repository
stringclasses
156 values
issue title
stringlengths
1
1.01k
labels
stringclasses
8 values
body
stringlengths
1
270k
tensorflowtensorflow
issue with tensorflow on window when use tensorforestestimator
Bug
I already have conda installation in my window I instal tensorflow use pip and it be work fine base on the testing the tensorflow installation test the tensorflow installation when I m use tensorforestestimator I m get the follow error lib site package tensorflow contrib tensor forest python op t...
tensorflowtensorflow
docker doesn t support gpu for mac os doc clarification request
Bug
the doc present instruction for instal via docker on macos gpu support isn t mention explicitly but it seem that gpu nvidia docker can t be support for macos see here if I m right that gpu be a no go for docker on mac os I d like to see this up front in the documentation e g write gpu support ...
tensorflowtensorflow
t eval build with slim output wrong prediction when input batch contain identical image
Bug
when evaluate a model build with slim on a batch that contain identical image the output of the batch will be wrong mostly the output will be the most frequent label for example if my evaluation set have 421 image and my batch size be 40 I fill the last batch with 19 identical image the 421st one to avoi...
tensorflowtensorflow
timeout do not work with session create with server
Bug
what related github issue or stackoverflow thread have you find by search the web for your problem none environment info operating system ubuntu instal version of cuda and cudnn please attach the output of ls l path to cuda lib libcud cpu only if instal from source provide ...
tensorflowtensorflow
cudnnlstm dropout take no effect
Bug
my environment be tensorflow 0 11 0rc2 in unbuntu 16 04 cuda8 0 cudnn5 1 gpu be gtx1080 I be use the cudnnlstm from tensorflow contrib cudnn rnn package I find that the dropout set in cudnnlstm seem take no effect and I check that there be no test for dropout in op unit test so I write a simple cod...
tensorflowtensorflow
slow adam sparse update in distribute tf
Bug
I be try to train a model with the tf nn embed lookup sparse operation small example when I run this code with num worker 1 the output be as follow 0 calc in 0 0176000595093 0 apply in 0 184364080429 0 calc in 0 0167639255524 0 apply in 0 189659118652 however when I inc...
tensorflowtensorflow
can not show stderr when use jupyter
Bug
hello could you please have a look about this I be use tf and jupyter but what make I confuse be that the log text can not be show in jupyter output cell but it output correctly in ipython I think it be because of the stderr this issue have be discuss before in 3047 you add several line to determ...
tensorflowtensorflow
outdate documentation of assert op
Bug
from assert ensure maximum element of x be small or equal to 1 assert op tf assert tf less equal tf reduce max x 1 x x tf with dependency assert op x tf with dependency do not exist see also a related stack overflow discussion
tensorflowtensorflow
confusion about cuda install doc for mac
Bug
environment info operating system mac osx 10 11 the doc have 2 issue relate to this section at the top the gpu version work well with cuda toolkit 8 0 and cudnn v5 other version be support cuda toolkit 7 0 and cudnn v3 only when instal from source please see cuda installation for deta...
tensorflowtensorflow
error in einsum with unspecified dimension
Bug
code import tensorflow as tf x tf placeholder dtype tf int32 a tf random normal 13 x 512 b tf random normal 512 3 c tf einsum ijk kl ijl a b with tf session as sess print sess run c x 7 output ...
tensorflowtensorflow
inconsistency in parameter name
Bug
tf nn conv1d value filter stride padding use cudnn on gpu none datum format none name none tf nn conv2d input filter stride padding use cudnn on gpu none datum format none name none tf nn conv3d input filter stride padding name none there s discrepancy between param...
tensorflowtensorflow
user contribute tutorial
Bug
it would be great if there be a section in the tensorflow org website where the community could provide tutorial for new user there be a lot of great tutorial in the web and have some on the tensorflow org website would be very valuable for people just start out thought here be an example of what I would love ...
tensorflowtensorflow
error 404 when download tensorflow on window
Bug
the link provide on the website to the whl use to install tensorflow on window seem to be break pip installation on window I m get an http error 404 I find the cpu build elsewhere but I can t find the gpu build which I would like tu use
tensorflowtensorflow
in debug mode get assertion cudagetlasterror cudasuccess fail
Bug
environment ubuntu 14 04 cuda 8 cudnn 5 1 tensorflow r0 12 build command bazel build c opt config cuda c dbg strip never tensorflow tool pip package build pip package previously follow the installation of tensorflow not in debug mode my code work well but after I rebuild with ...
