Skip to content
Snippets Groups Projects

Repository graph

You can move around the graph by using the arrow keys.
Select Git revision
  • hpvm-release-epochs
  • hpvm-release-epochs0
  • main default
  • v2.0
  • v2.0rc
  • v1.0
  • v0.5
7 results
Created with Raphaël 2.2.06May543130Apr201918171615141312111098765432130Mar292827262423222119181716151312116543228Feb27262524232019181716151413121110742131Jan3029282725242322212019181716151413Adding layer files for Imagenet DNNsmergingAdding an outer loop for multiple runs - VGG16_cifar10meringAdding source for AlexNet ImagenetAdding assertion to test for fp16 fields in table -- more such assertions neededAdding best config logging after every 100 iterationschange weight loading in graph_codegenchange code gen and graph ir for tensor runtimeFixing partial fixes to install-time tunerFix Machsuite benchmark makefilesAdd new Machsuite benchmarks and fix CBE comparison predicate bugAdding try/catch for top baseline time to be in config fileFixing linker flags for AlexNet2-HPVMEnabling debug messages by defaultFixing Linker flags for VGG16-100 HPVM benchmarkFixing linker flags for ResNet-HPVM benchmarkAdding -lcufft and -lstdc++fs to HPVM bencmark linker flagsmergingAdding validation code to install-time --- Still untested due to unavailable GPUAdding VGG16 and ResNet50 Imagenet versionsonnx files uploadedtvm & keras output matching uploadedAdding rules for building install-time binaries using FP16 routineMoving FP16_tuning setting as CMake tunable macroAdding install-time tuner that uses PROMISE+fp16+SoftwareApproxsAdding routine for Stastical confidence measurement for install-tunerFixing mem handling bug in Lenet_promise sourceUpdating piped scripts to use 5K imagesAdding promise translator support to dump Quantization rangesAdding auto garbage collection in inference loopAdding Quant Range computation across batches of 100 imagesAdding batched inference to get Quant Ranges -- needs more testingMergingReducing quantization percentiles triedprofile in nvprof/nvvpMerge branch 'approx_hpvm' of gitlab.engr.illinois.edu:llvm/hpvm into approx_hpvm1. Now support previously enexpected case of search value resulting to targetAdding test case for PROMISeupdate keras/onnx/tvm models