Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

EmbedLayer #1872

Closed
wants to merge 86 commits into from
Closed

EmbedLayer #1872

wants to merge 86 commits into from

Commits on Jan 25, 2015

  1. clarify draw_net.py usage: net prototxt, not caffemodel

    @shelhamer
    shelhamer committed Jan 25, 2015
    Configuration menu
    Copy the full SHA
    2f869e7 View commit details
    Browse the repository at this point in the history

  2. [docs] ask install + hardware questions on caffe-users

    @shelhamer
    shelhamer committed Jan 25, 2015
    Configuration menu
    Copy the full SHA
    61c63f6 View commit details
    Browse the repository at this point in the history

Commits on Jan 29, 2015

  1. [docs] send API link to class list

    @shelhamer
    shelhamer committed Jan 29, 2015
    Configuration menu
    Copy the full SHA
    4cc8195 View commit details
    Browse the repository at this point in the history

  2. [docs] add check mode hint to CPU-only mode error

    @shelhamer
    shelhamer committed Jan 29, 2015
    Configuration menu
    Copy the full SHA
    1f7c3de View commit details
    Browse the repository at this point in the history

Commits on Jan 30, 2015

  1. Brief explanation of SLICE layer's attributes 

    * A sample code was added.
* `slice_dim` and `slice_point` attributes were explained.
    @boechat107
    boechat107 committed Jan 30, 2015
    Configuration menu
    Copy the full SHA
    8b96472 View commit details
    Browse the repository at this point in the history

  2. lint 1f7c3de

    @shelhamer
    shelhamer committed Jan 30, 2015
    Configuration menu
    Copy the full SHA
    75d0e16 View commit details
    Browse the repository at this point in the history

  3. Merge pull request BVLC#1817 from boechat107/patch-1 

    [docs] brief explanation of SLICE layer's attributes
    @shelhamer
    shelhamer committed Jan 30, 2015
    Configuration menu
    Copy the full SHA
    e3c895b View commit details
    Browse the repository at this point in the history

Commits on Feb 16, 2015

  1. Correct 'epochs' to 'iterations' 

    See https://github.com/BVLC/caffe/blob/master/models/bvlc_reference_caffenet/solver.prototxt
    Brandon Amos committed Feb 16, 2015
    Configuration menu
    Copy the full SHA
    1e0d49a View commit details
    Browse the repository at this point in the history

  2. Merge pull request BVLC#1879 from bamos/patch-1 

    Correct 'epochs' to 'iterations'
    @sguada
    sguada committed Feb 16, 2015
    Configuration menu
    Copy the full SHA
    3e9b050 View commit details
    Browse the repository at this point in the history

Commits on Feb 20, 2015

  1. Merge pull request BVLC#1849 from BVLC/next 

    Next: release candidater
    @shelhamer
    shelhamer committed Feb 20, 2015
    Configuration menu
    Copy the full SHA
    f998127 View commit details
    Browse the repository at this point in the history

  2. Updated the path for get_ilsvrc_aux.sh to match what is found in the … 

    …current project
    e3 committed Feb 20, 2015
    Configuration menu
    Copy the full SHA
    af01b9c View commit details
    Browse the repository at this point in the history

  3. Merge pull request BVLC#1914 from eerwitt/master 

    fix Imagenet example path
    @shelhamer
    shelhamer committed Feb 20, 2015
    Configuration menu
    Copy the full SHA
    5ee85b7 View commit details
    Browse the repository at this point in the history

  4. [build] fix dynamic linking of tools 

    set the right rpath for tools and examples respectively

thanks for the report @mees!
    @shelhamer
    shelhamer committed Feb 20, 2015
    Configuration menu
    Copy the full SHA
    eabbccd View commit details
    Browse the repository at this point in the history

  5. Merge pull request BVLC#1921 from shelhamer/fix-tool-linking 

    [build] fix dynamic linking of tools
    @shelhamer
    shelhamer committed Feb 20, 2015
    Configuration menu
    Copy the full SHA
    682d9da View commit details
    Browse the repository at this point in the history

Commits on Feb 21, 2015

  1. check caffe tool runs in runtest

    @shelhamer
    shelhamer committed Feb 21, 2015
    Configuration menu
    Copy the full SHA
    5a26333 View commit details
    Browse the repository at this point in the history

Commits on Feb 22, 2015

  1. ignore pycharm files

    Anatoly Baksheev committed Feb 22, 2015
    Configuration menu
    Copy the full SHA
    a1e951d View commit details
    Browse the repository at this point in the history