tensorflowtensorflow
link to window build in os setup documentation be bad
Bug
build location be list under pip installation on window for example when try to retrieve I get the response xml nosuchkey the specify key do not exist
tensorflowtensorflow
grpc worker client fetch top out at 120 mb s
Bug
fetch datum as numpy array be slow for grpc session evaluate 128 mb variable repeatedly from session grpc localhost only get I 116 mb s compare to 5150 mb s for session python client transfer benchmark py profile 5150 04 mb per second 116 70 mb per second attach be cpu p...
tensorflowtensorflow
embed projector click on point doesn t select they
Bug
in chrome click on point regardless of dataset pca or t sne 2d or 3d will not select they win10 chrome version 55 0 2883 87 m
tensorflowtensorflow
format error on the website
Bug
hi on tensorflow org there be part of documentation accidentally format into code it s not a serious problem it s about initializer on r0 11 I didn t check if other version have the same problem just want to post it here for your information screen shoot 2016 12 09 at 20 41 19 link random ...
tensorflowtensorflow
extremely inefficient cuda event poll
Bug
I be run tensorflow with this application I be run on 4 nvidia k80 card and cuda 7 5 for my experiment I modify the code to only do 100 training iteration during the 100 iteration tensorflow destroy 2662 cuda event but how many time it poll event status 40669120 time this mean on average it poll 1...
tensorflowtensorflow
slice input producer output tensor of wrong shape under certain condition
Bug
from the verison tensorflow 0 12 0rc0 cp27 cp27mu manylinux1 x86 64 whl when use tf train slice input producer with a tensor list contain only a single tensor the output from the tf train batch have incorrect dimension code to reproduce it import tensorflow as tf datum tf random uniform shap...
tensorflowtensorflow
upgrade from r11 to r12 prodeuce variable not define when use any optimizer but gradientdescentoptimizer
Bug
after a recent upgrade to the late version of tensorflow in github several thing stop work I find out that all the optimizer such as adam or adagrad be now produce an error relate to variable scope that I have not manage to solve yet however gradientdescentoptimizer work fine it may be relate to the issu...
tensorflowtensorflow
embed projector hang when datum with string base feature be add
Bug
what related github issue or stackoverflow thread have you find by search the web for your problem none environment info operating system os x 10 11 6 answer to other prompt n a if possible provide a minimal reproducible example we usually don t have time to read hundred of line of...
tensorflowtensorflow
inconsistent api for one initializer zeros initializer
Bug
this cl have remove shape argument from one initializer however this argument remain for zero initializer this be a break change so maybe it should be mention in release md with instruction on how code should transition ie one initializer should be replace with one initializer curr...
tensorflowtensorflow
invalid proto descriptor for file tensorflow contrib tensorboard plugin trace trace info proto
Bug
trace info proto seem to be miss import tensorflow core protobuf meta graph proto it work for python protobuf implementation but fail with trace run on strict cpp implementation scratch py python scratch py contrib importlib import module tensorflow contrib file hom...
tensorflowtensorflow
inconsistent behavior for tf variable scope
Bug
problem tensorflow doesn t place op e g mul in pre exist variable scope and automatically create a new scope instead minimal reproducible example python with tf variable scope layer123 v tf get variable v initializer tf constant initializer 42 tf float32 ...
tensorflowtensorflow
wide n deep tutorial example not work
Bug
window 10 0 12 0 rc0 encounter error attributeerror nonetype object have no attribute bucketize when run the example without any modification full working and error log as follow warn tensorflow the default value of combiner will change from sum to sqrtn after 2016 11 01 warning ...
tensorflowtensorflow
quantization code in different path to where specify in help documentation
Bug
follow the documentation here to quantize seem to be wrong I have to change bazel build tensorflow contrib quantization tool quantize graph bazel bin tensorflow contrib quantization tool quantize graph input tmp classify image graph def pb output node name softmax output tmp qua...
