Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • ack
  • cleanup-ne
  • cuda-12-1-ci
  • feature/chimp
  • feature/cricketd
  • fix-cudnn-backend protected
  • free-host-mem
  • github/fork/nravic/master
  • gpu-cr
  • gpu-development
  • gpu-improvements
  • master default protected
  • refactor_main
  • server-bin-no-hash
  • share-object-support
  • socket-memcpy-multithreaded
  • 3.0
  • 2.0
  • 1.0
19 results
Created with Raphaël 2.2.017Jul1329Jun2622212016141312762125May24181615121110411Apr30Mar272410921Feb181716157Sep18Jul171620Jun1413924May29Apr2616Feb17Nov30Oct2116716Aug13921Jul13876524Jun17162131May252126Mar23222120199Feb313Jan1216Dec27Nov2523201918171613121110923Oct2221201629Sep31Aug181117Jun1014May29Apr27242398727Mar2625231918171610628Feb2213549Jan20Dec1618Nov28Oct2223Sep1727Aug819Jun5428May21181715131087330Apr29261716129Nov87fix no output on weird shells, e.g. sshfix using logger function before initializationcublas: remove usage of new APIs if we compile for CUDA 10add support for cuModuleLoadDataimprove debug output for cuModuleLoadimprove cublas implementation, add cudnnBackend implementationimprove docs/pytorch.mdimprove logging for unloading of modulesfix faulty if statement when intercepting dlopen callsadd cublas and cudnn functions to support mnistCUDNN sampleupdate readme with new server instructions and update pipeline status imageimplement three more cudnn tensor APIsimplement cudnn tensor functionsadd basic cuBLAS supportadd server side cudnn lrn implementations, fix some function namesadd cudnn LRN apiadd cudnn dependency to Dockerfilesimplement cudaMemset Async APIsadd cudnn activation and pooling apisadd more cuDNN APIsuse resource managers for cudnn apiadd cuDNN implementationadd cuDNN API stubsuse fixed size rpc array instead of opaque variable length array for cudaDevicePropadd cuDNN tests to tests/samplesStart moving gpu-checkpointing to cpu folder.gpu-developmentgpu-developmentimprove support for cuGetProcAddressadd libgl dependency to pytorch documentationadd v2 implementation of cudaGetDevicePropertiesupdate docs to not deactivate compression as we now support compressed pytorch kernelsuse uint64_t for decompressions to fix overflowing of range and length specifiers for very long compressed segmentsfix potential segfault because of missing variadic parameter in loggingfix decompression not working for long uncompressed lz4 segmentsfix elf decompression handling padding wrong in some circumstancesadd documentation on how to use pytorch to docs/pytorch.mdchange c standard to gnu11, improve loggingadd cpu-server-nvml headexclude some nvml definitions when compiling with an old CUDA version to make the CI happyadd nvml library to dockerfilesadd license to pytorch_minimal.py