ctuning / ck-tensorflow

Collective Knowledge components for TensorFlow (code, data sets, models, packages, workflows):
http://cKnowledge.org
BSD 3-Clause "New" or "Revised" License
93 stars 26 forks source link

ck-tensorflow/program/image-classification-tf-py/classify.py is not executable. #74

Closed taskset closed 6 years ago

taskset commented 6 years ago

$ ck run program:tensorflow --cmd_key=classify

OS CK UOA: linux-64 (4258b5fe54828a50)

OS name: Ubuntu 16.04.4 LTS Short OS name: Linux 4.15.0 Long OS name: Linux-4.15.0-29-generic-x86_64-with-debian-stretch-sid OS bits: 64 OS ABI: x86_64

Platform init UOA: -


Current directory: /home/hello/CK/ck-tensorflow/program/tensorflow/tmp


Resolving software dependencies ...

*** Dependency 1 = lib-tensorflow (TensorFlow library):

Resolved. CK environment UID = 28433f8d33a6f186

*** Dependency 2 = tensorflow-model (TensorFlow model (net and weights)):

Resolved. CK environment UID = c8f9d4473da5995d (version 20151205)

More than one dataset entry is found for this program:

0) image-jpeg-0001 (1aaaa23c44e588f9) 1) image-jpeg-dnn-cat (e496fdf046e6ac13) 2) image-jpeg-dnn-cat-gray (b6edb4314645efa0) 3) image-jpeg-dnn-computer-mouse (04df4936ba03c285) 4) image-jpeg-dnn-cropped-panda (0a4f26aa98034fd2) 5) image-jpeg-dnn-fish-bike (ebd96e7048ba3f7f) 6) image-jpeg-dnn-snake-224 (08d57fd1a6443ee9) 7) image-jpeg-dnn-surfers (b04c3400ba69fdc6) 8) image-jpeg-fgg (fgg, 21500da03e4ccfa3)

Select UOA (or press Enter for 0): 0

Cleaning output files and directories: stderr.log stderr2.log stderr.log stderr2.log



Prepared script:

! /bin/bash

. /home/hello/CK/local/env/28433f8d33a6f186/env.sh . /home/hello/CK/local/env/c8f9d4473da5995d/env.sh

export CK_DATASET_PATH=/home/hello/CK/ctuning-datasets-min/dataset/image-jpeg-0001/

export BATCH_SIZE=5 export CK_DATASET_FILENAME=data.jpg export NUM_BATCHES=5

echo executing code ... ${CK_ENV_COMPILER_PYTHON_FILE} ../classify.py --model_dir=${CK_ENV_MODEL_TENSORFLOW} --image_file=/home/hello/CK/ctuning-datasets-min/dataset/image-jpeg-0001/data.jpg > stderr.log 2> stderr2.log


(bash -c "chmod 755 ./tmp-iIf2Zt.sh; . ./tmp-iIf2Zt.sh")

(sleep 0.5 sec ...)

(run ...) executing code ...

(printing output files)

* stderr2.log

  ./tmp-iIf2Zt.sh: line 15: ../classify.py: Permission denied

* stderr.log

Execution time: 0.011 sec.

Program execution likely failed (return code 126 !=0 )!

we've check this file: $ ls ck-tensorflow/program/image-classification-tf-py/ -lh total 36K -rwxrwxr-x 1 hello hello 8.9K 6月 29 21:24 benchmark.nvidia-gtx1080.py -rwxrwxr-x 1 hello hello 8.9K 6月 29 21:24 benchmark.nvidia-tx1.py -rw-rw-r-- 1 hello hello 6.5K 6月 29 21:24 classify.py -rw-rw-r-- 1 hello hello 3.7K 6月 29 21:24 README.md

classify.py is not executable.

psyhtest commented 6 years ago

@w-simon Thanks for reporting! Let's try to fix this together.

In fact, classify.py shouldn't be executable. The problem is that ${CK_ENV_COMPILER_PYTHON_FILE} is not defined on your environment, because Python is not explicitly set as a dependency if this program. I guess we usually don't see this error, because we often use TensorFlow built from source (which depends on Python so Python does get defined transitively). Let me check in a fix...

gfursin commented 6 years ago

Hi @w-simon . Thank you for reporting the problem - I reproduced it. @psyhtest correctly said that the problem is actually because Python is not detected when you automatically detected your TF via CK. I believe I fixed the problem and checked it on my machine with latest TF 1.9.0. Can you please do the following to test it:

$ ck pull all
$ ck rm env:28433f8d33a6f186 -f
$  ck run program:tensorflow --cmd_key=classify

These commands will update all repos including my fix, remove bad env with currently detected TF, and will let you to select a CK TensorFlow package (CPU or GPU) to be automatically installed by the CK. I tested with 1.9.0 cuda and it now worked on my machine ... Thanks a lot!

taskset commented 6 years ago

@gfursin Thank you very much! This had been fixed and working well.

gfursin commented 6 years ago

Great! Thanks @w-simon! Glad to hear that!