tensorflowtensorflow
tfprof python3 incompatibility
Bug
this line in tfprof logger py l127 use dict iteritem which break my python 3 code
tensorflowtensorflow
tfrecord datalosserror see above for traceback corrupt record at xxx
Bug
I convert large datum from csv to tfrecord use tf python io tfrecordwriter in hadoop there re some error happen bug 1 if I use zlib or gzip when create tfrecord writer I can convert the csv successfully but when I loop all file to read it s stuck at some line without any error when use these tfr...
tensorflowtensorflow
opening show a white screen in chrome 55
Bug
what related github issue or stackoverflow thread have you find by search the web for your problem I have search for open github issue with the keyword webgl but none show up issue today I open in my browser but get a complete white screen the attached log show the error that be show i...
tensorflowtensorflow
newcheckpointreader do not work for the v2 checkpoint format
Bug
unable to read the new checkpoint v2 format that become default from r0 12 release use newcheckpointreader this break the tf slim api specifically l487 what related github issue or stackoverflow thread have you find by search the web for your problem topic discuss 3aalcbj 6z8 I start that ...
tensorflowtensorflow
contrib learn estimator do not work with multiple gpu
Bug
attempt to assign op to a gpu within model fn pass to an estimator produce the follow error tensorflow python framework error impl invalidargumenterror can not assign a device to node save shardedfilename 2 could not satisfy explicit device specification device gpu 1 because no support kernel for ...
tensorflowtensorflow
error need when load a graphdef create in a new version of tensorflow
Bug
there be a lot of issue e g 5628 that be cause by user loading graphdef that be create by a new version of tensorflow e g 0 12 that introduce saverv2 into old code basis this currently result in cryptic error late in the loading process but instead we should raise an error or at least a prominent war...
tensorflowtensorflow
different prediction result for tf learn quickstart
Bug
hello today I upgrade tensorflow package and read the tutorial from the beginning find the result of classifi prediction be different it be 1 1 I remember that the old version tf run out 1 2 so what s the problem the os be osx ei capitan and tensoflow be 0 12 0 rc0 pyth...
tensorflowtensorflow
flawed memory management allow growth true consume more memory cause out of memory
Bug
to prevent tensorflow tf from allocate the totality of graphic memory I always use the follow option when create session config tf configproto config gpu option allow growth true sess tf session config config however do so cause some experiment to run out of memory while not d...
tensorflowtensorflow
stride slice op check failure
Bug
two report of a check failure in the strided slice op have surface on stack overflow tensorflow print value of a tensor tensorflow evaluate abort core dump in both case the error message be f tensorflow core kernel stride slice op cc 316 check fail tmp copyfrom input slice ...
tensorflowtensorflow
rnn tutorial code can not find sample datum
Bug
follow the direction here on my mac recurrent neural network cd tensorflow model rnn ptb python ptb word lm py datum path tmp simple example datum model small traceback most recent call last file ptb word lm py line 368 in tf app run file user shleifer...
tensorflowtensorflow
ctc beam search decoder s log probability hold invalid value
Bug
environment info operating system os x 10 11 6 tf version 0 10 0rc0 no gpu example run log probability op create from python decode log probability ctc op ctc beam search decoder logit seq length ...
tensorflowtensorflow
for which input be gradient compute
Bug
what be a good way to determine which input of a tensorflow op be backpropagate through for example consider the crop and resize op the description sound rather straightforward extract crop from the input image tensor and bilinearly resize they however if you dig through the code this func...
tensorflowtensorflow
where be tensorflow directory on window
Bug
I have instal and verify tensorflow installation on window that s great but how I be suppose to to stuff like pyhthon3 cifar10 train py run that command get error no such file or directory also run the whole code in cifar10 py and get error there be no module name tensorflow modle importerror be unhand...
tensorflowtensorflow
computer freeze when feed a large numpy array as input in mnist tutorial
Bug
what related github issue or stackoverflow thread have you find by search the web for your problem none environment info intel i5 8 gb ram operating system ubuntu 14 04 5 linux 4 4 0 45 generic 66 14 04 1 ubuntu smp we d oct 19 15 05 38 utc 2016 x86 64 x86 64 x86 64 gnu linux instal ve...