  2. set proper CMAKE_INSTALL_RPATH for _caffe.so and tools

    Anatoly Baksheev committed Feb 22, 2015
    Configuration menu
    Copy the full SHA
    fca05c3 View commit details
    Browse the repository at this point in the history

  3. fixed bug in install-tree: _caffe.so installed by install(TARGET ...)… 

    … was overwritten with symlink created at build time and installed with install(DIRECTORY ...)
    Anatoly Baksheev committed Feb 22, 2015
    Configuration menu
    Copy the full SHA
    645aa03 View commit details
    Browse the repository at this point in the history

  4. minor cmake sumamry log fix

    Anatoly Baksheev committed Feb 22, 2015
    Configuration menu
    Copy the full SHA
    5e06d16 View commit details
    Browse the repository at this point in the history

  5. cpp_lint.py fails silently with Python3 (which is the default on some… 

    … systems).

This commit specifies Python2 with which cpp_lint.py works :-)
    @jsupancic
    jsupancic committed Feb 22, 2015
    Configuration menu
    Copy the full SHA
    569ae01 View commit details
    Browse the repository at this point in the history

  6. Merge pull request BVLC#1939 from Nerei/bugfix/install_rpath_for_pycaffe 

    [cmake] fix install rpath for pycaffe
    @shelhamer
    shelhamer committed Feb 22, 2015
    Configuration menu
    Copy the full SHA
    cb1f4d6 View commit details
    Browse the repository at this point in the history

Commits on Feb 24, 2015

  1. APPLE was misspelled. in Line 27

    @spmallick
    spmallick committed Feb 24, 2015
    Configuration menu
    Copy the full SHA
    845f9ea View commit details
    Browse the repository at this point in the history

  2. Merge pull request BVLC#1948 from spmallick/patch-1 

    APPLE was misspelled in Line 27
    @longjon
    longjon committed Feb 24, 2015
    Configuration menu
    Copy the full SHA
    486360d View commit details
    Browse the repository at this point in the history

  3. Merge pull request BVLC#1941 from jsupancic/cpp_lint_python2 

    fixes: cpp_lint.py fails silently with Python3
    @longjon
    longjon committed Feb 24, 2015
    Configuration menu
    Copy the full SHA
    c091197 View commit details
    Browse the repository at this point in the history

  4. Merge pull request BVLC#1926 from shelhamer/test-caffe-tool 

    Check caffe tool runs
    @longjon
    longjon committed Feb 24, 2015
    Configuration menu
    Copy the full SHA
    360dbfd View commit details
    Browse the repository at this point in the history

  5. Making python3 work with cmake and the new python wrapper

    @philkr
    philkr committed Feb 24, 2015
    Configuration menu
    Copy the full SHA
    54037d3 View commit details
    Browse the repository at this point in the history

  6. Merge pull request BVLC#1923 from philkr/python3_master 

    Making python3 work with cmake and the new python wrapper
    @longjon
    longjon committed Feb 24, 2015
    Configuration menu
    Copy the full SHA
    b915f9d View commit details
    Browse the repository at this point in the history

Commits on Feb 25, 2015

  1. Decoding the datum before feeding it into the reshaping data layer

    @philkr
    philkr committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    2cf5089 View commit details
    Browse the repository at this point in the history

  2. fixed matcaffe printout to specify num of args (now including train/t… 

    …est phase)
    @forresti @shelhamer
    forresti authored and shelhamer committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    4a3887a View commit details
    Browse the repository at this point in the history

  3. Replaced illegal tab in Makefile with spaces. 

    Commands, such as $(error ...), are not allowed to be indented with tabs
outside of targets, throwing an error instead of outputting the actual
error. The solution is to use innocuous spaces instead. Ideally, spaces
should be used everywhere outside targets, but since make does not mind
it if variable assignments are tab-indented outside targets, a complete
overhaul is not necessary. However, if more errors are added, it might
make more sense to be consistent.

Also, make will already add a period so I removed it.
    @gustavla
    gustavla committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    d2beb8a View commit details
    Browse the repository at this point in the history

  4. Makefile fix for OS X 10.10

    @sergeyk
    sergeyk committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    1377e1b View commit details
    Browse the repository at this point in the history

  5. Merge pull request BVLC#1961 from sergeyk/master 

    fix accelerate / veclib path for OS X 10.10
    @shelhamer
    shelhamer committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    3a1195a View commit details
    Browse the repository at this point in the history

  6. Merge pull request BVLC#1960 from gustavla/makefile_fix 

    Replaced illegal tab in Makefile with spaces.
    @longjon
    longjon committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    b9aa166 View commit details
    Browse the repository at this point in the history

  7. Fixing two bugs related to python3 and PROJECT_SOURCE_DIR

    @philkr
    philkr committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    7651970 View commit details
    Browse the repository at this point in the history

