platform/upstream/caffeonacl.git
8 years agoUse input_shape instead of input_dim in examples
Luke Yeager [Thu, 20 Aug 2015 21:29:02 +0000 (14:29 -0700)]
Use input_shape instead of input_dim in examples

8 years agoMerge pull request #2947 from BlGene/bdl_fixup
Jeff Donahue [Thu, 20 Aug 2015 20:16:14 +0000 (13:16 -0700)]
Merge pull request #2947 from BlGene/bdl_fixup

BasePrefetchingDataLayer fixup.

8 years agoMerge pull request #2930 from lukeyeager/pycaffe-layer_type_list
Jeff Donahue [Thu, 20 Aug 2015 18:04:05 +0000 (11:04 -0700)]
Merge pull request #2930 from lukeyeager/pycaffe-layer_type_list

Expose LayerFactory::LayerTypeList in pycaffe

8 years agoExpose LayerFactory::LayerTypeList in pycaffe
Luke Yeager [Fri, 14 Aug 2015 23:53:39 +0000 (16:53 -0700)]
Expose LayerFactory::LayerTypeList in pycaffe

Useful for validating NetParameters without crashing on SIGABRT

8 years agoIn BasePrefetchingDataLayer::Forward_cpu hanged top[0]->Reshape to
max argus [Thu, 20 Aug 2015 09:01:58 +0000 (09:01 +0000)]
In BasePrefetchingDataLayer::Forward_cpu hanged top[0]->Reshape to
top[0]->ReshapeLike, in line with other calls.

8 years agoMerge pull request #2812 from philkr/python_loss_weight
Jon Long [Mon, 17 Aug 2015 03:47:33 +0000 (20:47 -0700)]
Merge pull request #2812 from philkr/python_loss_weight

Expose blob loss weight to python

8 years agoMalloc at least 1 byte for MultiGPU P2PSync buffers
Ronghang Hu [Sun, 16 Aug 2015 02:02:57 +0000 (19:02 -0700)]
Malloc at least 1 byte for MultiGPU P2PSync buffers

8 years agoMerge pull request #2931 from ronghanghu/fix-gpu-test
Ronghang Hu [Sat, 15 Aug 2015 16:42:54 +0000 (09:42 -0700)]
Merge pull request #2931 from ronghanghu/fix-gpu-test

Fix MultiGPU solver test crash

8 years agoFix MultiGPU solver test with TEST_GPUID != 1
Ronghang Hu [Fri, 14 Aug 2015 20:52:01 +0000 (13:52 -0700)]
Fix MultiGPU solver test with TEST_GPUID != 1

This is a patch for multi-gpu testing issue (#2926). The problem fixed in this
commit is that when calling make runtest with TEST_GPUID != 0 on a MultiGPU
machine, solver tests will crash because gpu ids in multi-gpu tests doesn't
match that of single GPU test.

8 years agoMerge pull request #2928 from cypof/lib_make_target
Jeff Donahue [Fri, 14 Aug 2015 23:42:26 +0000 (16:42 -0700)]
Merge pull request #2928 from cypof/lib_make_target

New make target to only build the library.

8 years agoNew make target to only build the library.
Cyprien Noel [Fri, 14 Aug 2015 23:22:34 +0000 (16:22 -0700)]
New make target to only build the library.

8 years agoExposing blob loss weight to python
philkr [Thu, 23 Jul 2015 15:33:58 +0000 (08:33 -0700)]
Exposing blob loss weight to python

8 years agoMerge pull request #2925 from flx42/fix_classification_example
Jeff Donahue [Fri, 14 Aug 2015 18:54:06 +0000 (11:54 -0700)]
Merge pull request #2925 from flx42/fix_classification_example

Make classification.bin support models with less than 5 classes

8 years agoMake classification.bin support models with less than 5 classes
Felix Abecassis [Fri, 14 Aug 2015 18:15:56 +0000 (11:15 -0700)]
Make classification.bin support models with less than 5 classes

The example program would crash if the number of classes was less than
5, since it was still attempting to get the top 5 predictions.