tensorflowtensorflow
request for documentation on recommend flow in slim for train validation and test set
Bug
the example in the slim readme md give basic documentation for training and evaluate model when use separately however there be guidance miss on how to do the classic cycle of mini batch gradient descent use shuffle subset of the training set periodically evaluate validation set and then evaluate on the te...
tensorflowtensorflow
update cmake readme doc reflect 0 12r
Bug
now with 0 12r would be interesting to update readme as current status instruction or simply reference the release
tensorflowtensorflow
io no opkernel be register to support op spacetobatchnd with these attrs
Bug
I be try to use tf nn atrous conv2d to solve a segmentation problem it work on my mac but when I save this graph use convert variable to constant and tf train write graph then read it in use c api try to get it run on my ipad I get this error could not create tensorflow graph invalid argu...
tensorflowtensorflow
break link in a tool developer s guide for graph metric py
Bug
in this page graph metric py link break graph metric py be delete by this commit diff 2839c4fb2e611a1359978388f55f3583
tensorflowtensorflow
no documentation for send input parameter of request to tensorflow serve
Bug
I be run the sample iris program in tensorflow serve since it be a tf learn model I be export the model use the following classifier export export dir model dir signature fn my classification signature fn and the signature fn be define as show below def my classification signature fn example unu...
tensorflowtensorflow
strip unused assume all placeholder be the same type
Bug
strip unused assume that all of the placeholder be of the same type which may not be the case this method be also use by optimize for inference l82 l85 which lead to error when the model be load msg std 1 string input 0 of node dropout6 cond switch be pass float from placeholder 2 0 i...
tensorflowtensorflow
error in hessian and hessian vector product for tf nn sparse softmax cross entropy with logit
Bug
on a simple model that implement logistic regression construct the loss use tf nn sparse softmax cross entropy with logit make both hessian and hessian vector product return identically zero vector which be incorrect if I instead write the loss function manually use tf log tf sigmoid etc hessian...
tensorflowtensorflow
add zero sized layer cost huge amount of memory but without increase total of trainable parameter
Bug
in order to make code generic my code do append a layer with a variable sized layer which might be of length 0 I e tv concat tf concat 1 tf reshape tf slice x 0 time step 0 1 ncontext 1 tp current batch size embed size ncontext tv hh in the above code x be a tensor of shape b...
tensorflowtensorflow
tf with dependency isn t expose but it be use in documentation
Bug
there be some example that use tf with dependency ex tf assert in r0 11 assert but tf with dependency isn t expose to tensorflow module environment info branch master branch git rev parse head c7edafccc if possible provide a minimal reproducible example python3 c im...
tensorflowtensorflow
misformatte documentation after constant initializer
Bug
the doc for constant initializer fail to correctly close a code block in one of the example as a result many other section be display in raw markdown see constant initializer
tensorflowtensorflow
documentation for tf strided slice be nearly useless
Bug
the documentation for this op could almost be use as a teaching moment for how to not document something in order to understand what it do it be necessary to not only understand in detail how the relate numpy op work but also understand the mapping between the input of that op and the input of thetensorflow op ...
tensorflowtensorflow
optimization occur after graph be partition across device
Bug
the optimization sub expression constant folding of the graph take place after partition while this be important it prevent optimization of logic that be split across device by an unfortunate placement decision in my simple example constant be assign to my device graph and the downstream math node be...
tensorflowtensorflow
einsum ij ji rase error
Bug
hi I m run linux 64 bit os and I m get the follow error after run import tensorflow as tf I tf constant 1 0 0 0 1 0 0 0 1 ii tr tf einsum ij ji I I with tf session as sess re sess run ii tr print re assertionerror index have incorrect format ij ji c...
tensorflowtensorflow
don t imply top k be nondifferentiable
Bug
top k be in the evaluation evaluation section of the documentation which say the evaluation op be useful for measure the performance of a network since they be nondifferentiable they be typically use at evaluation time this be confuse since top k be differentiable point out by...
tensorflowtensorflow
api page could use some search optimization
Bug
when I search on google for tensorflow method the documentation be hard to distinguish in the search result the api page have irrelevant title and aren t as highly rank as they should be to reproduce search for tf extract image patch expect behavior I get a result with the api documentation obser...
