Hello Benno,
+1
Tested on :
Ubuntu 18.04 with SSL
8 GPUs per server
NVIDIA-SMI 418.56
Tested gpu workload with: tensorflow
Image used for testing: tensorflow/tensorflow:1.13.1-gpu-py3
Result:
------- versions ------
DISTRIB_ID=Ubuntu
VERSION_ID="16.04"
driver_version
418.56
CUDA Version 10.0.130
tf version: 1.13.1
---------------
TensorFlow: 1.13
Model: resnet50
Dataset: imagenet (synthetic)
Mode: training
SingleSess: False
Batch size: 32 global
32 per device
Num batches: 500
Num epochs: 0.01
Devices: ['/gpu:0']
NUMA bind: False
Data format: NCHW
Optimizer: sgd
Variables: parameter_server
==========
Generating training model
Initializing graph
Running warm up
Done warm up
...
Executing pre-exec command
'{"arguments":["ln","-s","/sys/fs/cgroup/cpu,cpuacct","/data0/mesos/work/provisioner/containers/3b1ccd4e-e2d6-44ba-bf8d-f7b29881f6a6/backends/overlay/rootfses/e06cb46b-07e6-4e87-8b2d-fa9af29e298b/sys/fs/cgroup/cpuacct"],"shell":false,"value":"ln"}'
Executing pre-exec command
'{"arguments":["ln","-s","/sys/fs/cgroup/cpu,cpuacct","/data0/mesos/work/provisioner/containers/3b1ccd4e-e2d6-44ba-bf8d-f7b29881f6a6/backends/overlay/rootfses/e06cb46b-07e6-4e87-8b2d-fa9af29e298b/sys/fs/cgroup/cpu"],"shell":false,"value":"ln"}'
Changing root to
/data0/mesos/work/provisioner/containers/3b1ccd4e-e2d6-44ba-bf8d-f7b29881f6a6/backends/overlay/rootfses/e06cb46b-07e6-4e87-8b2d-fa9af29e298b
2019-05-02 07:16:57.039394: I
tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports
instructions that this TensorFlow binary was not compiled to use: AVX2 FMA
2019-05-02 07:16:57.250080: I tensorflow/compiler/xla/service/service.cc:150]
XLA service 0x4ec62d0 executing computations on platform CUDA. Devices:
2019-05-02 07:16:57.250152: I tensorflow/compiler/xla/service/service.cc:158]
StreamExecutor device (0): Tesla V100-PCIE-16GB, Compute Capability 7.0
2019-05-02 07:16:57.273117: I
tensorflow/core/platform/profile_utils/cpu_utils.cc:94] CPU Frequency:
2594200000 Hz
2019-05-02 07:16:57.277123: I tensorflow/compiler/xla/service/service.cc:150]
XLA service 0x503da70 executing computations on platform Host. Devices:
2019-05-02 07:16:57.277177: I tensorflow/compiler/xla/service/service.cc:158]
StreamExecutor device (0): <undefined>, <undefined>
2019-05-02 07:16:57.278024: I
tensorflow/core/common_runtime/gpu/gpu_device.cc:1433] Found device 0 with
properties:
name: Tesla V100-PCIE-16GB major: 7 minor: 0 memoryClockRate(GHz): 1.38
pciBusID: 0000:83:00.0
totalMemory: 15.75GiB freeMemory: 15.44GiB
2019-05-02 07:16:57.278046: I
tensorflow/core/common_runtime/gpu/gpu_device.cc:1512] Adding visible gpu
devices: 0
> On 29 Apr 2019, at 22:05, Benno Evers <bev...@mesosphere.com> wrote:
>
> Hi Jorge,
>
> I'm admittedly not too familiar with CUDA and tensorflow but the error
> message you describe sounds to me more like a build issue, i.e. it sounds
> like the version of the nvidia driver is different between the docker image
> and the host system?