Close #2585

8 years agoMerge pull request #2921 from buaaliyi/multi_gpu
Ronghang Hu [Fri, 14 Aug 2015 17:54:11 +0000 (10:54 -0700)]
Merge pull request #2921 from buaaliyi/multi_gpu

Destroy CUDA stream when finished

8 years agoMerge pull request #2924 from ronghanghu/fix-malloc-empty
Ronghang Hu [Fri, 14 Aug 2015 17:03:58 +0000 (10:03 -0700)]
Merge pull request #2924 from ronghanghu/fix-malloc-empty

Malloc at least one byte in Parallel

8 years agoMalloc at least one byte in Parallel
Ronghang Hu [Fri, 14 Aug 2015 16:19:48 +0000 (09:19 -0700)]
Malloc at least one byte in Parallel

8 years agoMerge pull request #2920 from PatWie/master
Ronghang Hu [Fri, 14 Aug 2015 16:14:28 +0000 (09:14 -0700)]
Merge pull request #2920 from PatWie/master

information about new implemented solvers

8 years agoMerge pull request #2918 from ronghanghu/adam
Ronghang Hu [Fri, 14 Aug 2015 16:12:45 +0000 (09:12 -0700)]
Merge pull request #2918 from ronghanghu/adam

Adam solver

8 years agoCite Adam paper in solver.hpp
Ronghang Hu [Fri, 14 Aug 2015 05:41:21 +0000 (22:41 -0700)]
Cite Adam paper in solver.hpp

8 years agoAdam solver
PatWie [Mon, 3 Aug 2015 15:31:14 +0000 (17:31 +0200)]
Adam solver

This commit implements the Adam solver by Kingma et. al for CPU and
GPU. All solver parameters are defined in the caffe.proto. This also
adds an example for the MNIST dataset.

8 years agoDestroy CUDA stream when finished
LI Yi [Fri, 14 Aug 2015 12:51:45 +0000 (20:51 +0800)]
Destroy CUDA stream when finished

8 years agoinformation about new implemented solvers
PatWie [Fri, 14 Aug 2015 11:06:13 +0000 (13:06 +0200)]
information about new implemented solvers

8 years agoMerge pull request #2572 from flx42/optimize_inner_product_special_case
Jeff Donahue [Fri, 14 Aug 2015 04:35:13 +0000 (21:35 -0700)]
Merge pull request #2572 from flx42/optimize_inner_product_special_case

Optimize inner product layer for special case M == 1

8 years agoMerge pull request #2903 from ronghanghu/multi_gpu
Ronghang Hu [Thu, 13 Aug 2015 20:28:11 +0000 (13:28 -0700)]
Merge pull request #2903 from ronghanghu/multi_gpu

Multi-GPU Data Parallelism

8 years agoApply mutex only to shared layers and fix NVCC warning
Ronghang Hu [Wed, 12 Aug 2015 19:05:56 +0000 (12:05 -0700)]
Apply mutex only to shared layers and fix NVCC warning

8 years agoData Layers Parallel for Multi-GPU
Ronghang Hu [Wed, 12 Aug 2015 04:38:06 +0000 (21:38 -0700)]
Data Layers Parallel for Multi-GPU

Allow data layers (and also PythonLayer when used as data layer) to be shared
among worker solver's training net, and also test net for future-proof if one
wants to do Multi-GPU testing. Data layers are locked during forward to ensure
sequential forward.

8 years agoMerge pull request #2909 from jeffdonahue/learnable-param-id-fix
Jeff Donahue [Tue, 11 Aug 2015 21:02:56 +0000 (14:02 -0700)]
Merge pull request #2909 from jeffdonahue/learnable-param-id-fix

learnable_param_ids_ fix

8 years agoMerge pull request #2891 from fyu/fix-caffenetpy
Jon Long [Tue, 11 Aug 2015 20:58:41 +0000 (13:58 -0700)]
Merge pull request #2891 from fyu/fix-caffenetpy

fix examples/pycaffe/caffenet.py import

8 years agofix for learnable_param_ids_
Jeff Donahue [Tue, 11 Aug 2015 20:18:41 +0000 (13:18 -0700)]
fix for learnable_param_ids_

