Permalink
Commits on Dec 14, 2018
  1. Update running docs with new NCCL semantics (#705)

    alsrgv committed Dec 14, 2018
Commits on Dec 13, 2018
  1. Keras: pass optimizer parameters as dictionary instead of **kwargs (#701

    alsrgv committed Dec 13, 2018
    )
Commits on Dec 9, 2018
  1. Update to PyTorch 1.0.0 in Dockerfile (#688)

    alsrgv committed Dec 9, 2018
Commits on Dec 8, 2018
  1. Rebalance integration tests to cover TensorFlow 1.12.0 and PyTorch 1.…

    alsrgv committed Dec 8, 2018
    …0.0 (#539)
    
    * Rebalance integration tests to cover TensorFlow 1.12.0 and PyTorch 1.0
    
    * Switch from debian:sid to debian:buster
Commits on Nov 30, 2018
  1. Update Dockerfile to TensorFlow 1.12.0 (#668)

    alsrgv committed Nov 30, 2018
Commits on Nov 28, 2018
  1. Bump version to 0.15.2 (#664)

    alsrgv committed Nov 28, 2018
  2. Add CII Best Practices Badge (#662)

    alsrgv committed Nov 28, 2018
  3. Create CONTRIBUTING.md (#660)

    alsrgv committed Nov 28, 2018
Commits on Nov 26, 2018
Commits on Nov 22, 2018
  1. Move finalization logic into termination of BackgroundThread (#644)

    alsrgv committed Nov 22, 2018
  2. Fix Python3.6 integration tests (#643)

    alsrgv committed Nov 22, 2018
Commits on Nov 19, 2018
  1. Simplify PyTorch continuous-integration example to run faster (#632)

    alsrgv committed Nov 19, 2018
Commits on Nov 13, 2018
  1. Fix race on horovod_global.tensor_table (#619)

    alsrgv committed Nov 13, 2018
Commits on Oct 29, 2018
  1. PyTorch break: .dtype() -> .scalar_type() (#595)

    alsrgv committed Oct 29, 2018
    Details are in pytorch/pytorch#12766
Commits on Oct 25, 2018
  1. Bump version to 0.15.1 (#586)

    alsrgv committed Oct 25, 2018
Commits on Oct 24, 2018
  1. Added option to treat sparse gradients as dense tensors (#570)

    alsrgv committed Oct 24, 2018
Commits on Oct 23, 2018
  1. Add AVX intrinsics to vectorize & speed up FP16-CPU computations (#574)

    alsrgv committed Oct 23, 2018
    * Add AVX intrinsics to vectorize & speed up FP16-CPU computations
    
    * Add compilation test & runtime identification of AVX/F16C instinsics
Commits on Oct 17, 2018
  1. PyTorch master compatibility, improved errors in case of .so import f…

    alsrgv committed Oct 17, 2018
    …ailure (#569)
    
    * PyTorch master compatibility, improved errors in case of .so import failure
    
    * Temporarily add future dependency for PyTorch
    
    * Switch 0.4.2 (nonexistent) -> 1.0.0
  2. Add license scan badge from FOSSA (#568)

    alsrgv committed Oct 17, 2018
Commits on Oct 16, 2018
  1. FP16 allreduce on CPU (#563)

    alsrgv committed Oct 16, 2018
Commits on Oct 15, 2018
Commits on Oct 10, 2018
  1. PyTorch 1.0: Switch to new way to initialize CPU tensors (v2) (#557)

    alsrgv committed Oct 10, 2018
Commits on Oct 8, 2018
  1. PyTorch Synthetic Benchmark (#545)

    alsrgv committed Oct 8, 2018
  2. Upgrade Dockerfile to TensorFlow 0.11.0, PyTorch 0.4.1 and Open MPI 3…

    alsrgv committed Oct 8, 2018
    ….1.2 (#544)
    
    * Upgrade Dockerfile to TensorFlow 0.11.0 and PyTorch 0.4.1
    
    * Upgrade Open MPI to 3.1.2
Commits on Oct 2, 2018
  1. Bump version to 0.15.0 (#538)

    alsrgv committed Oct 2, 2018
  2. Fully de-initialize Horovod upon shutdown (#532)

    alsrgv committed Oct 2, 2018
    * Fully de-initialize Horovod upon shutdown
    
    * Move clearing initialization flag to just before starting initialization
    
    * Add missing shut_down behaviors as well
Commits on Oct 1, 2018
  1. Revert `future` dependency (#537)

    alsrgv committed Oct 1, 2018
Commits on Sep 27, 2018
  1. Don't import torch if HOROVOD_WITHOUT_PYTORCH is set (#524)

    alsrgv committed Sep 27, 2018
Commits on Sep 26, 2018
  1. PyTorch 1.0 readiness updates (#523)

    alsrgv committed Sep 26, 2018
    * PyTorch 1.0 readiness updates
    
    * Further updates, drop PyTorch 0.3.1 support
    
    * Add PyTorch version check
    
    * Switch to use LooseVersion for version comparison