Skip to content
Snippets Groups Projects
Select Git revision
  • accl default
  • adegendt/accl_clone
  • dev
  • full_accl_support
  • georg/accl-build
  • georg/accl-complete
  • georg/accl-finn
  • georg/end2end
  • main protected
9 results
You can move around the graph by using the arrow keys.
Created with Raphaël 2.2.014Oct131211108764130Sep292827242322212019171613121098731Aug30272320191716131210965426Jul23151312653230Jun282524232117151410987543130May28262520181716141110987653130Apr2927232221201918151412109826Mar25242317543226Feb242322181211108764131Jan29262120171517Dec161413983124Nov231996530Oct282726222120191514131211109876529Sep272625242321201817161514[Test] start working on lookup testAdded support for converting AvgPool + Trunc to QuantAvgPool node.Removed warnings check in test, since these warnings are unavoidable.Temporarily disabled setting of scale factors internal to the MultiThreshold node.BugfixInitial support for new 1D SWG in case of full SIMD unfolding[CustomOp] sketch CustomOp for lookup[HLS] packed i/o types for StreamingLookup[Refactor] add missing license text to some HLSCustomOpsMerge branch 'dev' into feature/lookup[Driver] fix driver DataType accessMoved GemmToMatMul from FINN to finn-base.[CustomOp] sketch outline for lookup opMerge pull request #393 from Xilinx/feature/faster_shape_infMoved RemoveIdentityOps transformation from FINN to finn-base.Preserve shapes and datatypes during gemm to matmul conversion.[Refactor] switch to CustomOp base-class impl for shape inf[Deps] update finn-baseMerge branch 'dev' into feature/faster_shape_inf[Deps] update finn-baseMerge pull request #390 from Xilinx/feature/fixed_point_datatype[Notebook] clear outputs for 1_brevitas_network_import.ipynbExtended analysis pass for test_QONNX_to_FINN to look for left over BinaryQuant nodes.Raised default for max_multithreshold_bit_width from 4 to 8, due to missing quantization for 8-bit input test networks (CNV and mobilenet).Added support for filtering which Quant nodes to convert to MultiThreshold nodes.[Notebook] update notebooks with DataType.X -> DataType["X"][Deps] update finn-base[Deps] update to latest finn-baseHandle the out_dtype attribute of MultiThreshold nodes more appropriately.[Test] add float32 case to test_npy2apintstream[Deps] update finn-baseSplit convert_qonnx_to_finn.py file into multiple.Renamed QONNX to FINN verification step to QONNX_TO_FINN_PYTHON.Made QONNX to FINN a default build step, changed verification step name and made running of conversion conditional on the existence of quant nodes.Install finn-base with commit from qonnx-branch + install qonnx w/o deps.[CustomOp] fix dtype name in Thresholding op msgSmall updates to QuantActBaseHandler documentation.[Refactor] use RandomNormal for faster/more compact shape inference[Deps] update finn-base to get faster shape inference base ops[Refactor] use get_accumulator_dt_cands as part of DataType rf.