8 years agoMerge pull request #2782 from matthiasplappert/adadelta
Ronghang Hu [Tue, 11 Aug 2015 07:35:00 +0000 (00:35 -0700)]
Merge pull request #2782 from matthiasplappert/adadelta

AdaDelta Solver (v3)

8 years agoMerge pull request #2873 from gut/master
Evan Shelhamer [Tue, 11 Aug 2015 01:54:20 +0000 (18:54 -0700)]
Merge pull request #2873 from gut/master

Fix gist download model and binary scripts

8 years agoMerge pull request #2897 from Russell91/patch-1
Jeff Donahue [Mon, 10 Aug 2015 21:30:04 +0000 (14:30 -0700)]
Merge pull request #2897 from Russell91/patch-1

Update net_spec.py

8 years agoUpdate net_spec.py
Russell Stewart [Mon, 10 Aug 2015 20:50:01 +0000 (13:50 -0700)]
Update net_spec.py

8 years agoClean up and modernize AdaDelta code; add learning rate support; add additional test...
Matthias Plappert [Sat, 18 Jul 2015 16:46:51 +0000 (18:46 +0200)]
Clean up and modernize AdaDelta code; add learning rate support; add additional test cases

8 years agoUpdated AdaDelta for modern Caffe; reduced iterations on multi-iter tests
Kevin Bache [Thu, 19 Mar 2015 22:56:51 +0000 (15:56 -0700)]
Updated AdaDelta for modern Caffe; reduced iterations on multi-iter tests

8 years agoImplement AdaDelta; add test cases; add mnist examples
Mohamed Omran [Sat, 20 Sep 2014 17:01:28 +0000 (19:01 +0200)]
Implement AdaDelta; add test cases; add mnist examples

8 years ago[docs] add multi-gpu usage note to interfaces
Evan Shelhamer [Fri, 7 Aug 2015 20:56:49 +0000 (13:56 -0700)]
[docs] add multi-gpu usage note to interfaces

8 years agoDetect topology corner cases and improve broadcast order
mhouston [Fri, 10 Jul 2015 23:05:48 +0000 (16:05 -0700)]
Detect topology corner cases and improve broadcast order

- Start with distant nodes in broadcast
- Fix outside loop to loop for full tree depth

8 years agoMulti-GPU
Cyprien Noel [Tue, 19 May 2015 18:11:05 +0000 (11:11 -0700)]
Multi-GPU

- Parallelize batches among GPUs and tree-reduce the gradients
- The effective batch size scales with the number of devices
- Batch size is multiplied by the number of devices
- Split batches between GPUs, and tree-reduce the gradients
- Detect machine topology (twin-GPU boards, P2P connectivity)
- Track device in syncedmem (thanks @thatguymike)
- Insert a callback in the solver for minimal code change
- Accept list for gpu flag of caffe tool, e.g. '-gpu 0,1' or '-gpu all'.
  Run on default GPU if no ID given.
- Add multi-GPU solver test
- Deterministic architecture for reproducible runs

8 years agoAllocate host memory through cudaMallocHost
Cyprien Noel [Tue, 19 May 2015 03:07:36 +0000 (20:07 -0700)]
Allocate host memory through cudaMallocHost

thanks to discussion by @thatguymike and @flx42

8 years agoAdd DataReader for parallel training with one DB session
Cyprien Noel [Tue, 19 May 2015 01:06:09 +0000 (18:06 -0700)]
Add DataReader for parallel training with one DB session

- Make sure each solver accesses a different subset of the data
- Sequential reading of DB for performance
- Prefetch a configurable amount of data to host memory
- Distribute data to solvers in round-robin way for determinism

8 years agoPersistent prefetch thread
Cyprien Noel [Tue, 19 May 2015 00:45:20 +0000 (17:45 -0700)]
Persistent prefetch thread

8 years agoChange the way threads are started and stopped
Cyprien Noel [Tue, 28 Apr 2015 21:46:20 +0000 (14:46 -0700)]
Change the way threads are started and stopped

- Interrupt the thread before waiting on join
- Provide a method for looping threads to exit on demand
- CHECK if start and stop succeed instead of returning an error