tensorflowtensorflow
tensorflow crash when use large image with 3d convolutional network
Bug
I m try to implement a 3d fully convolutional network on my gpu but for some reason I get a crash environment info operating system ubuntu 14 04 lt gpu geforce titan x instal version of cuda and cudnn 8 0 and 5 attach the output of ls l path to cuda lib libcud cud filelist txt ...
tensorflowtensorflow
possible typo in docstring of embed rnn seq2seq and embed attention seq2seq
Bug
so in the docstring of embed rnn seq2seq l296 it say this model first embed encoder input by a newly create embedding of shape num encoder symbol x input size then it run an rnn to encode should that instead read embed size x input size then it run an rnn to en...
tensorflowtensorflow
small typo in documentation
Bug
in the code example in replica device setter tf replica device setter be use instead of tf train replica device setter
tensorflowtensorflow
documentation show none default instead of actual op attr default
Bug
this be not a bug but a small mistake in the api the documentation assign say use lock default to none in tf assign that parameter be a boolean so none would not make a lot of sense in any case the source code make it clear the default value be actually false file tensorflow ...
tensorflowtensorflow
cifar10 multi gpu train py break with more than 1 gpu
Bug
environment info operating system ubuntu instal version of cuda and cudnn 8 0 and 5 1 the commit hash git rev parse head 3d41cf77d624aeee0482f92121a9300b29db2809 2 the output of bazel version build label 0 3 2 build target bazel out local fastbuild bin src main java com goo...
tensorflowtensorflow
documentation change for add a new op shapefn
Bug
in add a new op there be the follow line of code that describe how to set the shape function python tf registershape zeroout common shape call cpp shape fn I think this be wrong you do not want the sign since it s not be use as a decorator also have an explicit import would be h...
tensorflowtensorflow
add documentation on how to use bucketing function
Bug
can someone please add documentation on how to use tensorflow contrib training bucket and tensorflow contrib training bucket by sequence length function note only file github issue for bug and feature request all other topic will be close for general support from the community see stackoverfl...
tensorflowtensorflow
doc error relate to tf space to batch
Bug
in the api doc relate to tfspace to batchinput padding block size namenone space to batch tf space to batch the result of example case of 3 for the follow input of shape 1 4 4 1 and block size of 2 prettyprint x 1 2 3 4 5 6 7 8...
tensorflowtensorflow
io no opkernel to support truncatednormal
Bug
I have a nn similar to this one and my io project be base on the simple contrib example I write out the graph pb file like so and replace it with the original io one tf train write graph sess graph def nnmodel graph pb as text false but xcode crash on execution with this error run...
tensorflowtensorflow
inconsistent behavior for variable collection and output collection parameter contrib layer
Bug
use layer from contrib the output collection parameter accept either a string or list of string the variable collection parameter require a list of string grant the documentation do specify this variable collection optional list of collection for all the variable or a dictionay contain...
tensorflowtensorflow
constant fold doesn t remove control edge
Bug
I believe that when constant folding take place and a section of a graph be replace by a constant that only the datum output edge of the replace node be remove I believe that I can see that a graph of node end up in a div part of the gradient generation bit be replace by a const the output of the div ...
tensorflowtensorflow
sticky no opkernel be register to support op conv2d with these attrs
Bug
there be two issue here 1 tf nn conv2d do not support float64 tensor on cpu although the documentation state it should 2 after the first occurrence of the situation above tf nn conv2d keep crash with the same error even when we provide float32 tensor what related github issue or stackoverflow thre...
tensorflowtensorflow
supervisor should stop not work in tf distribute
Bug
run tensorflow 0 11 0rc2 and follow the example put it all together example trainer program the worker do not get a should stop signal from the master repro trainer py start
tensorflowtensorflow
deal with control flow context when copy op
Bug
in the current implementation of copy op both tf contrib copy graph and tf contrib graph editor the code of copy an op look like this python copy input input copy func op input copy control input control input copy func control input copy node def op def node def de...
tensorflowtensorflow
tensorflow core public session h file not find
Bug
I didn t find anything about this problem I m create the python package from the source with bazel the command use be the same of the official guide in the section create pip package create the pip package and install the file whl generate be work but doesn t have some include file that I use in a n...
