Commits

Branch selector

 master

User selector

jianlijianli

Datepicker

All time

Commit History

Commits on Apr 25, 2019

• Add test for bidirectional LSTM

  Show description for aeb3616
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  aeb3616

  Copy full SHA for aeb3616

• Update version for MUL.

  Show description for d97e898
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  d97e898

  Copy full SHA for d97e898

• Refactor bidi lstm test.

  Show description for cfe74fb
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  cfe74fb

  Copy full SHA for cfe74fb

• Remove comments that are out of date.

  Show description for db4cc16
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  db4cc16

  Copy full SHA for db4cc16

• Update aux input weights dimension. It should be cell * input.

  Show description for 5dcd9cb
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  5dcd9cb

  Copy full SHA for 5dcd9cb

Commits on Apr 24, 2019

• Add test case with project bias.

  Show description for 958b1ae
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  958b1ae

  Copy full SHA for 958b1ae

Commits on Apr 22, 2019

• Test neon code path for uint8 logistic by increasing depth from 8 to 24. At least 16 is needed.

  Show description for 3db32c9
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  3db32c9

  Copy full SHA for 3db32c9

Commits on Apr 19, 2019

• Add version to op property and set version during quantization.

  Show description for da56f87
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  da56f87

  Copy full SHA for da56f87

Commits on Apr 18, 2019

• Create a safer AveragePool with 32 bit accumulator and add the branching logic that [0, 16*16] filter uses 16 bit kernel and (16*16, inf] uses 32 bit kernel.

  Show description for 129ff2e
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  129ff2e

  Copy full SHA for 129ff2e

• Add versioning to Mean.

  Show description for d5e562e
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  d5e562e

  Copy full SHA for d5e562e

Commits on Apr 16, 2019

• Update comment in the quantizer. The tensors are in random order, it is the ops that are sorted.

  Show description for b8206b4
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  b8206b4

  Copy full SHA for b8206b4

Commits on Apr 15, 2019

• Remove subgraph quantizer.

  Show description for 1f8cabf
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  1f8cabf

  Copy full SHA for 1f8cabf

Commits on Apr 13, 2019

• Fix asan, tsan.

  Show description for c7ce6f4
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  c7ce6f4

  Copy full SHA for c7ce6f4

Commits on Apr 12, 2019

• Improve post-training quantizer

  Show description for d134aaf
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  d134aaf

  Copy full SHA for d134aaf

• Add utility methods to quantization utils for model quantization.

  Show description for 7e4c43b
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  7e4c43b

  Copy full SHA for 7e4c43b

• Add utility methods for model quantization.

  Show description for a83376e
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  a83376e

  Copy full SHA for a83376e

Commits on Apr 9, 2019

• Move subgraph level tests to model level.

  Show description for 331de44
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  331de44

  Copy full SHA for 331de44

Commits on Apr 4, 2019

• Support Concat in quantizer.

  Show description for 123e0d5
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  123e0d5

  Copy full SHA for 123e0d5

Commits on Apr 3, 2019

• Fix raspberry Pi build.

  Show description for 6f8ffb9
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  6f8ffb9

  Copy full SHA for 6f8ffb9

Commits on Mar 29, 2019

• Update comment on quantize operator to highlight the usage that the op can

  Show description for b2e4602
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  b2e4602

  Copy full SHA for b2e4602

Commits on Mar 28, 2019

• Modify the names of the original input/output tensor from "input"/"output" to "input_int8"/"output_int8" and leave the name "input"/"output" to the newly added quant/dequant/requant node.

  Show description for f5e3223
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  f5e3223

  Copy full SHA for f5e3223

Commits on Mar 27, 2019

• Add the capability to handle uint8 in quantize_model.

  Show description for 70aa89e
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  70aa89e

  Copy full SHA for 70aa89e

• Add test on tensor name.

  Show description for b9cebd3
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  b9cebd3

  Copy full SHA for b9cebd3

• Increase maxpool test depth. Depth larger than kPoolingAccTrancheSize is needed to properly test the neon code.

  Show description for ac88b10
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  ac88b10

  Copy full SHA for ac88b10

• Add fast path for requantization between int8 and uint8 with bitwise manipulation.

  Show description for 7cb3ab3
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  7cb3ab3

  Copy full SHA for 7cb3ab3

• Create optimized int8 AveragePool.

  Show description for 65d68ab
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  65d68ab

  Copy full SHA for 65d68ab

Commits on Mar 26, 2019

• Add requantization support to Quantize operation.

  Show description for a1d8d4a
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  a1d8d4a

  Copy full SHA for a1d8d4a

Commits on Mar 25, 2019

• Enable optimized max pooling.

  Show description for e56e4ab
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  e56e4ab

  Copy full SHA for e56e4ab

Commits on Mar 21, 2019

• Create optimized softmax.

  Show description for 21ca5b2
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  21ca5b2

  Copy full SHA for 21ca5b2

Commits on Mar 20, 2019

• Update DepthwiseConv test.

  Show description for f67c35d
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  f67c35d

  Copy full SHA for f67c35d

Commits on Mar 14, 2019

• Update quantized mean.

  Show description for 333b79b
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  333b79b

  Copy full SHA for 333b79b

• Create quantized mean.

  Show description for 872b9bd
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  872b9bd

  Copy full SHA for 872b9bd

Commits on Mar 12, 2019

• Add a guard when decomposing multiplier. When the scale is small, the shift can be less than -31 and that is failing RoundingDivideByPOT. This issue is only observed in per-channel quantization and…

  Show description for dc7d314
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  dc7d314

  Copy full SHA for dc7d314

Commits on Mar 5, 2019

• Update operator_test for Max

  Show description for 118c32f
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  118c32f

  Copy full SHA for 118c32f

• Update performance bast practice page

  Show description for f4bb6ba
  

  jianlijianli

  authored and

  tensorflower-gardener

  committed
  f4bb6ba

  Copy full SHA for f4bb6ba

Pagination

Previous

Next