8 years agoThread-local Caffe
Cyprien Noel [Tue, 28 Apr 2015 02:48:10 +0000 (19:48 -0700)]
Thread-local Caffe

8 years agoAdd BlockingQueue for inter-thread communication
Cyprien Noel [Tue, 28 Apr 2015 21:28:04 +0000 (14:28 -0700)]
Add BlockingQueue for inter-thread communication

8 years agofrom __future__ imports must occur at the beginning of the file
Fisher Yu [Sun, 9 Aug 2015 21:22:04 +0000 (14:22 -0700)]
from __future__ imports must occur at the beginning of the file

8 years agoMerge pull request #2887 from shelhamer/solver-test-data
Evan Shelhamer [Sun, 9 Aug 2015 19:35:04 +0000 (12:35 -0700)]
Merge pull request #2887 from shelhamer/solver-test-data

Test solvers on fixed hdf5 data

8 years agoUse net_->learnable_params() instead of net_->params() in RMSprop
Ronghang Hu [Sun, 9 Aug 2015 18:03:23 +0000 (11:03 -0700)]
Use net_->learnable_params() instead of net_->params() in RMSprop

In RMSProp solver, use const vector<Blob<Dtype>*>& net_params = this->net_->learnable_params(); instead of const vector<shared_ptr<Blob<Dtype> > >& net_params = this->net_->params();

8 years agoMerge pull request #2888 from ronghanghu/rms-prop-fix-tests
Ronghang Hu [Sun, 9 Aug 2015 08:45:09 +0000 (01:45 -0700)]
Merge pull request #2888 from ronghanghu/rms-prop-fix-tests

Encapsulate kRMSDecay in solver tests

8 years agoEncapsulate kRMSDecay in solver tests
Ronghang Hu [Sun, 9 Aug 2015 08:32:25 +0000 (01:32 -0700)]
Encapsulate kRMSDecay in solver tests

Instead of introducing another argument kRMSDecay and setting it for every test, this param could be set by the RMSProp test class for encapsulation.

8 years agoMerge pull request #2867 from ronghanghu/rms-prop
Ronghang Hu [Sun, 9 Aug 2015 07:35:27 +0000 (00:35 -0700)]
Merge pull request #2867 from ronghanghu/rms-prop

RMSProp clean up and rebase

8 years agoImplement RMSProp Solver
Eren Golge [Sun, 9 Aug 2015 06:45:08 +0000 (23:45 -0700)]
Implement RMSProp Solver

Implement RMSProp solver and cleaned up to adjust to new solver interface that uses
accumulated gradients and refactored regularization.

8 years agoTestGradientBasedSolver: replace dummy data with hdf5
Evan Shelhamer [Sat, 8 Aug 2015 20:07:13 +0000 (13:07 -0700)]
TestGradientBasedSolver: replace dummy data with hdf5

Rely on fixed hdf5 data for determinism of solver tests.

- draw random Gaussian data and targets for test and save to hdf5
- use the same data for all tests without constant / gaussian condition
  previously needed for accumulation
- avoid test artifacts due to order of random draws in dummy data

8 years agoTestGradientBasedSolver: drop doubled seed inititialization
Evan Shelhamer [Sat, 8 Aug 2015 19:57:45 +0000 (12:57 -0700)]
TestGradientBasedSolver: drop doubled seed inititialization

8 years agoMerge pull request #2813 from longjon/net-spec-imp
Evan Shelhamer [Sat, 8 Aug 2015 04:06:29 +0000 (21:06 -0700)]
Merge pull request #2813 from longjon/net-spec-imp

Python net spec cleanups and top-less layers

8 years agoMerge pull request #2886 from jeffdonahue/master
Jeff Donahue [Sat, 8 Aug 2015 03:44:43 +0000 (20:44 -0700)]
Merge pull request #2886 from jeffdonahue/master

temporarily switch the snapshot_format default back to BINARYPROTO

8 years agotemporarily switch the snapshot_format default back to BINARYPROTO
Jeff Donahue [Sat, 8 Aug 2015 03:03:04 +0000 (20:03 -0700)]
temporarily switch the snapshot_format default back to BINARYPROTO

out of anticipation for user issues due to issue #2885, which causes
Caffe to crash when it attempts to snapshot nets with duplicate layer
names