>
> Maybe you could continue investigating to see if this is related to the
> release itself or caused by some external cause, and create a JIRA ticket to
> capture your findings?
>
> Thanks,
> Benno
>
> On Fri, Apr 26, 2019 at 9:55 PM Jorge Machado <jom...@me.com
> <mailto:jom...@me.com>> wrote:
> Hi all,
>
> did someone tested it on ubuntu 18.04 + nvidia-docker2 ? We are having some
> issues using the cuda 10+ images when doing real processing. We still need to
> check some things but basically we get:
> kernel version 418.56.0 does not match DSO version 410.48.0 -- cannot find
> working devices in this configuration
>
> Logs:
> I0424 13:27:14.000586 30 executor.cpp:726] Forked command at 73
> Preparing rootfs at
> '/data0/mesos/work/provisioner/containers/548d3cae-30b5-4530-a8db-f94b00215718/backends/overlay/rootfses/e1ceb89e-3abc-4587-a87c-d63037b7ae8b'
> Marked '/' as rslave
> Executing pre-exec command
> '{"arguments":["ln","-s","/sys/fs/cgroup/cpu,cpuacct","/data0/mesos/work/provisioner/containers/548d3cae-30b5-4530-a8db-f94b00215718/backends/overlay/rootfses/e1ceb89e-3abc-4587-a87c-d63037b7ae8b/sys/fs/cgroup/cpuacct"],"shell":false,"value":"ln"}'
> Executing pre-exec command
> '{"arguments":["ln","-s","/sys/fs/cgroup/cpu,cpuacct","/data0/mesos/work/provisioner/containers/548d3cae-30b5-4530-a8db-f94b00215718/backends/overlay/rootfses/e1ceb89e-3abc-4587-a87c-d63037b7ae8b/sys/fs/cgroup/cpu"],"shell":false,"value":"ln"}'
> Changing root to
> /data0/mesos/work/provisioner/containers/548d3cae-30b5-4530-a8db-f94b00215718/backends/overlay/rootfses/e1ceb89e-3abc-4587-a87c-d63037b7ae8b
> 2019-04-24 13:27:18.346994: I
> tensorflow/core/platform/cpu_feature_guard.cc:141
> <http://cpu_feature_guard.cc:141/>] Your CPU supports instructions that this
> TensorFlow binary was not compiled to use: AVX2 FMA
> 2019-04-24 13:27:18.352203: E
> tensorflow/stream_executor/cuda/cuda_driver.cc:300
> <http://cuda_driver.cc:300/>] failed call to cuInit: CUDA_ERROR_UNKNOWN:
> unknown error
> 2019-04-24 13:27:18.352243: I
> tensorflow/stream_executor/cuda/cuda_diagnostics.cc:161
> <http://cuda_diagnostics.cc:161/>] retrieving CUDA diagnostic information for
> host: __host__
> 2019-04-24 13:27:18.352252: I
> tensorflow/stream_executor/cuda/cuda_diagnostics.cc:168
> <http://cuda_diagnostics.cc:168/>] hostname: __host__
> 2019-04-24 13:27:18.352295: I
> tensorflow/stream_executor/cuda/cuda_diagnostics.cc:192
> <http://cuda_diagnostics.cc:192/>] libcuda reported version is: 410.48.0
> 2019-04-24 13:27:18.352329: I
> tensorflow/stream_executor/cuda/cuda_diagnostics.cc:196
> <http://cuda_diagnostics.cc:196/>] kernel reported version is: 418.56.0
> 2019-04-24 13:27:18.352338: E
> tensorflow/stream_executor/cuda/cuda_diagnostics.cc:306
> <http://cuda_diagnostics.cc:306/>] kernel version 418.56.0 does not match DSO
> version 410.48.0 -- cannot find working devices in this configuration
> 2019-04-24 13:27:18.374940: I
> tensorflow/core/platform/profile_utils/cpu_utils.cc:94
> <http://cpu_utils.cc:94/>] CPU Frequency: 2593920000 Hz
> 2019-04-24 13:27:18.378793: I tensorflow/compiler/xla/service/service.cc:150
> <http://service.cc:150/>] XLA service 0x4f41e10 executing computations on
> platform Host. Devices:
> 2019-04-24 13:27:18.378821: I tensorflow/compiler/xla/service/service.cc:158
> <http://service.cc:158/>] StreamExecutor device (0): <undefined>,
> <undefined>
> W0424 13:27:18.385210 140191267731200 deprecation.py:323] From
> /usr/local/lib/python2.7/dist-packages/tensorflow/python/framework/op_def_library.py:263:
> colocate_with (from tensorflow.python.framework.ops) is deprecated and will
> be removed in a future version.
