Commits

Branch selector

 master

User selector

liufengdb

Datepicker

All time

Commit History

Commits on Jul 6, 2020

• Remove the constraint the QConst couldn't be shared

  Show description for 108aa90
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  108aa90

  Copy full SHA for 108aa90

Commits on Jul 3, 2020

• Expand the min/max range to prevent extremely small scales

  Show description for d31913d
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  d31913d

  Copy full SHA for d31913d

Commits on Jul 1, 2020

• Add a trait verifier to the same operands and results scales trait

  Show description for ac9969a
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  ac9969a

  Copy full SHA for ac9969a

Commits on Jun 30, 2020

• Cleanup the logic for duplicating constants

  Show description for e5d8b05
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  e5d8b05

  Copy full SHA for e5d8b05

Commits on Jun 23, 2020

• Fuse the back to back tfr.cast with unranked input to tf.EnsureShape

  Show description for 557916f
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  557916f

  Copy full SHA for 557916f

Commits on Jun 22, 2020

• Add quantization support for tfl.transpose_conv

  Show description for 31acd85
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  31acd85

  Copy full SHA for 31acd85

• Apply default quantization parameter before quantization pass

  Show description for 22f6939
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  22f6939

  Copy full SHA for 22f6939

Commits on Jun 3, 2020

• Expose the fully_quantize flag for the new mlir quantizer

  Show description for 2fc2651
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  2fc2651

  Copy full SHA for 2fc2651

• Fix the l2norm zip tests and re-enable all the fully_quantize tests through the new converter

  Show description for ab7c949
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  ab7c949

  Copy full SHA for ab7c949

Commits on Jun 1, 2020

• Define the fixed output range interface and fix the logistic quantization error

  Show description for 25af574
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  25af574

  Copy full SHA for 25af574

• Respect the quantization parameters from QAT when applying post-training quantization

  Show description for ecb473d
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  ecb473d

  Copy full SHA for ecb473d

• Consolidate all the quantization configurations and validations

  Show description for 38e7a44
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  38e7a44

  Copy full SHA for 38e7a44

Commits on May 29, 2020

• Remove the redundant tfl.quantizet/tfl.dequantize pair in the post-quantize

  Show description for 6115eda
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  6115eda

  Copy full SHA for 6115eda

Commits on May 28, 2020

• Make quantized_input_stats optional inside the converter

  Show description for 6133029
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  6133029

  Copy full SHA for 6133029

Commits on May 19, 2020

• Remove trivial quantize op

  Show description for 97aed8f
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  97aed8f

  Copy full SHA for 97aed8f

Commits on May 18, 2020

• Support int8 in tflite_convert

  Show description for 83b8556
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  83b8556

  Copy full SHA for 83b8556

Commits on May 15, 2020

• Support int16 quantization type

  Show description for da27ac6
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  da27ac6

  Copy full SHA for da27ac6

Commits on May 14, 2020

• Expose inference type in the mlir quantizer

  Show description for d33cb73
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  d33cb73

  Copy full SHA for d33cb73

Commits on May 7, 2020

• Fix the real multiplier calculation

  Show description for 59e53e9
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  59e53e9

  Copy full SHA for 59e53e9

• Run shape inference before any new non-tf ops are introduced

  Show description for fac30b7
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  fac30b7

  Copy full SHA for fac30b7

Commits on May 5, 2020

• Use the calculated activation range in depthwise_conv

  Show description for 7c39403
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  7c39403

  Copy full SHA for 7c39403

Commits on May 1, 2020

• Add the same scale decompose function

  Show description for ae76544
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  ae76544

  Copy full SHA for ae76544

Commits on Apr 30, 2020

• Add the scale propagation pass to the pipeline

  Show description for bac40c0
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  bac40c0

  Copy full SHA for bac40c0

Commits on Apr 22, 2020

• Legalize tf.Round to xla_hlo.round

  Show description for 59688cc
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  59688cc

  Copy full SHA for 59688cc

Commits on Apr 21, 2020

• Fix a typo in the device target method

  Show description for 9bcc7a8
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  9bcc7a8

  Copy full SHA for 9bcc7a8

Commits on Apr 16, 2020

• Add the scale decomposition interface to the device spec

  Show description for de9ade2
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  de9ade2

  Copy full SHA for de9ade2

Commits on Apr 15, 2020

• Add utility method to calculate the final activation range

  Show description for 5c182c2
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  5c182c2

  Copy full SHA for 5c182c2

Commits on Apr 14, 2020

• Reorder the element-wise operations and move operations

  Show description for 4e80982
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  4e80982

  Copy full SHA for 4e80982

• Add an utility method to decompose a double to two integers

  Show description for 7c115c1
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  7c115c1

  Copy full SHA for 7c115c1

Commits on Apr 11, 2020

• Add an utility method to decompose a double to two integers

  Show description for 9640308
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  9640308

  Copy full SHA for 9640308

Commits on Apr 10, 2020

• Convert Softmax to custom op

  Show description for dc9eec0
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  dc9eec0

  Copy full SHA for dc9eec0

• Remove the experimental xla quantization work from open source (II)

  Show description for f221748
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  f221748

  Copy full SHA for f221748

• Remove experimental xla hlo quantization from open source (I)

  Show description for 03c15fe
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  03c15fe

  Copy full SHA for 03c15fe

Commits on Apr 8, 2020

• Add the the kernel fusion pass to the pipeline and tests

  Show description for eea6e5e
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  eea6e5e

  Copy full SHA for eea6e5e

• Fuse Relu(X) to the affine ops

  Show description for 789d6cf
  

  liufengdb

  authored and

  tensorflower-gardener

  committed
  789d6cf

  Copy full SHA for 789d6cf

Pagination

Previous

Next