Code

Vitis Accelerator IP Flow#1134

Code

Draft

steltze wants to merge 90 commits intofastmachinelearning:mainfastmachinelearning/hls4ml:mainfrom steltze:vitis_accelerator_ip_flowsteltze/hls4ml:vitis_accelerator_ip_flowCopy head branch name to clipboard

+1,621-15

Commits

Commits on Apr 19, 2024

• Initial commit

  steltze

  committed
  312832f

• Set change the backend

  steltze

  committed
  d2b5a15

• Change the accelerator config script

  steltze

  committed
  02659dd

• Set the vitis accelerator template

  steltze

  committed
  56296b6

• Set vitis accelerator writer

  steltze

  committed
  7dd0173

• Fix writes init

  steltze

  committed
  6f181b8

Commits on May 10, 2024

• Include separable convolution resource implementation

  steltze

  committed
  bd2e52e

Commits on May 29, 2024

• Separate depthwise resource strategy to 3 cases

  steltze

  committed
  b795240

• Complete vitis accelerator wrapper for io_stream case

  steltze

  committed
  eeb04d4

Commits on May 31, 2024

• Fix call to wrong backend writer

  steltze

  committed
  7e47c85

• Fix vitis accelerator writer

  steltze

  committed
  5a2a38f

Commits on Jun 11, 2024

• Fix include in axi wrapper header file writer

  steltze

  committed
  99f9429

• Change python-cpp bridge writer

  steltze

  committed
  b9609dc

• Fix tlast handling in axis wrapper writer

  steltze

  committed
  4f69c16

• Extend convert_data to handle stream type, use that for the bridge

  steltze

  committed
  014a7b2

Commits on Jun 14, 2024

• Add zcu102 to the supported boards json

  steltze

  committed
  723073e

Commits on Jun 20, 2024

• Fix some c synthesis warnings

  steltze

  committed
  290896b

• Group more tests per YAML to reduce the number of envs created

  

  vloncar

  authored and

  steltze

  committed
  c9dfcf2

• Support negative_slope in quantized_relu

  

  vloncar

  authored and

  steltze

  committed
  d3b8e20

• [pre-commit.ci] auto fixes from pre-commit hooks

  

  pre-commit-ci[bot]

  authored and

  steltze

  committed
  b32984f

• Fix activation check in profiling

  

  vloncar

  authored and

  steltze

  committed
  98273a0

• Stage initial set of changes for the Catapult backend (fastmachinelearning#956)

  Show description for 1640c4b
  

  5 peopleauthored and

  steltze

  committed
  1640c4b

• [pre-commit.ci] pre-commit autoupdate

  Show description for 2a71a83
  

  pre-commit-ci[bot]

  authored and

  steltze

  committed
  2a71a83

• fix unwanted tested file change in fastmachinelearning#956

  

  calad0i

  authored and

  steltze

  committed
  6ac964c

• Fix SR backend synth missing variables

  

  bo3z

  authored and

  steltze

  committed
  ec95e01

• Test for SR backend config

  

  vloncar

  authored and

  steltze

  committed
  5de1bf5

• Upsampling support for PyTorch models

  

  vloncar

  authored and

  steltze

  committed
  a6fec36

• Split Catapult types into separate file

  

  vloncar

  authored and

  steltze

  committed
  1b72b19

• Split Quartus types into separate file

  

  vloncar

  authored and

  steltze

  committed
  28521d0

• Split Vivado types into separate file

  

  vloncar

  authored and

  steltze

  committed
  a44707d

• Increase precision of Softsign test

  

  vloncar

  authored and

  steltze

  committed
  cefab60

• Use quantized input in binary CNN test

  

  vloncar

  authored and

  steltze

  committed
  440901b

• Add UnspecifiedPrecisionType

  

  vloncar

  authored and

  steltze

  committed
  c351a02

• Rudimentary optimizer to infer 'auto' precision

  

  vloncar

  authored and

  steltze

  committed
  4d9d35a

• Auto precision test

  

  vloncar

  authored and

  steltze

  committed
  32ae9b6

• Sepconv fixes

  

  vloncar

  authored and

  steltze

  committed
  932b01e

• update precision propagation for signed, select im2col for quartus parallel conv

  

  jmitrevs

  authored and

  steltze

  committed
  6a65fed

• Make inferring no_bias a configurable option of the optimizer

  

  vloncar

  authored and

  steltze

  committed
  41b7e98

• updates to infering precision from qonnx branch

  

  jmitrevs

  authored and

  steltze

  committed
  24253e1

• remove count, become more selective on when True is returned

  

  jmitrevs

  authored and

  steltze

  committed
  6ee8189

• fix pooling precision

  

  calad0i

  authored and

  steltze

  committed
  b5add0c

• remove typing

  

  calad0i

  authored and

  steltze

  committed
  665c904

• Fix avg pooling op check

  

  vloncar

  authored and

  steltze

  committed
  b366d24

• Optimizer to remove expensive Transpose that serves as Flatten

  

  vloncar

  authored and

  steltze

  committed
  f0ca865

• Generalize removal of Transpose after flatten so it works on 1D as well

  

