77 Commits (a6beb96dd0235c236336f2db31df875b33db6635)

Author	SHA1	Message	Date
cc	75eec3d1f6	Post training quantization supports optimize model by fusing (#24822) ... * Post_training_quantization supports optimize model by fusing, test=develop	6 years ago
Wojciech Uss	78d4f0cc91	add option to exclude ops by id from quantization (#24689)	6 years ago
cc	dbcd7c69e9	Update sigmoid output from Y to out, test=develop (#24765)	6 years ago
cc	88e9d74a75	Collecting concat output threshold, test=develop (#24742)	6 years ago
cc	6c89ca2157	Add output threshold for ops that have several output activations, test=develop (#24726)	6 years ago
cc	4d35112255	[Fix bug] Init scale node in OutScaleForTrainingPass and enable test_quantization_scale_pass UT (#24393) ... * Init scale node in OutScaleForTrainingPass, test=develop * Enable test_quantization_scale, test=develop	6 years ago
Wojciech Uss	db052009c7	Enabled quantize all and skip missing in QAT (#24281) ... * Enabled quantize all and skip missing in QAT	6 years ago
Sylwester Fraczek	e1a7a88057	added reshape transpose matmul fuse pass (#23754)	6 years ago
arlesniak	d31a174f51	added fusing matmul-transpose-reshape pass (#23866)	6 years ago
Wojciech Uss	3d744162dd	QAT: support for new models (#23928) ... * QAT: support range-based quantization and scales from attribute * added support for channelwise	6 years ago
cc	40aa14ec77	Weight quantization support channel_wise_abs_max method to achieve higher accuracy (#23629) ... * Weight quantization support channel_wise_abs_max method to achieve higher accuracy	6 years ago
joanna.wozna.intel	12ba05ce0c	Add scale-matmul fuse pass (#23734)	6 years ago
Wojciech Uss	1753860dd0	Enable matmul and cleanup in QAT2 (#23657)	6 years ago
cc	25628587f1	Collect output scale for quantized op and fused op (#23369) ... * Collect output scale for quantized op and fused op * Post_training_quantizaion sets batch_generator to support lod tensor	6 years ago
Wojciech Uss	9fd9067455	handle conv2d activations in older QAT models (#23202)	6 years ago
Wojciech Uss	be2ac9cc3a	separated QAT1 and QAT2 (#23284)	6 years ago
Wojciech Uss	f836c8aa8f	add check for scales and a message (#23119)	6 years ago
cc	bd80903333	Add activation_type in AddQuantDequantPass to be compatible with paddleslim, test=develop (#23221)	6 years ago
cc	589cd8782f	Post_training_quantizaion supports min_max methon (#23078) ... * Post_training_quantizaion supports min_max methon	6 years ago
tianshuo78520a	d2ba91aad1	fix typo words (#22653)	6 years ago
Wojciech Uss	4cddb43c5c	Add support for Ernie NLP model to the Slim QAT (#22506) ... * a test for Ernie QAT INT8 accuracy check test=develop * Remove NLP comparison test to split PRs test=develop * Fix typo and tabs, delete commented lines test=develop * re-combine the 2 PRs, test=develop Co-authored-by: Michał Gallus <sand3r@interia.eu> Co-authored-by: bingyanghuang <33643817+bingyanghuang@users.noreply.github.com>	6 years ago
cc	d143f70a09	Post_training_quantization support set quant 8/16 bits (#22492) ... * post_training_quantization support set bits, test=develop * up, test=develop	6 years ago
cc	197913ebe1	Add weight quantization in post_training_quanzitaion (#22445) ... * support weight quantization in post_training_quanzitaion, test=develop * add test for weight quantization, test=develop	6 years ago
juncaipeng	8f7372ca81	add mul and matmul quantization, test=develop (#22054) ... * add mul and matmul quantization, test=develop * add test for matmul, test=develop	6 years ago
juncaipeng	8b74fc4fa7	Fix post training quantization (#21745) ... * fix post training quantization bug of memory constrained, support the input be different, test=develop	6 years ago
juncaipeng	1f57ac1241	delete concat in AddQuantDequantPass, test=develop (#21454)	6 years ago
lidanqing	c0aa13672e	Fp32 vs int8 qat C++ performance (#21244) ... * add ut for comparing FP32 and QAT INT8 * add save qat transformed model python script test=develop * updated * added missing file * add "with_label" test=develop * performance benchmark as unit test test=develop * change names of unnecessary thing * Change CMakeList.txt for model downloading and UT test=develop * change names of functions and params for more readable code test=develop * Change PADDLE_ENFORCE messages test=develop * fix indent problems test=develop * indent problems test=develop	6 years ago
itminner	07e6a94268	paddleslim quantization skip pattern support list of string (#21141)	6 years ago
juncaipeng	84865b806b	add resnet50 test for post trainint quantization, test=develop (#21272)	6 years ago
juncaipeng	29b63f0aa1	support set model_filename and params_filename in post_training_quantization, test=develop (#21213) ... * support set model_filename and params_filename in post_training_quantization, test=develop	6 years ago
juncaipeng	00b11a4a1e	Support more ops in post training quantization, test=develop (#21073) ... * Support more ops in post training quantization, and save the output scale in quantized op. * Update docs in post training quantization and qat	6 years ago
joanna.wozna.intel	37e0e7a96b	QAT int8 accuracy little improvement (#21074) ... test=develop	6 years ago
juncaipeng	fa522dffa0	Fix bug in add_quant_dequant_pass, test=develop (#21018) ... * Fix bug for inserting add_quant_dequant_op to same variable repetitively in add_quant_dequant_pass, test=develop	6 years ago