  8. Small fix (visualization) on SLICE layer's documentation 

    The sample was missing some additional spaces to be correctly rendered on the HTML. The mistake was mine.
    @boechat107
    boechat107 committed Feb 25, 2015
    Configuration menu
    Copy the full SHA
    25cdd35 View commit details
    Browse the repository at this point in the history

Commits on Feb 26, 2015

  1. Merge pull request BVLC#1955 from philkr/reshaping_encoded 

    Decoding the datum before feeding it into the reshaping data layer
    @shelhamer
    shelhamer committed Feb 26, 2015
    Configuration menu
    Copy the full SHA
    a677076 View commit details
    Browse the repository at this point in the history

Commits on Feb 27, 2015

  1. Merge pull request BVLC#1999 from boechat107/patch-2 

    Small fix (visualization) on SLICE layer's documentation
    @longjon
    longjon committed Feb 27, 2015
    Configuration menu
    Copy the full SHA
    4fba3da View commit details
    Browse the repository at this point in the history

Commits on Mar 3, 2015

  1. Blobs are ND arrays (for N not necessarily equals 4). 

    vector<int> shape_ instead of (num, channels, height, width).
    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    4 Configuration menu
    Copy the full SHA
    1434e87 View commit details
    Browse the repository at this point in the history

  2. Add BlobShape message; use for Net input shapes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    5407f82 View commit details
    Browse the repository at this point in the history

  3. add offset, {data,diff}_at nd blob accessors

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    119a1c6 View commit details
    Browse the repository at this point in the history

  4. TestBlob: test that legacy BlobProtos are correctly handled by ShapeE… 

    …quals
    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    c4e9ec4 View commit details
    Browse the repository at this point in the history

  5. InnerProductLayer weights are 2D; biases are 1D

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    559ff3a View commit details
    Browse the repository at this point in the history

  6. Fix sparse GaussianFiller for new IPLayer weight axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    cf9fdda View commit details
    Browse the repository at this point in the history

  7. InnerProductLayer can multiply along any axis

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    29581e6 View commit details
    Browse the repository at this point in the history

  8. ConvLayer biases are 1D

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    94179cc View commit details
    Browse the repository at this point in the history

  9. LossLayer output is 0D (scalar)

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    a0fa2a9 View commit details
    Browse the repository at this point in the history

  10. AccuracyLayer output is 0D (scalar)

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    2 Configuration menu
    Copy the full SHA
    d8c6aeb View commit details
    Browse the repository at this point in the history

  11. AccuracyLayer generalized to N instance axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    6b8a765 View commit details
    Browse the repository at this point in the history

  12. Test{Net,Solver} fixes for AccuracyLayer generalization

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    8e96445 View commit details
    Browse the repository at this point in the history

  13. EltwiseLayer need not assume old 4D dim names

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    bf73cb1 View commit details
    Browse the repository at this point in the history

  14. FlattenLayer: generalized Blob axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    1b97c06 View commit details
    Browse the repository at this point in the history

  15. common_layers.hpp: remove unused "Blob col_bob_"

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    fb9caee View commit details
    Browse the repository at this point in the history

  16. TestConcatLayer: fix style errors

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    704e524 View commit details
    Browse the repository at this point in the history

  17. TestConcatLayer: add forward/gradient tests for concatenation along num

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    d52e9a8 View commit details
    Browse the repository at this point in the history

  18. ConcatLayer: generalized Blob axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    8afdcd0 View commit details
    Browse the repository at this point in the history

  19. SliceLayer: generalized Blob axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    b868916 View commit details
    Browse the repository at this point in the history

  20. SoftmaxLayer: generalized Blob axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    abec302 View commit details
    Browse the repository at this point in the history

  21. CuDNNSoftmaxLayer: generalized Blob axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    60c288b View commit details
    Browse the repository at this point in the history

  22. SoftmaxLossLayer generalized like SoftmaxLayer

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    94d93da View commit details
    Browse the repository at this point in the history

  23. SplitLayer: change Reshape(n,h,c,w) to ReshapeLike(...)

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    e2bc9f9 View commit details
    Browse the repository at this point in the history

  24. HDF5DataLayer shapes output according to HDF5 shape

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    e6468e9 View commit details
    Browse the repository at this point in the history

  25. DataLayer outputs 1D labels

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    e56377d View commit details
    Browse the repository at this point in the history

  26. MemoryDataLayer outputs 1D labels

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    7c8725b View commit details
    Browse the repository at this point in the history

  27. ImageDataLayer outputs 1D labels

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    c87a136 View commit details
    Browse the repository at this point in the history

