Commits · 6f4d1039840ebf828aa377dd50d0b5b3dc0c17bf · 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

Jun 28, 2020
- [StreamingFC] bugfix in bram estimation · 6f4d1039
  Yaman Umuroglu authored 4 years ago
  
  6f4d1039
Jun 27, 2020
- [Streamline] AbsorbSignBiasIntoMultiThreshold part of Streamline · 19b3e6d2
  Yaman Umuroglu authored 4 years ago
  
  19b3e6d2
- [Transform] introduce AbsorbSignBiasIntoMultiThreshold · d7769677
  Yaman Umuroglu authored 4 years ago
  
  d7769677
Jun 26, 2020
- [Core] allow subgraph exec in execute_onnx · c5ce46c4
  Yaman Umuroglu authored 4 years ago
  
  c5ce46c4
Jun 25, 2020
- Fix comment and infer_node_datatype for cases when in data type != out data type · 5e33a8f6
  Tobi-Alonso authored 4 years ago
  
  5e33a8f6
- [Streamline] Add AbsorbScalarMulIntoTopK transformation · d54cb43c
  auphelia authored 4 years ago
  
  d54cb43c
Jun 24, 2020
- [Streamline] in MoveMulPastDWConv raise warning if mul weights are not set · ad33d8d0
  auphelia authored 4 years ago
  
  ad33d8d0
- [CustomOp] Update QuantAvgPool2d · 5905f974
  auphelia authored 4 years ago
  
  5905f974
- [Transform] Add QuantAvgPool to infer datalayout trafo · 883863e8
  auphelia authored 4 years ago
  
  883863e8
- [Streamline] Update MoveMulPastDWConv transformation · 945208d4
  auphelia authored 4 years ago
  
  945208d4
- [CustomOp] Update make_shape_compatible fct in QuantAvgPool2d to guarantee... · f525398c
  auphelia authored 4 years ago
  
  [CustomOp] Update make_shape_compatible fct in QuantAvgPool2d to guarantee that InferShapes works directly after brevitas export
  f525398c
- [Transform] Update change datalayout trafo for quantavgpool · 011d9a68
  auphelia authored 4 years ago
  
  011d9a68
- [CustomOp] Integrate NHWC functionality into QUantAvgPool2d node · b291eb06
  auphelia authored 4 years ago
  
  b291eb06
- [Registry] Remove obsolete QuantAvgPool2dNHWC node · d1c7a151
  auphelia authored 4 years ago
  
  d1c7a151
- Delete obsolete files · 2da0fe14
  auphelia authored 4 years ago
  
  2da0fe14
Jun 23, 2020
- [HLSCustomOp] remove unroll limits in hls synthesis tcl · d000c97d
  Yaman Umuroglu authored 4 years ago
  
  d000c97d
- [StreamingFC] get_weightstream_width is 0 except for decoupled · 3d505e2c
  Yaman Umuroglu authored 4 years ago
  
  3d505e2c
- [Transform] Add trafo to change QuantAvgPool2d (N,C,H,W) -> (N,H,W,C) · 6cf428f4
  auphelia authored 4 years ago
  
  6cf428f4
- [CustomOp] Add QuantAvgPool2dNHWC node · 769dbc7b
  auphelia authored 4 years ago
  
  769dbc7b
Jun 22, 2020
- [TLastMarker] fix typo · 2250a368
  Yaman Umuroglu authored 4 years ago
  
  2250a368
- [TLastMarker] kernel2kernel -> internal · 24c1dda3
  Yaman Umuroglu authored 4 years ago
  
  24c1dda3
- [InsertTLastMarker] update DynIters value types · 67a4544c
  Yaman Umuroglu authored 4 years ago
  
  67a4544c
- [TLastMarker] update comments, restrict values · c2c3712b
  Yaman Umuroglu authored 4 years ago
  
  c2c3712b
Jun 21, 2020
- [Streamline] remove identity ops as part of streamlining · cbf24078
  Yaman Umuroglu authored 4 years ago
  
  cbf24078
- [Transform] license + more flexibility in RemoveIdentityOps · 91e3a3f5
  Yaman Umuroglu authored 4 years ago
  
  91e3a3f5
Jun 18, 2020
- [Transform] Limit removing of identity ops to scalar ops · ab36ab55
  auphelia authored 4 years ago
  
  ab36ab55
- [Transform] Add trafo to remove identity ops · 62ead99d
  auphelia authored 4 years ago
  
  62ead99d
- [Core] allow disabling qnt sanitization w SANITIZE_QUANT_TENSORS=0 · 79bff3ba
  Yaman Umuroglu authored 4 years ago
  
  79bff3ba
- [Transform] Add trafo to move channelwise mul past dw conv · 52302ca6
  auphelia authored 4 years ago
  
  52302ca6
- [Transform] set wdt/odt as float32 for CollapseRepeated · e0f7e034
  Yaman Umuroglu authored 4 years ago
  
  e0f7e034
- [Utils] error thresh should be float · 7102f289
  Yaman Umuroglu authored 4 years ago
  
  7102f289
- [fpgadataflow] Add InferPool_Batch transformation to infer Pool_Batch op · 056934e8
  Tobi-Alonso authored 4 years ago
  
  056934e8
- [HLSCustomOp] Add new Pool_Batch op to support pooling layers with kernel_size != stride · 9627536a
  Tobi-Alonso authored 4 years ago
  
  9627536a
- [fpgadataflow] Remove print used for debug · c61e83d2
  Tobi-Alonso authored 4 years ago
  
  c61e83d2
- [HLSCustomOp] Update FMPadding_Batch. New attribute: SIMD · 687f3261
  Tobi-Alonso authored 4 years ago
  
  687f3261
- [Util] Error threshold for max error in rounding for finn execution as env var · 4dae8677
  auphelia authored 4 years ago
  
  4dae8677
Jun 17, 2020
- [Core] rename validate -> sanitize_quant_values, add comments · 6bb39ed7
  Yaman Umuroglu authored 4 years ago
  
  plus a small bugfix (break -> continue)
  6bb39ed7
- [Transform] add more MultiThreshold checks in MoveMaxPoolPastMultiThreshold · 6232549f
  Yaman Umuroglu authored 4 years ago
  
  6232549f
- [Transform] check MaxPool padding before skipping · 28cd2b05
  Yaman Umuroglu authored 4 years ago
  
  28cd2b05
- [Transform] use DataType.signed() in MoveMaxPoolPastMultiThreshold · 92665b53
  Yaman Umuroglu authored 4 years ago
  
  added dt inference to test case
  92665b53