search

slaclab / anarel-manage

Manage the conda based analysis release environments that include psana at LCLS.
Other
3 stars 3 forks source link

update tensroflow #25

Closed davidslac closed 7 years ago

davidslac commented 7 years ago

now that our gpu's have cuda 8, can update to latest tensorflow and cudnn

davidslac commented 7 years ago

from the dev-rhel7, made tf and tf-gpu environments. Now that they put it on pip, can change the build scripts. This is what happens when do pip natively:

(tf) (psreldev) psel701: /reg/g/psdm/sw/conda/manage/recipes/external $ pip install tensorflow
Collecting tensorflow
  Downloading tensorflow-1.0.0-cp27-cp27mu-manylinux1_x86_64.whl (44.1MB)
    100% |################################| 44.1MB 15kB/s 
Collecting mock>=2.0.0 (from tensorflow)
  Using cached mock-2.0.0-py2.py3-none-any.whl
Collecting numpy>=1.11.0 (from tensorflow)
  Using cached numpy-1.12.0-cp27-cp27mu-manylinux1_x86_64.whl
Collecting protobuf>=3.1.0 (from tensorflow)
  Downloading protobuf-3.2.0-cp27-cp27mu-manylinux1_x86_64.whl (5.6MB)
    100% |################################| 5.6MB 107kB/s 
Requirement already satisfied: wheel in /reg/g/psdm/sw/conda/inst/miniconda2-dev-rhel7/envs/tf/lib/python2.7/site-packages (from tensorflow)
Collecting six>=1.10.0 (from tensorflow)
  Using cached six-1.10.0-py2.py3-none-any.whl
Collecting funcsigs>=1; python_version < "3.3" (from mock>=2.0.0->tensorflow)
  Using cached funcsigs-1.0.2-py2.py3-none-any.whl
Collecting pbr>=0.11 (from mock>=2.0.0->tensorflow)
  Using cached pbr-1.10.0-py2.py3-none-any.whl
Requirement already satisfied: setuptools in /reg/g/psdm/sw/conda/inst/miniconda2-dev-rhel7/envs/tf/lib/python2.7/site-packages/setuptools-27.2.0-py2.7.egg (from protobuf>=3.1.0->tensorflow)
Installing collected packages: six, funcsigs, pbr, mock, numpy, protobuf, tensorflow
Successfully installed funcsigs-1.0.2 mock-2.0.0 numpy-1.12.0 pbr-1.10.0 protobuf-3.2.0 six-1.10.0 tensorflow-1.0.0

and for the gpu:

davidslac commented 7 years ago

tf 1.0.0, after creating a session, reports some interesting messages, it would be good to build tensorflow ourselves with optimized switches:

W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE3 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.1 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use SSE4.2 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use AVX2 instructions, but these are available on your machine and could speed up CPU computations.
W tensorflow/core/platform/cpu_feature_guard.cc:45] The TensorFlow library wasn't compiled to use FMA instructions, but these are available on your machine and could speed up CPU computations.