 0 else abs(float(expectedText)*tolerance)
        return abs(float(expectedText)-float(actualText)) <= epsilon
      else:
        return False;
    return True
 class TestRunResult:
  def __init__(self):
    self.succeeded = False;
    self.testCaseRunResults = [] # list of TestCaseRunResult
  @staticmethod
  def fatalError(name, diagnostics, logFile = None):
    r = TestRunResult()
    r.testCaseRunResults.append(TestCaseRunResult(name, False, diagnostics))
    r.logFile = logFile
    return r
 class TestCaseRunResult:
  def __init__(self, testCaseName, succeeded, diagnostics = None):
    self.testCaseName = testCaseName
    self.succeeded = succeeded
    self.diagnostics = diagnostics
    self.expectedLines = [] # list of remaining unmatched expected lines from the baseline file for this test case run
 # Lists all available tests
 def listCommand(args):
  for t in Test.allTestsIndexedByFullName.values():
    print t.fullName
 # Runs given test(s) or all tests
 def runCommand(args):
  if len(args.test) > 0:
     testsToRun = []
     for name in args.test:
       if name.lower() in Test.allTestsIndexedByFullName:
         testsToRun.append(Test.allTestsIndexedByFullName[name.lower()])
       else:
         print >>sys.stderr, "ERROR: test not found", name
         return 1
  else:
     testsToRun = Test.allTestsIndexedByFullName.values()
  devices = ["cpu", "gpu"]
  if (args.device):
    args.device = args.device.lower()
    if not args.device in devices:
      print >>sys.stderr, "--device must be one of", devices
      return 1
    devices = [args.device]
  flavors = ["debug", "release"]
  if (args.flavor):
    args.flavor = args.flavor.lower()
    if not args.flavor in flavors:
      print >>sys.stderr, "--flavor must be one of", flavors
      return 1
    flavors = [args.flavor]
  print "CNTK Test Driver is started"
  print "Running tests:  ", " ".join([y.fullName for y in testsToRun])
  print "Build location: ", args.build_location
  print "Run location:   ", args.run_dir
  print "Flavors:        ", " ".join(flavors)
  print "Devices:        ", " ".join(devices)
  if (args.update_baseline):
    print "*** Running in automatic baseline update mode ***"
  print ""
  succeededCount, totalCount = 0, 0
  for test in testsToRun:
    for flavor in flavors:
      for device in devices:
        totalCount = totalCount + 1
        # Printing the test which is about to run (without terminating the line)
        sys.stdout.write("Running test {0} ({1} {2}) - ".format(test.fullName, flavor, device));
        # in verbose mode, terminate the line, since there will be a lot of output
        if args.verbose:
          sys.stdout.write("\n");
        sys.stdout.flush()
        # Running the test and collecting a run results
        result = test.run(flavor, device, args)
        if args.verbose:
          # writing the test name one more time (after possibly long verbose output)
          sys.stdout.write("Test finished {0} ({1} {2}) - ".format(test.fullName, flavor, device));
        if result.succeeded:
          succeededCount = succeededCount + 1
          # in no-verbose mode this will be printed in the same line as 'Running test...'
          print "[OK]"
        else:
          print "[FAILED]"
        # Showing per-test-case results:
        for testCaseRunResult in result.testCaseRunResults:
           if testCaseRunResult.succeeded:
             # Printing 'OK' test cases only in verbose mode
             if (args.verbose):
               print(" [OK] " + testCaseRunResult.testCaseName);
           else:
             # 'FAILED' + detailed diagnostics with proper indendtation
             print(" [FAILED] " + testCaseRunResult.testCaseName);
             if testCaseRunResult.diagnostics:
               for line in testCaseRunResult.diagnostics.split('\n'):
                 print "    " + line;
             # In non-verbose mode log wasn't piped to the stdout, showing log file path for conveniencce
        if not result.succeeded and not args.verbose and result.logFile:
          print "  See log file for details:", result.logFile
  if args.update_baseline:
    print "{0}/{1} baselines updated, {2} failed".format(succeededCount, totalCount, totalCount - succeededCount)
  else:
    print "{0}/{1} tests passed, {2} failed".format(succeededCount, totalCount, totalCount - succeededCount)
  if succeededCount != totalCount:
    sys.exit(10)
 # ======================= Entry point =======================
 parser = argparse.ArgumentParser(description="TestDriver - CNTK Test Driver")
 subparsers = parser.add_subparsers(help="command to execute. Run TestDriver.py <command> --help for command-specific help")
 runSubparser = subparsers.add_parser("run", help="run test(s)")
 runSubparser.add_argument("test", nargs="*",
                    help="optional test name(s) to run, specified as Suite/TestName. "
                         "Use list command to list available tests. "
                         "If not specified then all tests will be run.")
 #TODO: port paths to Windows
 defaultBuildLocation=os.path.realpath(os.path.join(thisDir, "..", "bin"))
 runSubparser.add_argument("-b", "--build-location", default=defaultBuildLocation, help="location of the CNTK build to run")
 runSubparser.add_argument("-d", "--device", help="cpu|gpu - run on a specific device")
 runSubparser.add_argument("-f", "--flavor", help="release|debug - run only a specific flavor")
 #TODO: port paths to Windows
 defaultRunDir=os.path.join("/tmp", "cntk-test-{0}.{1}".format(time.strftime("%Y%m%d%H%M%S"), random.randint(0,1000000)))
 runSubparser.add_argument("-r", "--run-dir", default=defaultRunDir, help="directory where to store test output, default: a random dir within /tmp")
 runSubparser.add_argument("--update-baseline", action='store_true', help="update baseline file(s) instead of matching them")
 runSubparser.add_argument("-v", "--verbose", action='store_true', help="verbose output - dump all output of test script")
 runSubparser.set_defaults(func=runCommand)
 listSubparser = subparsers.add_parser("list", help="list available tests")
 listSubparser.set_defaults(func=listCommand)
 if len(sys.argv)==1:
    parser.print_help()
    sys.exit(1)
 args = parser.parse_args(sys.argv[1:])
 # discover all the tests
 Test.discoverAllTests()
 # execute the command
 args.func(args)
--- a/ThirdPartyNotices.txt
+++ b/ThirdPartyNotices.txt
@ -0,0 +1,42 @@
 === AN4 dataset ====
 Contents of Tests/Speech/Data directory is a modified version of AN4 dataset pre-processed and optimized for CNTK end-to-end testing.
 AN4 dataset is a part of CMU audio databases located at http://www.speech.cs.cmu.edu/databases/an4
 This modified version of dataset is distributed under the terms of a AN4 license:
 /* ====================================================================
 * Copyright (c) 1991-2005 Carnegie Mellon University.  All rights
 * reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer. 
 *
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in
 *    the documentation and/or other materials provided with the
 *    distribution.
 *
 * This work was supported in part by funding from the Defense Advanced 
 * Research Projects Agency and the National Science Foundation of the 
 * United States of America, and the CMU Sphinx Speech Consortium.
 *
 * THIS SOFTWARE IS PROVIDED BY CARNEGIE MELLON UNIVERSITY ``AS IS'' AND 
 * ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, 
 * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL CARNEGIE MELLON UNIVERSITY
 * NOR ITS EMPLOYEES BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 
 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 
 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 
 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 
 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 
 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 *
 * ====================================================================
 */