Commits · f9f6f60fad59d061f9075c7ae2de4c757b121522 · streichg / Finn · GitLab

Snippets Groups Projects

To receive notifications about scheduled maintenance, please subscribe to the mailing-list gitlab-operations@sympa.ethz.ch. You can subscribe to the mailing-list at https://sympa.ethz.ch

Oct 08, 2021
- Small refactoring for QONNX activation handlers. · f9f6f60f
  Hendrik Borras authored 3 years ago
  
  f9f6f60f
- Add optional support for QONNX ingestion in the dataflow builder. · e7eeed0f
  Hendrik Borras authored 3 years ago
  
  e7eeed0f
- Update qonnx commit. · 2ef7525e
  Hendrik Borras authored 3 years ago
  
  2ef7525e
Oct 07, 2021
- Added mobilenet to test_QONNX_to_FINN test. · 93461641
  Hendrik Borras authored 3 years ago
  
  93461641
- Correct for fp accuracy issues during Quant constant folding. · bf5d9124
  Hendrik Borras authored 3 years ago
  
  bf5d9124
- Added support for converting Gemm to MatMul. · acdaebe2
  Hendrik Borras authored 3 years ago
  
  acdaebe2
- Updated QONNX commit. · 1ca2c541
  Hendrik Borras authored 3 years ago
  
  1ca2c541
- Catch unsupported constant folding for SCALED datatypes. · b5bcd320
  Hendrik Borras authored 3 years ago
  
  b5bcd320
- Added support for FoldQuantWeights into Add nodes. · 0207bc3d
  Hendrik Borras authored 3 years ago
  
  0207bc3d
- Updated QONNX commit. · 9b5a7f9f
  Hendrik Borras authored 3 years ago
  
  9b5a7f9f
Oct 04, 2021
- Made node insertion optional if nothing would change. · 5fc87983
  Hendrik Borras authored 3 years ago
  
  5fc87983
- Automatically move scalar and bias node into MultiThreshold node where possible. · 658fe4ef
  Hendrik Borras authored 3 years ago
  
  658fe4ef
Oct 01, 2021
- Moved removal of FINN datatypes into main QONNX to FINN transformation. · 58d28b11
  Hendrik Borras authored 3 years ago
  
  58d28b11
- Moved QONNX transformations to a new folder and moved activation handlers to their own file. · 18547a2c
  Hendrik Borras authored 3 years ago
  
  18547a2c
- Moved QONNX transformations to a new folder and moved activation handlers to their own file. · dab28d46
  Hendrik Borras authored 3 years ago
  
  dab28d46
- Fixed a potential bug where removing the FINN datatype on a tensor would also... · bdce7541
  Hendrik Borras authored 3 years ago
  
  Fixed a potential bug where removing the FINN datatype on a tensor would also remove other quantization information.
  bdce7541
- Resolved UserWarnings after QONNX to FINN conversion. · 12ea8977
  Hendrik Borras authored 3 years ago
  
  12ea8977
- Added support for catching UserWarnings emitted by FINN during onnx execution as errors. · dbee47b9
  Hendrik Borras authored 3 years ago
  
  dbee47b9
Sep 30, 2021
- Added analysis test for QONNX to FINN conversion checking for Quant nodes. · c59cc447
  Hendrik Borras authored 3 years ago
  
  c59cc447
- Added test for QONNX to FINN conversion with FINN sample models. · ec408810
  Hendrik Borras authored 3 years ago
  
  ec408810
- Added Copyright disclaimer. · 749f6019
  Hendrik Borras authored 3 years ago
  
  749f6019
- Added rudimentary doc strings. · 1b420228
  Hendrik Borras authored 3 years ago
  
  1b420228
- Moved allowed_predecessors definition into handler classes. · a74bb6f0
  Hendrik Borras authored 3 years ago
  
  a74bb6f0
- Added overall transformation for converting QONNX to FINN-ONNX. · e0da0526
  Hendrik Borras authored 3 years ago
  
  e0da0526
- Added compatibility check for identity and relu activation conversion. · bf430b61
  Hendrik Borras authored 3 years ago
  
  bf430b61
- Added support for constant folding Quant weight nodes with per-channel scaling for convolutions. · 3e7c1de6
  Hendrik Borras authored 3 years ago
  
  3e7c1de6
Sep 29, 2021
- Added handler for QuantRelu activations. · d8495f34
  Hendrik Borras authored 3 years ago
  
  d8495f34
- Added transformation for constant folding quant nodes used for scaled weights. · c7381f09
  Hendrik Borras authored 3 years ago
  
  c7381f09
- Added support for Quant idendity operations at the beginning of the ONNX graph. · e91f5133
  Hendrik Borras authored 3 years ago
  
  e91f5133
- Modified activation handler selection to consider predecessor instead of successor. · 565b42cc
  Hendrik Borras authored 3 years ago
  
  565b42cc
- Refactored code to use handler classes to convert quant to MultiThreshold. · 86acb28f
  Hendrik Borras authored 3 years ago
  
  86acb28f
Sep 28, 2021
- Added preliminary implementation for converting activations with Quant node to... · 84ad5645
  Hendrik Borras authored 3 years ago
  
  Added preliminary implementation for converting activations with Quant node to MultiThreshold nodes.
  84ad5645
Sep 27, 2021
- Replaced finn-base with qonnx. · e7c71a97
  Hendrik Borras authored 3 years ago
  
  e7c71a97
Sep 24, 2021
- Merge pull request #374 from Xilinx/feature/objdet_fixes · df872f7c
  Yaman Umuroglu authored 3 years ago
  
  Fixes & features for multi-output/object detection networks
  Unverified
  
  df872f7c
- [Partition] fix dataflow partition location · 7b90ab65
  Yaman Umuroglu authored 3 years ago
  
  7b90ab65
Sep 23, 2021
- [Test] remove extra driver gen in test_end2end_bnn_pynq · b5fb5854
  Yaman Umuroglu authored 3 years ago
  
  b5fb5854
- [Test] disable rtlsim tracing in test_end2end_bnn_pynq · 3aae8022
  Yaman Umuroglu authored 3 years ago
  
  3aae8022
Sep 22, 2021
- [Test] fix failing end2end BNN-PYNQ tests · 18edc210
  Yaman Umuroglu authored 3 years ago
  
  18edc210
- [Transform] exclude input fifos from RemoveShallowFIFOs · fef04ada
  Yaman Umuroglu authored 3 years ago
  
  fef04ada
- [Build] better names for created partition dirs · 9f35779a
  Yaman Umuroglu authored 3 years ago
  
  9f35779a