Skip to content
Snippets Groups Projects
Select Git revision
  • DevPTQ
  • DevQAT
  • Fix_ptq
  • MulCompensation_Nodes
  • dev
  • leaky_relu_support
  • main default protected
  • mobilenetv1_support
  • qat
  • quantization_cast
  • real_quantization_cast
  • refact_removing_scaling_nodes
  • scaling_node_for_rescale_tensor
  • storing_sf_as_producer_attributes
  • tag_instead_of_metaop
  • test_LeNet
  • true_quantization
  • v0.1.2
  • v0.1.1
  • v0.1.0
20 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.015Apr10131Mar28272519181716141211654328Feb2724212019171411531Jan3029282723222119171513863220Dec1817161310953229Nov28272625181331Oct30292823211698743227Sep26231896543230Aug23229865231Jul302926252423221918161283128Jun2019121110727May23222114136229Apr24178528Mar824Jan191822Nov151424Oct236425Sep2221145429Aug242322211185432131Jul28272625CHanging assert messageFix_ptqFix_ptqSmall fixes in the main PTQ function: Assert to block user from using castQuantizeGraph with optimizeSign/removing pybind default value for fold_graphFixed attributes cloningomobilenetv1_sup…mobilenetv1_supportMerge branch 'version_update' into 'dev'Update Pybind 2.10.4 -> 2.13.6AddPrecisionTag function that tag each node of the graph with the correct DTtrue_quantizationtrue_quantizationMerge branch 'quantization_cast' into 'dev'Real quantization cast for PTQLast Fixquantization_castquantization_castFixing resetInput() with MetaOpsAdding ONNX assetsnewINtegrating the new ptq testsuite into the CIFixing clearGraphViewInputNode chich automatically cast all input of the GV into targetType instead of resetting every inputCorrecting issue with targettypeAdding clearInput method to supress Int32 inputs after PTQAdding bitshift rounding optionRebasing real quantization cast onto devAdding changesfirst changeMerge branch 'add_matmul' into 'dev'avoid getTensorAbsoluteMax() redefinition (+ minor changes)minor changesMerge branch 'dev' into 'add_matmul'modify prepareNetwork()handle the MatMuls in prepareNetwork()minor changefix the LSQ workspace cuda allocationMerge branch 'fix_backend' into 'dev'remove commented codeedit quantizeNormalizedNetwork() (MatMul support)modify normalizeActivations() for MatMul supportmake use of determineBackend()integration of matmul (insert scalings + norm parameters)enhance reorderMatMulInputs() and remove replaceMatMulWithFC()draft the determineBackend() routinehotfix the backend issuewipminor changes

Copyright © Eclipse Foundation, Inc. All Rights Reserved.     Privacy Policy | Terms of Use | Copyright Agent