Commits

Branch selector

 master

User selector

mgouicem

Datepicker

All time

Commit History

Commits on Apr 17, 2020

• benchdnn: rnn: skip unsupported int8 cases properly

  mgouicem

  committed
  74baa19

  Copy full SHA for 74baa19

• src: cpu: rnn: fix weights layout check in leading dimension computation

  mgouicem

  committed
  fd8e39a

  Copy full SHA for fd8e39a

• src: cpu: rnn: default initialize parameters to balance211

  mgouicem

  committed
  14fb7ec

  Copy full SHA for 14fb7ec

• src: cpu: rnn: distinguish weights oc and data LD in weights packing

  mgouicem

  committed
  62a87a3

  Copy full SHA for 62a87a3

• api: rnn: use the proper error message when setting qparams

  mgouicem

  committed
  a5017e9

  Copy full SHA for a5017e9

• src: cpu: rnn: remove spurious type conversion

  mgouicem

  committed
  7457706

  Copy full SHA for 7457706

• src: cpu: reorder: rnn: fix f32 -> s8 reorder for I=1

  mgouicem

  committed
  60f7ea2

  Copy full SHA for 60f7ea2

• benchdnn: rnn: simplify handling of quantization parameters

  mgouicem

  committed
  ebb21b9

  Copy full SHA for ebb21b9

• src: cpu: rnn: saturate properly when accumulator is larger then 2^31

  mgouicem

  committed
  841dada

  Copy full SHA for 841dada

• benchdnn: rnn: print trivial-strides in reproducer line

  mgouicem

  committed
  202c590

  Copy full SHA for 202c590

• src: cpu: rnn: check for gemm with zero dimension

  mgouicem

  committed
  06306cf

  Copy full SHA for 06306cf

• src: cpu: rnn: minor cleanup

  mgouicem

  committed
  eb01322

  Copy full SHA for eb01322

• fixup! benchdnn: rnn: fix norm checks in case of NaNs

  mgouicem

  committed
  97a19da

  Copy full SHA for 97a19da

Commits on Apr 14, 2020

• src: cpu: reorder: rnn: improv the performance of f32-u8 data reorder

  mgouicem

  committed
  d5ffdcd

  Copy full SHA for d5ffdcd

Commits on Apr 2, 2020

• benchdnn: rnn: check for out of memory

  mgouicem

  committed
  db4d52c

  Copy full SHA for db4d52c

Commits on Mar 31, 2020

• benchdnn: rnn: add nullptr check after malloc

  mgouicem

  committed
  b1aaf99

  Copy full SHA for b1aaf99

• src: cpu: rnn: use std::memcpy instead of memcpy

  mgouicem

  committed
  93812d9

  Copy full SHA for 93812d9

Commits on Mar 30, 2020

• benchdnn: rnn: add option to choose trivial strides or not

  mgouicem

  committed
  1b6b9c6

  Copy full SHA for 1b6b9c6

• benchdnn: rnn: fix norm checks in case of NaNs

  mgouicem

  committed
  7b24fc6

  Copy full SHA for 7b24fc6

Commits on Mar 28, 2020

• Apply suggestion to src/cpu/rnn/ref_rnn.cpp

  mgouicem

  committed
  d1681fa

  Copy full SHA for d1681fa

Commits on Mar 27, 2020

• src: cpu: rnn: fix issue in bwd when merge_gemm_layer, skip_dst_iter and L > 1

  mgouicem

  committed
  fdd2cf9

  Copy full SHA for fdd2cf9

• src,tests,doc: rnn: change definition of lstmp

  Show description for eb1d184
  mgouicem

  committed
  eb1d184

  Copy full SHA for eb1d184

Commits on Mar 20, 2020

• api: rnn: introduce projection lstm

  Show description for 9d4e1f9
  

  mgouicem

  authored and

  Fomenko, Evarist M

  committed
  9d4e1f9

  Copy full SHA for 9d4e1f9

Commits on Feb 18, 2020

• src: cpu: rnn: remove lambda call in omp simd loop

  Show description for a337d4a
  mgouicem

  committed
  a337d4a

  Copy full SHA for a337d4a

Commits on Feb 15, 2020

• doc: reformat tensor names in math mode

  mgouicem

  committed
  6934076

  Copy full SHA for 6934076

• doc: add execution argument index table for all primitives

  mgouicem

  committed
  660911c

  Copy full SHA for 660911c

• doc: add macros to enable uniform rendering of tensor names

  mgouicem

  committed
  1b48644

  Copy full SHA for 1b48644

• doc: rnn: restructure to match other primitives

  mgouicem

  committed
  9a5e882

  Copy full SHA for 9a5e882

• doc: typeset diff_* as operatorname

  Show description for 6b0d0fb
  mgouicem

  committed
  6b0d0fb

  Copy full SHA for 6b0d0fb

Commits on Feb 6, 2020

• src: cpu: rnn: jit: pass alpha and beta appropriately for vanilla_rnn

  mgouicem

  committed
  cc5b89b

  Copy full SHA for cc5b89b

• src: cpu: rnn: jit: fix rnn+relu bwd computation

  mgouicem

  committed
  7a58860

  Copy full SHA for 7a58860

Commits on Jan 27, 2020

• gtests: rnn: cleanup

  mgouicem

  committed
  004381e

  Copy full SHA for 004381e

• gtests: rnn: check if skipping optional parameters behaves properly

  mgouicem

  committed
  d136d7b

  Copy full SHA for d136d7b

• src: cpu: rnn: fix a missed gemm computation when we skip dst_iter copy

  mgouicem

  committed
  42fc68d

  Copy full SHA for 42fc68d

Commits on Jan 13, 2020

• src,tests: rnn: add s8->s8 weights reorder implementation and testing

  mgouicem

  committed
  dae3142

  Copy full SHA for dae3142

Pagination

Previous

Next