Skip to content
This repository has been archived by the owner on Aug 5, 2022. It is now read-only.

Commit

Permalink
task-gold_u1_release: mkl_gold_u1_release
Browse files Browse the repository at this point in the history
  • Loading branch information
pnoga authored and ptbuilder committed Nov 3, 2016
2 parents 29649bb + 331f9d5 commit e2c1e23
Show file tree
Hide file tree
Showing 9 changed files with 50 additions and 7,537 deletions.
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -40,7 +40,7 @@ to use more than one thread per core. When less than required cores are specifie
limit execution of OpenMP threads to specified cores only.

## Best performance solution
Please read [release notes](https://github.com/intel/caffe/blob/master/docs/release_notes.md) for our recommendations and configuration to achieve best performance on Intel CPUs.
Please read [our Wiki](https://github.com/intel/caffe/wiki/Recommendations-to-achieve-best-performance) for our recommendations and configuration to achieve best performance on Intel CPUs.

## Multinode Training
Intel® Distribution of Caffe* multi-node allows you to execute deep neural network training on multiple machines.
Expand Down
86 changes: 46 additions & 40 deletions docs/release_notes.md
Original file line number Diff line number Diff line change
@@ -1,39 +1,37 @@
#
# All modification made by Intel Corporation: © 2016 Intel Corporation
#
# All contributions by the University of California:
# Copyright (c) 2014, 2015, The Regents of the University of California (Regents)
# All rights reserved.
#
# All other contributions:
# Copyright (c) 2014, 2015, the respective contributors
# All rights reserved.
# For the list of contributors go to https://github.com/BVLC/caffe/blob/master/CONTRIBUTORS.md
#
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# * Redistributions of source code must retain the above copyright notice,
# this list of conditions and the following disclaimer.
# * 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.
# * Neither the name of Intel Corporation nor the names of its contributors
# may be used to endorse or promote products derived from this software
# without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS 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 THE COPYRIGHT OWNER OR CONTRIBUTORS 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.
#
All modification made by Intel Corporation: © 2016 Intel Corporation

All contributions by the University of California:
Copyright (c) 2014, 2015, The Regents of the University of California (Regents)
All rights reserved.

All other contributions:
Copyright (c) 2014, 2015, the respective contributors
All rights reserved.
For the list of contributors go to https://github.com/BVLC/caffe/blob/master/CONTRIBUTORS.md

Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions are met:

* Redistributions of source code must retain the above copyright notice,
this list of conditions and the following disclaimer.
* 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.
* Neither the name of Intel Corporation nor the names of its contributors
may be used to endorse or promote products derived from this software
without specific prior written permission.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
AND ANY EXPRESS 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 THE COPYRIGHT OWNER OR CONTRIBUTORS 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.
---

---
title: Release Notes
Expand Down Expand Up @@ -216,9 +214,14 @@ Berkeley Vision runs Caffe with K40s, K20s, and Titans including models at Image
There is an unofficial Windows port of Caffe at [niuzhiheng/caffe:windows](https://github.com/niuzhiheng/caffe). Thanks [@niuzhiheng](https://github.com/niuzhiheng)!

# Change log
03-11-2016
* integration with MKL2017 update1 (providing better performance solution)
* minor changes to provide optimal performance on default prototxt files describing topologies (for AlexNet, GoogleNet v2).
* fixed Dockerfiles - for Ubuntu and Centos.

1-09-2016
* added RNN support
* moved form MKL2017 beta update 1 engine to MKL2017 (providing better performance solution)
* moved form MKL2017 beta update 1 engine to MKL2017
* added official support for ResNet50, GoogleNet v2, VGG-19. (List of currenlty supported topologies: AlexNet, GoogleNet, GoogleNet v2, ResNet50, VGG-19)
* added official support for multinode on GoogleNet with MKL2017 engine
* added DataLayer optimizations
Expand All @@ -234,12 +237,15 @@ Workaround: For older processors use MKL2017 GEMM engine: set USE_MKL2017_AS_DEF
Workaround: Use GEMM engine in normalization layer (in prototxt file set `engine:=caffe` for that layer) for topologies that use LRN within channel like cifar.

* Performance results may be lower when Data Layer is provided in txt files (uncompressed list of jpg files)
Workaround: We recommend to always use LMDB Data Layer
Workaround: We recommend to always use compressed LMDB Data Layer

* LeNet, Cifar, Squeeznet currently are not optimized in terms of performance in Intel MKL2017
Workaround: better performance results might be achieved with GEMM engine: `set USE_MKL2017_AS_DEFAULT_ENGINE := 0` in `Makefile.config`.

* We observe convergence problems with some publicly presented hyper parameters (recommended for GPUs) for Googlenet and ResNet50. For CPU tuning of hyper parameters might be needed.
* We observe convergence problems with some publicly presented hyper parameters (recommended for GPUs) for Googlenet, ResNet50, VGG-19. For CPU tuning of hyper parameters might be needed.

* MKL2017 doesn't allow access to mean & variance statistics in batch normalization layer which prohibits their accumulation (in global-stats mode). This is affecting batch 1 scoring accuracy with topologies using batch normalization layer (resnet50, googlenet v2).
Workaround: use batch 32 or higher for accuracy measurements.


# Recommendations to achieve best performance
Expand Down Expand Up @@ -296,4 +302,4 @@ In folder `/examples/imagenet/` we provide scripts and instructions `readme.md`
Caffe is released under the [BSD 2-Clause license](https://github.com/BVLC/caffe/blob/master/LICENSE). The BVLC reference models are released for unrestricted use.

***
*Other names and brands may be claimed as the property of others
*Other names and brands may be claimed as the property of others
6 changes: 3 additions & 3 deletions external/mkl/prepare_mkl.sh
Original file line number Diff line number Diff line change
Expand Up @@ -56,10 +56,10 @@ echo $VERSION_LINE # Return Version Line
# MKL
DST=`dirname $0`
OMP=0
VERSION_MATCH=20160706
ARCHIVE_BASENAME=mklml_lnx_2017.0.0.20160801.tgz
VERSION_MATCH=20160906
ARCHIVE_BASENAME=mklml_lnx_2017.0.1.20161005.tgz
MKL_CONTENT_DIR=`echo $ARCHIVE_BASENAME | rev | cut -d "." -f 2- | rev`
GITHUB_RELEASE_TAG=self_containted_MKLGOLD
GITHUB_RELEASE_TAG=self_containted_MKLGOLD_u1
MKLURL="https://github.com/intel/caffe/releases/download/$GITHUB_RELEASE_TAG/$ARCHIVE_BASENAME"
# there are diffrent MKL lib to be used for GCC and for ICC
reg='^[0-9]+$'
Expand Down
Loading

0 comments on commit e2c1e23

Please sign in to comment.