  28. WindowDataLayer outputs 1D labels

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    9505001 View commit details
    Browse the repository at this point in the history

  29. EuclideanLossLayer: generalized Blob axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    fcbb933 View commit details
    Browse the repository at this point in the history

  30. DummyDataLayer outputs blobs of arbitrary shape

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    7462c84 View commit details
    Browse the repository at this point in the history

  31. Add CHECK_EQ(4, ...)s to "vision layers" to enforce that the 

    num/channnels/height/width indexing is valid.
    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    69fc1f6 View commit details
    Browse the repository at this point in the history

  32. PyBlobs support generalized axes

    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    269dafa View commit details
    Browse the repository at this point in the history

  33. Add option not to reshape to Blob::FromProto; use when loading Blobs 

    from saved NetParameter

Want to keep the param Blob shape the layer has set, and not necessarily
adopt the one from the saved net (e.g. want to keep new 1D bias shape,
rather than take the (1 x 1 x 1 x D) shape from a legacy net).
    @jeffdonahue
    jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    71df6f9 View commit details
    Browse the repository at this point in the history

  34. [pycaffe] expose Blob.reshape as *args function

    @longjon @jeffdonahue
    longjon authored and jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    aa242aa View commit details
    Browse the repository at this point in the history

  35. [pytest] use non-4d blobs in test_python_layer

    @longjon @jeffdonahue
    longjon authored and jeffdonahue committed Mar 3, 2015
    Configuration menu
    Copy the full SHA
    8c79d65 View commit details
    Browse the repository at this point in the history

Commits on Mar 4, 2015

  1. Merge pull request BVLC#1970 from jeffdonahue/tensor-blob 

    Blobs are N-D arrays (for N not necessarily equals 4)
    @shelhamer
    shelhamer committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    85bb397 View commit details
    Browse the repository at this point in the history

  2. Merge pull request BVLC#1966 from philkr/python_fixes 

    cmake and python3 bugfixes for BVLC#1939 and BVLC#1923
    @longjon
    longjon committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    136139b View commit details
    Browse the repository at this point in the history

  3. Add error checking for image mean 

    When setting the mean, assert that it is either one pixel or an array with
shape equal to the input data size.
    @lukeyeager
    lukeyeager committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    642619b View commit details
    Browse the repository at this point in the history

  4. Merge pull request BVLC#2031 from NVIDIA/image_mean 

    Check shape of input mean
    @shelhamer
    shelhamer committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    c0bc17c View commit details
    Browse the repository at this point in the history

  5. fix comment I forgot about from @shelhamer's review of BVLC#1970

    @jeffdonahue
    jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    dec148e View commit details
    Browse the repository at this point in the history

  6. zero-init param diffs and accumulate gradients 

    (With layers whose backwards accumlate gradients), this effectively
decouples the computational batch from the SGD minibatch. Each
iteration accumulates gradients over iter_size batches, then parameters
are updated.
    @longjon @jeffdonahue
    longjon authored and jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    fdf9846 View commit details
    Browse the repository at this point in the history

  7. zero-init param diffs in gradient checker

    @longjon @jeffdonahue
    longjon authored and jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    3285efc View commit details
    Browse the repository at this point in the history

  8. accumulate gradients in inner product layer

    @sguada @jeffdonahue
    sguada authored and jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    0410dab View commit details
    Browse the repository at this point in the history

  9. accumulate gradients in (de)conv layers

    @longjon @jeffdonahue
    longjon authored and jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    32a7cad View commit details
    Browse the repository at this point in the history

  10. accumulate gradients in cudnn conv layer

    @longjon @jeffdonahue
    longjon authored and jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    db16606 View commit details
    Browse the repository at this point in the history

  11. Add gpu_util.cuh, with caffe_gpu_atomic_add 

    (double impl from NVIDIA dev docs; float impl included in CUDA as
"atomicAdd")
    @jeffdonahue
    jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    894d063 View commit details
    Browse the repository at this point in the history

  12. test_gradient_check_util: check_bottom < -1 only checks params

    @jeffdonahue
    jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    ab68f93 View commit details
    Browse the repository at this point in the history

  13. Add EmbedLayer for inner products with sparse input (one-hot vectors), 

    with unit tests
    @jeffdonahue
    jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    6e38bb7 View commit details
    Browse the repository at this point in the history

  14. EmbedBackward with no loops -- use caffe_gpu_atomic_add instead

    @jeffdonahue
    jeffdonahue committed Mar 4, 2015
    Configuration menu
    Copy the full SHA
    2db4482 View commit details
    Browse the repository at this point in the history