tensorflowtensorflow
tensorboard not work with 0 11 rc1
Bug
all I see be a blank page when I go to localhost 6006 browser chrome version 54 0 2840 71 64 bit here s the full stack trace 127 0 0 1 02 nov 2016 08 59 41 get http 1 1 200 127 0 0 1 02 nov 2016 08 59 41 code 404 message not find 127 0 0 1 02 nov 2016 08 59 41 g...
tensorflowtensorflow
regression pyunicodeucs4 asutf8stre on 0 11 0rc0 to 0 11 0rc2
Bug
note only file github issue for bug and feature request all other topic will be close for general support from the community see stackoverflow to make bug and feature request more easy to find and organize we close issue that be deem out of scope for github issue and point people to stackoverflow ...
tensorflowtensorflow
filesystem fileexist interface be break
Bug
the filesystem define the follow interface which can t handle temporary runtime error properly but just swallow they like gcs hdfs and any customize dfs c virtual bool fileexist const string fname 0 it should be change to c virtual status fileexist const string fname 0 ...
tensorflowtensorflow
bug reduce logsumexp example do not work in 0 11 0rc1
Bug
version 0 11 0rc1 these example do not work x be 0 0 0 0 0 0 tf reduce logsumexp x 0 log 2 log 2 log 2 tf reduce logsumexp x 1 log 3 log 3 because tf squeeze t squeeze dim do not take integer for squeeze dim
tensorflowtensorflow
tf learn metricspec doesn t work with streaming mean
Bug
metricspec always pass both value and label to metric mean metric which don t use label like streaming mean be break
tensorflowtensorflow
tf contrib learn output shape shape 1 and be incompatible
Bug
I try to train a binary dnnclassifier similar as the example on tf contrib learn quickstart I get the follow traceback traceback txt I explore the tensorflow source code and find that it may relate to get in out shape function in tensorflow tensorflow contrib learn python learn learn io datum feeder py ...
tensorflowtensorflow
eigen implement cpu op be 10 time slow than openmp
Bug
I implement a phase cpu operator consist of four loop level I couldn t find any eigen tensor doc at that stage so I use openmp to trivially parallelise the outer loop I recently find the eigen tensor documentation so I think I d take advantage of it and get all the multithreade avx sse goody for free unfortunate...
tensorflowtensorflow
oom error message should show which gpu be out of memory
Bug
what related github issue or stackoverflow thread have you find by search the web for your problem I ve create a few overflow thread in regard to balance seq2seq memory load over multiple gpu here environment info operating system ubuntu 14 04 cuda 7 5 tensorflow 0 11 if possible provide...
tensorflowtensorflow
tf while loop seg fault upon set parallel iteration 0
Bug
note only file github issue for bug and feature request all other topic will be close for general support from the community see stackoverflow to make bug and feature request more easy to find and organize we close issue that be deem out of scope for github issue and point people to stackoverflow fo...
tensorflowtensorflow
let user to choice whether to do bazel clean during configure
Bug
currently run the configure will do bazel clean expunge by default it be ok for the first time to run configure but when I need to rerun the configure because some error occur I will retry I have to spend much time to download file from internet again each time during to the execution of bazel cle...
tensorflowtensorflow
attributeerror module object have no attribute save v2
Bug
note only file github issue for bug and feature request all other topic will be close for general support from the community see stackoverflow to make bug and feature request more easy to find and organize we close issue that be deem out of scope for github issue and point people to stackoverflow fo...
tensorflowtensorflow
mean squared error give warn about sum of squared error
Bug
I use tf contrib loss mean square error but during training stage tensorflow give I warn warn tensorflow sum of square from tensorflow contrib loss python loss loss op be deprecate and will be remove after 2016 10 01 but I already use mean squared error so this warning should not show also the mean ...
tensorflowtensorflow
tf nn softmax on gpu cause cuda error illegal address okay on cpu
Bug
cento 7 tensorflow 0 10 0 titan x pascal 367 44 I restore a model previously save with tf train saver and try to compute probability of output for a give input batch whenever I try to execute tf nn softmax on gpu I get an error e tensorflow stream executor cuda cuda driver cc 1140 could not...
tensorflowtensorflow
timeout break fifoqueue
Bug
ubuntu 14 04 5 lts 0 10 0rc0 use timeout with notebook be very useful in case you dequeue an empty queue or enqueue a full one the problem be that after a timeout occur a enqueue or dequeueop throw an error import tensorflow as tf with tf device cpu 0 ph tf placeholder tf float32 q tf...