> Instructions for updating:
> Colocations handled automatically by placer.
> W0424 13:27:18.399287 140191267731200 deprecation.py:323] From
> /user/tf-benchmarks-113/scripts/tf_cnn_benchmarks/convnet_builder.py:129:
> conv2d (from tensorflow.python.layers.convolutional) is deprecated and will
> be removed in a future version.
> Instructions for updating:
> Use keras.layers.conv2d instead.
> W0424 13:27:18.433226 140191267731200 deprecation.py:323] From
> /user/tf-benchmarks-113/scripts/tf_cnn_benchmarks/convnet_builder.py:261:
> max_pooling2d (from tensorflow.python.layers.pooling) is deprecated and will
> be removed in a future version.
> Instructions for updating:
> Use keras.layers.max_pooling2d instead.
> W0424 13:27:20.197937 140191267731200 deprecation.py:323] From
> /usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/losses/losses_impl.py:209:
> to_float (from tensorflow.python.ops.math_ops) is deprecated and will be
> removed in a future version.
> Instructions for updating:
> Use tf.cast instead.
> W0424 13:27:20.312573 140191267731200 deprecation.py:323] From
> /usr/local/lib/python2.7/dist-packages/tensorflow/python/ops/math_ops.py:3066:
> to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be
> removed in a future version.
> Instructions for updating:
> Use tf.cast instead.
> W0424 13:27:21.082763 140191267731200 deprecation.py:323] From
> /user/tf-benchmarks-113/scripts/tf_cnn_benchmarks/benchmark_cnn.py:2238:
> __init__ (from tensorflow.python.training.supervisor) is deprecated and will
> be removed in a future version.
> Instructions for updating:
> Please switch to tf.train.MonitoredTrainingSession
> I0424 13:27:22.013817 140191267731200 session_manager.py:491] Running
> local_init_op.
> I0424 13:27:22.193911 140191267731200 session_manager.py:493] Done running
> local_init_op.
> 2019-04-24 13:27:23.181740: E tensorflow/core/common_runtime/executor.cc:624
> <http://executor.cc:624/>] Executor failed to create kernel. Invalid
> argument: Default MaxPoolingOp only supports NHWC on device type CPU
> [[{{node tower_0/v/cg/mpool0/MaxPool}}]]
> I0424 13:27:23.262847 140191267731200 coordinator.py:224] Error reported to
> Coordinator: <class
> 'tensorflow.python.framework.errors_impl.InvalidArgumentError'>, Default
> MaxPoolingOp only supports NHWC on device type CPU
> [[node tower_0/v/cg/mpool0/MaxPool (defined at
> /user/tf-benchmarks-113/scripts/tf_cnn_benchmarks/convnet_builder.py:261) ]
> running this on nvidia-docker2 works fine.
> image used: tensorflow/tensorflow:latest-gpu
> command: python
> /user/tf-benchmarks-113/scripts/tf_cnn_benchmarks/tf_cnn_benchmarks.py
> --num_gpus=1 --batch_size=32 --model=resnet50
> --variable_update=parameter_server
> on the host nvidia-smi says: NVIDIA-SMI 418.56 Driver Version: 418.56
> CUDA Version: 10.1
> thx
> Jorge
>> On 26 Apr 2019, at 18:28, Benno Evers <bev...@mesosphere.com
>> <mailto:bev...@mesosphere.com>> wrote:
>>
>> Hi all,
>>
>> Please vote on releasing the following candidate as Apache Mesos 1.8.0.