8 years agoMerge pull request #2877 from jeffdonahue/pycaffe-shape-accessor
Jon Long [Fri, 7 Aug 2015 23:28:23 +0000 (16:28 -0700)]
Merge pull request #2877 from jeffdonahue/pycaffe-shape-accessor

[pycaffe] add Blob shape accessor

8 years agoMerge pull request #2866 from jeffdonahue/fix-weight-sharing
Jeff Donahue [Fri, 7 Aug 2015 23:05:00 +0000 (16:05 -0700)]
Merge pull request #2866 from jeffdonahue/fix-weight-sharing

Fix weight sharing

8 years agoNet: add learnable_params_ used by solvers to correctly handle shared params
Jeff Donahue [Tue, 21 Jul 2015 03:50:50 +0000 (20:50 -0700)]
Net: add learnable_params_ used by solvers to correctly handle shared params

-Params now share diffs as well as data (works due to layers
accumulating gradients into param diffs, rather than overwriting)
-It's now required that any shared params with specified lr_mult's,
decay_mult's match
-TestGradientBasedSolver checks that behavior remains correct with
shared weights

8 years agoTestGradientBasedSolver: make tests across solver types more consistent
Jeff Donahue [Fri, 7 Aug 2015 22:34:00 +0000 (15:34 -0700)]
TestGradientBasedSolver: make tests across solver types more consistent

8 years agoTestGradientBasedSolver: restore Gaussian filler to all tests except
Jeff Donahue [Fri, 7 Aug 2015 22:27:02 +0000 (15:27 -0700)]
TestGradientBasedSolver: restore Gaussian filler to all tests except
accumulation one

8 years agoMerge pull request #2836 from erictzeng/hdf5_snapshot
Jeff Donahue [Fri, 7 Aug 2015 22:22:41 +0000 (15:22 -0700)]
Merge pull request #2836 from erictzeng/hdf5_snapshot

Snapshot model weights/solver state to HDF5 files

8 years agoUpdate example bash scripts to expect .h5, new extensions in .gitignore
Eric Tzeng [Fri, 7 Aug 2015 20:17:26 +0000 (13:17 -0700)]
Update example bash scripts to expect .h5, new extensions in .gitignore

8 years agoTestSnapshot expects .h5 snapshots, explicitly checks history.
Eric Tzeng [Thu, 30 Jul 2015 01:40:38 +0000 (18:40 -0700)]
TestSnapshot expects .h5 snapshots, explicitly checks history.

8 years agoSnapshot model weights/solver state to HDF5 files.
Eric Tzeng [Wed, 22 Jul 2015 23:17:01 +0000 (16:17 -0700)]
Snapshot model weights/solver state to HDF5 files.

Summary of changes:
- HDF5 helper functions were moved into a separate file util/hdf5.cpp
- hdf5_save_nd_dataset now saves n-d blobs, can save diffs instead of
  data
- Minor fix for memory leak in HDF5 functions (delete instead of
  delete[])
- Extra methods have been added to both Net/Solver enabling
  snapshotting and restoring from HDF5 files
- snapshot_format was added to SolverParameters, with possible values
  HDF5 or BINARYPROTO (default HDF5)
- kMaxBlobAxes was reduced to 32 to match the limitations of HDF5

8 years agopycaffe: add shape accessor
Jeff Donahue [Thu, 6 Aug 2015 20:46:11 +0000 (13:46 -0700)]
pycaffe: add shape accessor

8 years agoMerge pull request #2884 from tianzhi0549/patch-1
Jon Long [Fri, 7 Aug 2015 20:51:52 +0000 (13:51 -0700)]
Merge pull request #2884 from tianzhi0549/patch-1

add [] to "delete pixels".