tensorflowtensorflow
please make the documentation headline linkable clickable
Bug
that would make it easy to refer to particular part anchor in the documentation
tensorflowtensorflow
fastgfile resourceexhaustederror
Bug
in cshallue change some tf gfile fastgfile call to open because two redhat user report in 4685 that it cause resourceexhaustederror even though resource don t appear to be exhaust we ve also have the same issue report in tensorflow model 531 tensorflow model 489 and tensorflow model 480 but we r...
tensorflowtensorflow
slim namespace have non slim symbol
Bug
tensorflow contrib slim init py make symbol from unrelated namespace available in its own py from tensorflow contrib framework python ops arg scope import from tensorflow contrib framework python op variable import from tensorflow contrib layers python layers import from tensorflow contrib la...
tensorflowtensorflow
example digit not work
Bug
hello everyone I use the very last docker container gpu enable nvidia docker everything work fine for what I could have test so far host ubuntu 14 04 server with cuda and nvidia driver up to date except this example be not work anymore can t manage to make it work and understand the error i...
tensorflowtensorflow
tutorial on contrib load csv no long work please suggest solution
Bug
I try to follow the turotial here but I get the error 221 tensorflow tutorial I python contrib learn quick origin py traceback most recent call last file contrib learn quick origin py line 13 in training set tf contrib learn dataset base load csv filename iris train attributeerror ...
tensorflowtensorflow
saver return path that cause getmatchingpath to go dig around in parent dir
Bug
note this work without problem in 0 10 0rc0 default pip install ed accord to website environment info operating system ubuntu 15 10 instal version of cuda and cudnn please attach the output of ls l path to cuda lib libcud rw r r 1 root root 189170 jan 1 2016 usr local cu...
tensorflowtensorflow
tensorboard command break
Bug
run tensorboard result in this logdir os path expanduser flag logdir attributeerror nonetype object have no attribute logdir good practice in this case be arg opt parse
tensorflowtensorflow
multi gpu cifa10 example put op outside of towerloss to cpu actually hurt performance
Bug
from cifa 10 example tutorial in train py with tf graph as default tf device cpu 0 put all the op but in tower loss to cpu as in the tutorial cifa 10 tutorial cifar 10 model this setup require that all gpu share the model parameter a well know fact be that transfer datum to and from gpu...
tensorflowtensorflow
example mnist rnn not work with docker image
Bug
issue example mnist rnn do run on docker image importerror traceback most recent call last in 3 from future import print function 4 5 from sklearn import metric pr...
tensorflowtensorflow
method doc contain unclosed verbatim which cause exceed verbatim block in html doc
Bug
the html doc for tf constant initializer at constant initializer contain unclose verbatim block which continue past the method and contain several follow method see for example the tf random normal initializer I be not much familiar with the process of documentation generation but my guess be that ...
tensorflowtensorflow
request for documentation quantize training in python
Bug
poke around I find however there be no document that I can find describe how this can be use can we add some example and also documentation around it so that people can start to play with it
tensorflowtensorflow
request for documentation loop implementation
Bug
I be try to understand the implementation of tf while loop and everything that be build on top of it because I be implement a custom tf graph subclass and find that the way tf while loop be handle during gradient computation be important for what I be do however I can not find any documentation on...
tensorflowtensorflow
minor instruction issue in functional op py scan
Bug
I have be use functional op py scan for implement rnn I ve just check the instruction and think this be little bit wrong in master branch from line 419 in functional op py it say args fn the callable to be perform it accept two argument the first will have the same possibly n...
tensorflowtensorflow
atrous convolution do not preserve tensor shape
Bug
for an input with an undefined batch size atrous conv2d emit tensor where all except the final dimension be undefined python input tf placeholder tf float32 none 256 256 3 conv tf nn conv2d input tf zero 3 3 3 16 stride 1 1 1 1 padding same print conv get sha...
tensorflowtensorflow
feature request include function argument default in the documentation
Bug
I have recently start use function from tf contrib they speed thing up a lot thank you I do notice that some of the prototype in the document have their argument replace with args kwargs see tf contrib layer layer contrib perhaps this be automatically do if the number of argument exceed some nu...