Issue · PaddlePaddle / Paddle

新建issue

conv_transpose_mkldnn has accuracy diff

#17535 · opened 5月 21, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 14
更新于5月 28, 2019
Request for trained models with program descriptors

#17365 · opened 5月 13, 2019 by saxon_zh Intel int8
- 已关闭
- 8
更新于5月 21, 2019
Create the Pass to convert QAT graph to MKL-DNN executable graph

#17364 · opened 5月 13, 2019 by saxon_zh v1.5 for Intel Intel int8
- 已关闭
- 1
更新于6月 26, 2019
INT8 mul MKLDNN kernel need be created

#17336 · opened 5月 10, 2019 by saxon_zh v1.5 for Intel Intel int8
- 已关闭
- 3
更新于7月 24, 2019
INT8 FC MKLDNN kernel need be created

#17335 · opened 5月 10, 2019 by saxon_zh Intel int8
- 已关闭
- 0
更新于11月 27, 2019
sequence_unpad OP (bwd) need be further optimized for CPU training

#17272 · opened 5月 08, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 1
更新于5月 16, 2019
sequence_pad (fwd) need be further optimized for CPU training

#17271 · opened 5月 08, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 0
更新于5月 15, 2019
sequence_reverse OP (fwd) need be optimized for CPU training

#17270 · opened 5月 08, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 0
更新于5月 13, 2019
Transpose2 (fwd+bwd) need be further optimized for CPU training

#17269 · opened 5月 08, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 1
更新于5月 15, 2019
Softmax OP (fwd+bwd) need be further optimized

#17268 · opened 5月 08, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 46
更新于5月 25, 2019
Enable INT8 Infererence with model trained/saved from QAT (Quantization-Aware Training)

#17267 · opened 5月 08, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 1
更新于7月 24, 2019
Task for v1.5: Optimize BERT L12 (dim=768)

#17196 · opened 5月 05, 2019 by saxon_zh Intel
- 已关闭
- 0
更新于5月 08, 2019
关于离线量化中calibration代码的一些疑惑

#17184 · opened 4月 30, 2019 by saxon_zh Intel int8
- 已关闭
- 4
更新于5月 08, 2019
test_calibration fails on E5-2650 V4

#17159 · opened 4月 29, 2019 by saxon_zh Intel int8
- 已关闭
- 9
更新于4月 30, 2019
test_analyzer_transformer runs into assertion failure when ngraph is used

#17157 · opened 4月 29, 2019 by saxon_zh Intel NGraph
- 已关闭
- 5
更新于12月 09, 2019
fc_op slows on multi-instance inference

#17153 · opened 4月 29, 2019 by saxon_zh v1.5 for Intel Intel
- 已关闭
- 23
更新于5月 07, 2019
PR_CI(PADDLE) is not stable.

#17150 · opened 4月 29, 2019 by saxon_zh Intel
- 已关闭
- 3
更新于7月 05, 2019
test_analyzer_bert runs into assertion failure when ngraph is used

#17147 · opened 4月 29, 2019 by saxon_zh Intel NGraph
- 已关闭
- 1
更新于6月 11, 2019
One issue found on test_conv2d_int8_mkldnn_op case

#17057 · opened 4月 23, 2019 by saxon_zh Intel
- 已关闭
- 0
更新于4月 24, 2019
all pass can only apply to OPs in block 0

#17003 · opened 4月 22, 2019 by saxon_zh Intel
- 已关闭
- 10
更新于12月 09, 2019

PaddlePaddle / Paddle 大约 1 年 前同步成功

PaddlePaddle / Paddle
大约 1 年前同步成功