8 years agoTestGradientBasedSolver: add TestSnapshot to verify behavior when
Jeff Donahue [Thu, 30 Jul 2015 00:27:58 +0000 (17:27 -0700)]
TestGradientBasedSolver: add TestSnapshot to verify behavior when
restoring net/solver from snapshot

8 years agoadd double_data, double_diff to BlobProto for weights/snapshots saved
Jeff Donahue [Thu, 30 Jul 2015 00:27:04 +0000 (17:27 -0700)]
add double_data, double_diff to BlobProto for weights/snapshots saved
when using Dtype == double

8 years agoMerge pull request #2634 from mlopezantequera/patch-2
Jeff Donahue [Fri, 7 Aug 2015 18:58:19 +0000 (11:58 -0700)]
Merge pull request #2634 from mlopezantequera/patch-2

Update parse_log.py

8 years agoadd [] to "delete pixels".
Tian Zhi [Fri, 7 Aug 2015 18:31:19 +0000 (02:31 +0800)]
add [] to "delete pixels".

see https://isocpp.org/wiki/faq/freestore-mgmt#delete-array-built-ins.

8 years agoMerge pull request #2880 from koki0702/typo
Evan Shelhamer [Fri, 7 Aug 2015 16:23:20 +0000 (09:23 -0700)]
Merge pull request #2880 from koki0702/typo

[docs] Fix layer typo

8 years agoFix typo
koki1.saitoh [Fri, 7 Aug 2015 09:22:28 +0000 (18:22 +0900)]
Fix typo

8 years agoMerge pull request #2871 from shelhamer/python-layer-arg
Evan Shelhamer [Fri, 7 Aug 2015 01:41:02 +0000 (18:41 -0700)]
Merge pull request #2871 from shelhamer/python-layer-arg

[pycaffe] PythonLayer takes parameters by string

8 years agoPythonLayer takes parameters by string
Takuya Narihira [Sat, 28 Feb 2015 03:34:51 +0000 (19:34 -0800)]
PythonLayer takes parameters by string

8 years agoMerge pull request #2876 from shelhamer/fix-py-include
Evan Shelhamer [Thu, 6 Aug 2015 20:04:42 +0000 (13:04 -0700)]
Merge pull request #2876 from shelhamer/fix-py-include

[pycaffe,build] include Python first in caffe tool

8 years ago[pytest] open exception file with mode for python3
Evan Shelhamer [Thu, 6 Aug 2015 20:04:15 +0000 (13:04 -0700)]
[pytest] open exception file with mode for python3

8 years ago[pycaffe,build] include Python first in caffe tool
Evan Shelhamer [Thu, 6 Aug 2015 20:03:50 +0000 (13:03 -0700)]
[pycaffe,build] include Python first in caffe tool

8 years agoFix download model binary script to get correct lines on parsing table
Gustavo Serra Scalet [Thu, 6 Aug 2015 17:37:20 +0000 (14:37 -0300)]
Fix download model binary script to get correct lines on parsing table

The base reference of "bottom" variable was relative to the "top+1" and
not to the whole readlines output. It ended up without all the lines.

That could work for some gists however for the model I was looking for (see
below) the sha1 key was not being parsed, as it was missing the last line.

tested with the following gist:
longjon/1bf3aa1e0b8e788d7e1d

8 years agoFix download model script to use zip archive
Gustavo Serra Scalet [Thu, 6 Aug 2015 17:20:55 +0000 (14:20 -0300)]
Fix download model script to use zip archive

Currently GitHub is not using tarballs as archive for downloading gists
therefore the script was broken as actually a zip archive was being downloaded.

8 years agoMerge pull request #2462 from longjon/correct-python-exceptions
Evan Shelhamer [Thu, 6 Aug 2015 07:27:59 +0000 (00:27 -0700)]
Merge pull request #2462 from longjon/correct-python-exceptions

Handle Python layer exceptions correctly

8 years agoMerge pull request #2859 from philkr/image_data
Evan Shelhamer [Tue, 4 Aug 2015 18:57:51 +0000 (11:57 -0700)]
Merge pull request #2859 from philkr/image_data

ImageData layer default batch size of 1, and check for zero batch size

8 years agoImageData layer default batch size of 1, and check for zero batch size
philkr [Tue, 4 Aug 2015 17:27:34 +0000 (10:27 -0700)]
ImageData layer default batch size of 1, and check for zero batch size

