summaryrefslogtreecommitdiff
path: root/caffe2/quantization
AgeCommit message (Expand)AuthorFilesLines
2019-04-15unit test with multiple op invocations (#19118)Jongsoo Park11-122/+117
2019-04-09amend D14778810 (#18902)Summer Deng5-9/+36
2019-04-08Fix a dev mode bug in activation distribution observer (#19004)Summer Deng2-3/+9
2019-04-06fix bug when falling back to acc32 when weight is prepacked (#18974)Summer Deng4-122/+213
2019-04-05Revert D14778810: [caffe2/int8] fix bug when falling back to acc32 when weigh...Junjie Bai4-210/+122
2019-04-05fix bug when falling back to acc32 when weight is prepacked (#18881)Summer Deng4-122/+210
2019-04-03fold col offset into bias; optimize A symmetric quant (#17026)Jongsoo Park4-53/+205
2019-04-03use optimization in D14020675 (#16945)Jongsoo Park1-2/+6
2019-04-01add Int8FCRelu (#18673)Jongsoo Park4-24/+59
2019-04-01Revert "Tensor construction codemod(raw_mutable_data) (#16373)" (#18680)Junjie Bai1-1/+2
2019-04-01add ConvRelu schema (#18693)Jongsoo Park2-3/+7
2019-04-01use acc16 only when n>128 and k>128 in Skylake (#18672)Jongsoo Park1-6/+18
2019-03-31handle a rare case of histogram min is inf/nan (#18239)Jongsoo Park2-6/+14
2019-03-29Tensor construction codemod(raw_mutable_data) (#16373)Jerry Zhang1-2/+1
2019-03-29Change dnnlowp log level from warning to v2 (#18576)Summer Deng1-13/+12
2019-03-25Add more options to the quantization model exporter (#18383)Summer Deng1-15/+17
2019-03-22Fix alignment issues for Fake BFP16 fp32 -> bfp16 rounding routines (#18321)Jianyu Huang1-1/+1
2019-03-22add more Python interface functions to make quantization simpler (#18246)Jongsoo Park1-0/+89
2019-03-21Replace the remaining usages of IntList in caffe2 to IntArrayRefJunjie Bai2-2/+2
2019-03-20handle dst_bin_width==0 case properly (#18240)Jongsoo Park1-2/+2
2019-03-18fix bug in pool_dnnlowp_op_avx2.cc (#18141)Jongsoo Park1-1/+1
2019-03-13fixes for AVX detection (#17915)Thomas Viehmann1-1/+1
2019-03-12fix overly restrictive assertion (#17939)Jongsoo Park2-44/+106
2019-03-08Int8 FC performance debugging (#17700)Summer Deng1-47/+102
2019-03-07Change ConvPoolOp<Context>::SetOutputSize to ConvPoolOp<Context>::GetOutputSi...Jerry Zhang5-18/+34
2019-03-04print warnings when DNNLOWP_16 or DNNLOWP_ROWWISE_16 engine is used (#17176)Jongsoo Park3-1/+24
2019-02-27optimize elementwise sum (#17456)Jongsoo Park2-61/+136
2019-02-22optimize max pool 2d (#17418)Jongsoo Park4-21/+142
2019-02-22Revert D14181620: [caffe2/int8] optimize max pool 2dLu Fang3-141/+21
2019-02-22optimize max pool 2d (#17391)Jongsoo Park3-21/+141
2019-02-15merge fully_connected_rowwise_dnnlowp_op into fully_connected_dnnlowp_op (#17...Jongsoo Park13-569/+246
2019-02-15bug fix when we prepack weight and bias together (#17145)Jongsoo Park2-11/+14
2019-02-14add pre-packing operation in README.md (#17151)Jongsoo Park1-0/+47
2019-02-14Minor fix of the histogram observer in FBL eval flows (#17118)Summer Deng1-6/+8
2019-02-12Fold col offsets into bias; optimize A symmetric quant (#16942)Jongsoo Park4-64/+170
2019-02-07Activation histogram net observer with multiple histogram files as output (#1...Summer Deng2-4/+18
2019-02-07Separate elementwise level2 math functions (#16753)Xiaomeng Yang1-1/+1
2019-02-06int8 SpatialBN (#16796)Jongsoo Park4-0/+295
2019-02-06call istringstream clear after str (#16820)Jongsoo Park2-0/+4
2019-02-04Bug fix in l2 quantization (#16749)Summer Deng2-5/+7
2019-01-30Back out "[pt1][tensor] Change ConvPoolOp<Context>::SetOutputSize to ConvPool...Jerry Zhang5-34/+18
2019-01-28Change ConvPoolOp<Context>::SetOutputSize to ConvPoolOp<Context>::GetOutputSi...Jerry Zhang5-18/+34
2019-01-22Fix formating in caffe2/quantization/server/README.mdJongsoo Park1-1/+1
2019-01-20Separate Moments from math and optimize it (#16175)Xiaomeng Yang1-8/+13
2019-01-19Allow for concurrent quantization in FullyConnectedDNNLowPOp (#16174)Kjell Schubert1-15/+15
2019-01-15use fbgemm gconv in dnnlowp (#16020)Jongsoo Park5-15/+188
2019-01-13simplify lambda function use in conv dnnlowp ops to fix #15911 (#15996)Jongsoo Park3-234/+212
2019-01-12fix compile error reported in issue #15911 (#15953)Jongsoo Park2-4/+4
2019-01-113x3x3 depthwise convolution with per channel quantization (#15775)Jongsoo Park4-53/+178
2019-01-11Make it consistent for OperatorBase usage (#15908)Jianyu Huang17-78/+71
generated by cgit v1.2.3 (git 2.25.1) at 2024-11-26 22:04:56 +0000