Skip to content
Snippets Groups Projects
Select Git revision
  • Athos-Float
  • EzPC-Secfloat
  • covid
  • cryptflow2
  • docker
  • grouped
  • master default protected
  • onnx-demo
  • onnx-demo-float
  • random_forest
  • updatedCompilerChanges
11 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.021Jan2015121123Dec222126Nov2531Aug242321201819171816151426Jul14118229Jun242220191817161512104228May24222017161312111021Apr191713108765431Mar3028262524232018171613111098543227Feb262423222120191817121017Dec5Nov431Oct30292522218427Sep25242019181614431Aug302315Jul1211927Jun2522Mar16FebImplement Depthwise Separable ConvolutionAdd instructions to compile any tf modelChange default save_weights to true[README] Instructions to run the networks using scripts.add setup section in readmeUpdate README.mdAdd script to install dependencies and compile everythingAdd script to compile and run models in Networks directory.testFixes output mismatch in reduce_mean and reduce_sum.Fix reduce_mean/sum of tensors to 1 dimension.Add debug flag in compile scriptUpdate test cases and add timeoutUpdate disclaimerAdd unittests and related scriptsFix CreateTensor signature for 64 bit caseTypo fix in seedotHandle TransformGraph for cases where constant outputs where converted to vars.Minor improvements to compiler scripts, handle const to var better for some opsAdd support to directly link code with SCI, PorthosAlways checkout Eigen3 and SEAL while building SCIPass key directory as argument for Porthos.Add grappler opts and new script to compile graphs.Improve assert message for broadcasts in typeinferAdd support for ":0" indexing of nodes in graphdefFix typo for Conv2D truncationConvert Slice operator to compile-time codgen.Add FusedBatchNorm in GarbageCollector and bugfix.Improvements to compiler scripts.Fix tf size inference for multiple output tensors.Add a new garbage collector pass.Add support for broadcasting semantics for binops.Support for reduced meanSupport for Split operationTaint Analysis added to type inferenceFix codegen for unary negation of tensorsDon't add clearmem calls for freeing scalarsAdd frontend support for Tanh, Sigmoid and [R]SqrtImplement StopGradient as no-opAutomatically scale down final ouptut.