juncaipeng	175ba39c03	Add post_training_quantization (#20800) ... * add post training quantization, test=develop * specify the quantizable op type, test=develop	7 years ago
juncaipeng	f201b465ec	Move pool2d to add_quant_dequant_pass, test=develop (#20586) ... * move pool2d to add_quant_dequant_pass, test=develop	7 years ago
Liufang Sang	86c2c362ae	fix fuse_reduce_op quantization bug (#20306) ... * fix fuse_reduce_op quantization bug test=develop * close fuse_all_reduce_ops in PaddleSlim, test=develop	7 years ago
Michał Gallus	540935a825	[Bug-fix][1.6] Improve QAT accuracy (#20174) ... * Leave fake quantization around mul * Replace Fake with Real Quantized Mul * Gather all scales from fake_quantize_ops * Enable uint8 in conv_relu tensors * Disable int8 mul and restore fake mul * Fix buf for running QAT on VGG16 and 19	7 years ago
bingyanghuang	9de6772510	Follow comment of Merged QAT PR 18970 (#19979) ... * Follow Wangzhen's comment in PR 18970, test=develop * Review comments, test=develop * Leave fake quantization around mul test=develop * Replace Fake with Real Quantized Mul test=develop * Fix bug in quantize placement pass Nodes in the graph now have checked type instead of node name when they are to be marked for quantization test=develop	7 years ago
Wojciech Uss	4286a6270d	Add support for new QAT models (#18970) ... * Add support for new QAT models test=develop Co-Authored-By: Michał Gallus <michal.gallus@intel.com> Co-Authored-By: Wojciech Uss <wojciech.uss@intel.com> * fixed fps results test=develop * fix top5 accuracy drop problem * updated for new QAT models * skip quantizing average pooling - dirty but working * add missing pass * added missing conv+brelu fuse pass * removed a call to non-existent pass test=develop * renamed pass test=develop * Adjust finding pooling scale to newest QAT models * Remove unnecessary code from quantization_mkldnn_pass * Copy Pooling input scale to output scale in QAT * Refactor & remove unused code in QAT * Incorporate fp32 FC into QAT test=develop * Enable graph drawing with debug flag test=develop * Add tests for QATv2 * Fix paths for QATv2 models test=develop * Add option to save transformed int8 qat model test=develop * Remove redundant lines from qat mkldnn pass test=develop * Delegate disablement of avg pooling to qat test=develop * fix CI bug, test=develop * Follow Wangzhen's Review, test=develop * Update API.spec test=develop * Name False in (is_unsigned, TensorScale) tuple test=develop	7 years ago
whs	bdb3e376d0	[PaddleSlim] Enhence compressor api in PaddleSlim (#19894) ... 1. Support customize eval function instead of eval program. 2. Fix loading checkpoint in quantization strategy. 3. Support saving eval model when saving a checkpoint. 4. Fix decoder of loading context in PaddleSlim. 5. Fix restoring from the checkpoint of uniform prune strategy. 6. Support saving eval model and infer model during training. 7. Add ‘unitest’ for saving eval model, saving infer model and uniform pruning restoring from the checkpoint. 8. Fix pruning of depthwise_conv_grad op by updating the groups.	7 years ago
juncaipeng	b0ceed6fb4	add fake_quant_dequant_op for average pool2d, test=develop (#19880) ... * add fake_quant_dequant_op for average pool2d * add test	7 years ago
lidanqing	ba368bf696	clean up intel labeled TODOs (#19476) ... test=develop	7 years ago
liu zhengxi	32598ffd8f	Python infer api update and add unit test (#19353) ... * python inference api supports numpy and add unit test, fix unit test fail in test_slim_int8_googlenet and test_slim_int8_mobilenet	7 years ago
Zhen Wang	0fe72469ea	Add the max-pool2d quantization support and the partial quantization support. (#19310) ... * add pool2d quantization support, only for max-pooling. * add the partial quantization support.	7 years ago
bingyanghuang	a25be53cb5	QAT int8 MKL-DNN transformation pass with MUL (#18322)	7 years ago
翟飞跃	802ea50956	fix spelling errors (#17941) ... * fix spelling errors; test=develop * Update API.spec update md5 * Update API.spec * change the order of api;test=develop	7 years ago
Kaipeng Deng	96ee528e3e	fix logging basicConfig cannot be setting after import paddle (#17786) ... * fix logging unable. test=develop * unset sys.stdout for stream handler. test=develop * fix newly add basicConfig. test=develop * fix import error. test=develop	7 years ago
bingyanghuang	90ebce9ead	QAT int8 MKL-DNN transformation pass (#17819)	7 years ago
翟飞跃	993c703bcc	INT8 MKL-DNN v2 integrate to slim (#17634) ... * refactor PR 16865 * delete mergetool files * test=develop * test=develop * test=develop * test=develop * create dir for int8 model before call SaveOptimModel * test=develop * mkldnn int8 only support linux; test=develop * refine code; test=develop * remove comment; test=develop * refine code; test=develop * fix bug; test=develop * add exception for mkldnn_post_training_strategy * reuse int8v2 CAPI dataset; test=develop * fix accuracy check bug; test=develop * remove tab * convert files to unix format * test=develop * reduce CI time;test=develop * reduce CI time and refine code;test=develop * refine comment; test=develop * add cmake FLAGS;test=develop * remove predict_num;test=develop	7 years ago
Zhen Wang	3398f99608	Adding AddQuantDequantPass for TensorRT int8 (#17529) ... * add quant_dequant_pass, test=develop * Add quant_dequant before some ops, such as the elementwise_add op. This is required by TensorRT. test=develop	7 years ago