>>
>>
>> 1.8.0 includes the following:
>> --------------------------------------------------------------------------------
>> * Greatly reduced allocator cycle time.
>> * Operation feedback for v1 schedulers.
>> * Per-framework minimum allocatable resources.
>> * New CLI subcommands `task attach` and `task exec`.
>> * New `linux/seccomp` isolator.
>> * Support for Docker v2 Schema2 manifest format.
>> * XFS quota for persistent volumes.
>> * **Experimental** Support for the new CSI v1 API.
>>
>> In addition, 1.8.0-rc2 includes the following changes:
>> ---------------------------------------------------------------------------------
>> * Docker manifest v2s2 config with image GC.
>> * Expanded `highlights` section in the CHANGELOG.
>>
>> In addition, 1.8.0-rc3 includes the following changes:
>> ---------------------------------------------------------------------------------
>> * Relaxed protobuf union validation strictness. (MESOS-9740)
>> * Fixed a bug causing non-uniform random results in the random sorter.
>> (MESOS-9733)
>>
>>
>> The CHANGELOG for the release is available at:
>> https://gitbox.apache.org/repos/asf?p=mesos.git;a=blob_plain;f=CHANGELOG;hb=1.8.0-rc3
>>
>> <https://gitbox.apache.org/repos/asf?p=mesos.git;a=blob_plain;f=CHANGELOG;hb=1.8.0-rc3>
>> --------------------------------------------------------------------------------
>>
>> The candidate for Mesos 1.8.0 release is available at:
>> https://dist.apache.org/repos/dist/dev/mesos/1.8.0-rc3/mesos-1.8.0.tar.gz
>> <https://dist.apache.org/repos/dist/dev/mesos/1.8.0-rc3/mesos-1.8.0.tar.gz>
>>
>> The tag to be voted on is 1.8.0-rc3:
>> https://gitbox.apache.org/repos/asf?p=mesos.git;a=commit;h=1.8.0-rc3
>> <https://gitbox.apache.org/repos/asf?p=mesos.git;a=commit;h=1.8.0-rc3>
>>
>> The SHA512 checksum of the tarball can be found at:
>> https://dist.apache.org/repos/dist/dev/mesos/1.8.0-rc3/mesos-1.8.0.tar.gz.sha512
>>
>> <https://dist.apache.org/repos/dist/dev/mesos/1.8.0-rc3/mesos-1.8.0.tar.gz.sha512>
>>
>> The signature of the tarball can be found at:
>> https://dist.apache.org/repos/dist/dev/mesos/1.8.0-rc3/mesos-1.8.0.tar.gz.asc
>>
>> <https://dist.apache.org/repos/dist/dev/mesos/1.8.0-rc3/mesos-1.8.0.tar.gz.asc>
>>
>> The PGP key used to sign the release is here:
>> https://dist.apache.org/repos/dist/release/mesos/KEYS
>> <https://dist.apache.org/repos/dist/release/mesos/KEYS>
>>
>> The JAR is in a staging repository here:
>> https://repository.apache.org/content/repositories/orgapachemesos-1253
>> <https://repository.apache.org/content/repositories/orgapachemesos-1253>
>>
>> Please vote on releasing this package as Apache Mesos 1.8.0!
>>
>> The vote is open until and passes if a majority of at least 3 +1 PMC votes
>> are cast.
>>
>> [ ] +1 Release this package as Apache Mesos 1.8.0
>> [ ] -1 Do not release this package because ...
>>
>> Thanks,
>> Benno and Joseph
>
>
>
> --
> Benno Evers
> Software Engineer, Mesosphere