8 years agoMerge pull request #2583 from lukeyeager/fix-log-levels
Jeff Donahue [Mon, 3 Aug 2015 23:19:47 +0000 (16:19 -0700)]
Merge pull request #2583 from lukeyeager/fix-log-levels

Change log levels in upgrade_proto

8 years agoChange log levels in upgrade_proto
Luke Yeager [Wed, 10 Jun 2015 00:50:40 +0000 (17:50 -0700)]
Change log levels in upgrade_proto

It's not an error if the upgrade succeeds

8 years agoMerge pull request #2837 from longjon/contributing-file
Jon Long [Sun, 2 Aug 2015 21:34:10 +0000 (14:34 -0700)]
Merge pull request #2837 from longjon/contributing-file

Add a CONTRIBUTING.md file, to appear on new Issue/PR pages

8 years ago[docs] add CONTRIBUTING.md which will appear on GitHub new Issue/PR pages
Jonathan L Long [Thu, 30 Jul 2015 03:16:31 +0000 (20:16 -0700)]
[docs] add CONTRIBUTING.md which will appear on GitHub new Issue/PR pages

8 years ago[docs] fix contrastive loss eq
Evan Shelhamer [Thu, 30 Jul 2015 00:20:31 +0000 (17:20 -0700)]
[docs] fix contrastive loss eq

make documented equation match the correct implementation of the
`max(margin - d, 0)^2` term in the loss. see #2321

8 years ago[docs] fix lmdb fetch url and path
Evan Shelhamer [Tue, 28 Jul 2015 17:27:39 +0000 (10:27 -0700)]
[docs] fix lmdb fetch url and path

8 years ago[docs] clear up PYTHONPATH confusion
Evan Shelhamer [Mon, 27 Jul 2015 17:58:11 +0000 (10:58 -0700)]
[docs] clear up PYTHONPATH confusion

Use the same language as the installation page to explain the Python
module path.

reported by @sdemyanov

8 years ago[pytest] simple test of top-less layers
Jonathan L Long [Fri, 24 Jul 2015 03:41:40 +0000 (20:41 -0700)]
[pytest] simple test of top-less layers

8 years ago[pycaffe] net spec layers can have ntop=0
Jonathan L Long [Fri, 24 Jul 2015 03:40:02 +0000 (20:40 -0700)]
[pycaffe] net spec layers can have ntop=0

In this case, the Function is returned instead of a Top, which can be
assigned a name if desired.

Thanks @philkr for an earlier implementation of this.

8 years ago[pycaffe] allow layers to have names different from their first tops
Jonathan L Long [Fri, 24 Jul 2015 03:35:42 +0000 (20:35 -0700)]
[pycaffe] allow layers to have names different from their first tops

Previously, net spec only allowed names to be assigned to Tops, giving
layers the names of their first tops. Now, names can be assigned to
Functions, which become layer names in serialization. Unnamed Functions
still get named after their first top, if present, or autogenerated, if
not. (This will allow top-less layers in a natural way.)

8 years ago[pycaffe] add Top._to_proto convenience function
Jonathan L Long [Fri, 24 Jul 2015 03:32:04 +0000 (20:32 -0700)]
[pycaffe] add Top._to_proto convenience function

This makes it possible to serialize Functions or Tops with a uniform
interface.

8 years ago[pycaffe] use a Counter instead of a dict for counting net spec names
Jonathan L Long [Fri, 24 Jul 2015 03:26:46 +0000 (20:26 -0700)]
[pycaffe] use a Counter instead of a dict for counting net spec names

8 years ago[pycaffe] remove dead code
Jonathan L Long [Fri, 24 Jul 2015 03:23:42 +0000 (20:23 -0700)]
[pycaffe] remove dead code

8 years agoMerge pull request #2807 from Franck-Dernoncourt/patch-1
Jon Long [Thu, 23 Jul 2015 03:37:17 +0000 (20:37 -0700)]
Merge pull request #2807 from Franck-Dernoncourt/patch-1

Fix path to mnist_autoencoder.prototxt

8 years agoFix path to mnist_autoencoder.prototxt
Franck Dernoncourt [Thu, 23 Jul 2015 00:32:12 +0000 (17:32 -0700)]
Fix path to mnist_autoencoder.prototxt