  vloncar

  authored and

  steltze

  committed
  1e416b5

• Remove transpose of input if n_chan=1

  

  vloncar

  authored and

  steltze

  committed
  2a5d8de

• SepConv1d/2d for io_parallel w/ Latency strategy

  

  vloncar

  authored and

  steltze

  committed
  3969523

• Cosmetic parameter config fixes

  

  vloncar

  authored and

  steltze

  committed
  52252ca

• Tests for SepConv io_parallel

  

  vloncar

  authored and

  steltze

  committed
  be56b93

• [pre-commit.ci] pre-commit autoupdate

  Show description for b0085a1
  

  pre-commit-ci[bot]

  authored and

  steltze

  committed
  b0085a1

• Update pytest docker image to 0.5.4

  

  jmitrevs

  authored and

  steltze

  committed
  44bc8f3

• bump to 0.5.5

  

  jmitrevs

  authored and

  steltze

  committed
  a7826e0

• fix pre-commit warning

  

  jmitrevs

  authored and

  steltze

  committed
  41ab6af

• change writing of obsolete ".h5" to ".keras" files

  

  jmitrevs

  authored and

  steltze

  committed
  c0f8d9f

• Fix extension test for Keras v3

  

  vloncar

  authored and

  steltze

  committed
  bcfd685

• Support ParallelizationFactor in SepConv1D/2D

  

  vloncar

  authored and

  steltze

  committed
  8c09595

• updated pytest docker image

  

  jmitrevs

  authored and

  steltze

  committed
  11819ac

• Don't test io_parallel for Catapult test and reduce the size of test to speed it up

  

  vloncar

  authored and

  steltze

  committed
  39d9232

• Add explicit DepthwiseConv tests and simpligy SepConv tests

  

  vloncar

  authored and

  steltze

  committed
  68a83d6

• [pre-commit.ci] pre-commit autoupdate

  Show description for 8a9d556
  

  pre-commit-ci[bot]

  authored and

  steltze

  committed
  8a9d556

• Initial commit

  steltze

  committed
  ad86387

• Stage initial set of changes for the Catapult backend (fastmachinelearning#956)

  Show description for 4ea329b
  

  5 peopleauthored and

  steltze

  committed
  4ea329b

• Rudimentary optimizer to infer 'auto' precision

  

  vloncar

  authored and

  steltze

  committed
  992b9b7

• Sepconv fixes

  

  vloncar

  authored and

  steltze

  committed
  8174465

• Optimizer to remove expensive Transpose that serves as Flatten

  

  vloncar

  authored and

  steltze

  committed
  84ff2c6

• Remove transpose of input if n_chan=1

  

  vloncar

  authored and

  steltze

  committed
  518796d

• Optimizer to remove expensive Transpose that serves as Flatten

  

  vloncar

  authored and

  steltze

  committed
  238e35c

• Remove transpose of input if n_chan=1

  

  vloncar

  authored and

  steltze

  committed
  c10dd82

• fix up automatic precision inferrence

  

  jmitrevs

  authored and

  steltze

  committed
  d6fe369

• starting towards being able to split seperable

  

  jmitrevs

  authored and

  steltze

  committed
  7290a29

• complete implementation of seperable -> dw + pw, untested

  

  jmitrevs

  authored and

  steltze

  committed
  13fcf0a

• make conv_same_pad also trigger on depthwise, varius bug fixes

  

  jmitrevs

  authored and

  steltze

  committed
  92e7222

• add parsing of depth multiplier for 1D depthwise conv

  

  jmitrevs

  authored and

  steltze

  committed
  f12a7ea

Commits on Aug 26, 2024

• Merge remote-tracking branch 'upstream/main' into vitis_accelerator_ip_flow

  

  stzelepi

  committed
  4d24e4e

• Finish resolving conficts with main

  

  stzelepi

  committed
  e2d270e

Commits on Nov 18, 2024

• Supress removing tar for now

  steltze

  committed
  fa6bd66

• Fix csynth and cosim

  steltze

  committed
  b42210d

• Fix tcl script to find cosim report

  steltze

  committed
  1303bba

• Correct PYNQ Z2 vivado tcl script, bitstream generated

  steltze

  committed
  8d3a1f2

Commits on Nov 19, 2024

• Clean pynq tcl script

  steltze

  committed
  a8e0497

• Fix compatibility of nnet helper functions with vitis axis

  steltze

  committed
  48686d3

• Setup vivado tcl script for zcu102

  steltze

  committed
  bae450b

• Rename backend to VitisAcceleratorIPFLow to prevent conflicts with kernel flow-versal

  steltze

  committed
  dde9124

Commits on Nov 20, 2024

• Fix compatiblity between axi stream and io parallel

  steltze

  committed
  663181f

• Update pynq driver for zcu102

  steltze

  committed
  e32f4d0

• Run pre-commit

  steltze

  committed
  c52ec75

• Remove unused file

  steltze

  committed
  9d9e645

• Remove unused xclbin generator

  steltze

  committed
  80697c0

Commits on Nov 27, 2024

• Clean backends init

  steltze

  committed
  f467829

• Fix backend import sequence

  steltze

  committed
  4c74550