\;jVddlZddlZddlZddlZddlZddlmZmZddlm Z ddl m Z m Z m Z mZmZmZmZe eej&Zdd lmZdd lmZdd l mZmZmZmZgd Zd gZgdZddgZ dgZ!e"ejFjIZ%e"ejLjIZ'gdZ(dZ)d ddddggddddddddddZ*dZ+dZ,dZ-ddddddddddidddddfdZ.d!d Z/y)"N)IrGraphcore) get_logger)AddQuantDequantForResidualAddQuantDequantPassConvertToInt8PassOutScaleForInferencePassOutScaleForTrainingPassQuantizationFreezePassQuantizationTransformPass)level) quant_config)PostTrainingQuantizationProgram)AddQuantDequantForInferencePassAddQuantDequantPassV2QuantizationTransformPassV2QuantWeightPass)abs_maxchannel_wise_abs_max range_abs_maxmoving_average_abs_maxr)rrrrrint8)mulconv2dpool2ddepthwise_conv2delementwise_add leaky_reluz*./mapping_table_for_saving_inference_model skip_quant)rrri'g?FT)weight_quantize_typeactivation_quantize_type weight_bitsactivation_bitsnot_quant_patternquantize_op_typesdtype window_size moving_rate for_tensorrtis_full_quantize onnx_formatquant_post_firstscale_trainablecttd5}|j}tj|}|cdddS#1swYyxYw)Nr)openVARS_MAPPING_TABLEreadjsonloads)filedatas kG:\00. PROJECTS\API\Inventory\templateJSON\kerjaOCR\Lib\site-packages\paddle/static/quantization/quanter.py load_dictr<us7  # &$yy{zz$ ' & &s 'AA cttd5}|jtj|dddy#1swYyxYw)Nw)r4r5writer7dumps)tabler9s r; save_dictrB|s0  # &$ 4::e$% ' & &s %?Achtjt}|j|t |dt rt |dt sJd|drt }t}d}nt}t}d}|d|vsJdj|d|||d|vsJd j|d||t |d tsJd |d d k\r|d d ksJdt |dtsJd|dd k\r|dd ksJdt |dttfsJdt |dtsJd|dr t|d<nS|drttz|d<n=|dD]5}|tvr |tvrJdj|ttzt |dtsJd|dt vsJddj#t zt |dtsJdt |dt$sJd|S)z check if user's configs are valid. Args: user_config(dict): user's config. Return: configs(dict): final configs will be used. r-r.z8'for_tensorrt' and 'is_full_quantize' must both be bool'TensorRT PaddleLiter$z6Unknown weight_quantize_type: {}. {} only supports {} r%z9Unknown activation_quantize_type: {}. {} only supports {}r&zweight_bits must be int value.rz'weight_bits should be between 1 and 16.r'z"activation_bits must be int value.z+activation_bits should be between 1 and 16.r(z%not_quant_pattern must be list or strr)z quantize_op_types must be a listzF{} is not support, now support op types are {}r*zdtype must be a str.zdtype can only be  r+z^window_size must be int value, window size for 'range_abs_max' quantization, default is 10000.r,zYmoving_rate must be float value, The decay coefficient of moving average, default is 0.9.)copydeepcopy_quant_config_defaultupdate isinstancebool"WEIGHT_QUANTIZATION_TYPES_TENSORRT&ACTIVATION_QUANTIZATION_TYPES_TENSORRTWEIGHT_QUANTIZATION_TYPESformatintliststrTENSORRT_OP_TYPESTRANSFORM_PASS_OP_TYPESQUANT_DEQUANT_PASS_OP_TYPES VALID_DTYPESjoinfloat) user_configconfigs weight_typesactivation_typesplatformop_types r;_parse_configsras mm12G NN; gn-t 4"#T:BAB  ~9 A0 4&'<7?FF&'<7 *+/??BII*+X7G?   ('(   !# (>"(D101D !"C ,+,  !"a'G4E,F",L545L #$tSk /./  #$d *)* ~'8#$ # $ #&A A #$23G::22 55;V03NN6 4 gg& ,D.DD , 7 | +-ACHHE. +   hgh    cbc  Nc |stjjn|}|t}n"t |t sJdt |}tjd|g}g}|rdn|d }|dr{|ryd| vr|d| d<| tjjn| }t||fd||d |dd | }|j}|j}t|j}ntt!j"|j$| }t|j}g}g}d |vr$|d r|d j&}|d j(}n>|dD]6}|t*vr|j-||t.vs&|j-|8t1|d kDr|drt2nt4}|d(id|d|d|dd|dd|dd|dd|dd|dd|d|dd|d|d|d|d| d| d|}|D]}|j7|t9|||d| }|D]} |j7t1|d kDrA|drt:nt<}!|!|||d|d|d||!}"|D]}|"j7|t?|||d||"}#|D]}|#j7|||U|sS|dsNtjd#tjd$t@d%|D]}tC|jD| r|jGd&d'|s| r|jI}$n)tjjK|jL}$|r|$|||fS|$S))a7Add quantization and dequantization operators to "program" for quantization training or testing. Args: program(paddle.static.Program): training or testing ``program``. place(paddle.CPUPlace or paddle.CUDAPlace): This parameter represents the executor run on which device. config(dict, optional): configs for quantization. if None, will use default config. Default: None. scope(paddle.static.Scope): Scope records the mapping between variable names and variables, similar to brackets in programming languages. Usually users can use `paddle.static.global_scope `_. When ``None`` will use `paddle.static.global_scope() `_ . Default: ``None``. for_test(bool): If the 'program' parameter is a test program, this parameter should be set to ``True``. Otherwise, set to ``False``.Default: False weight_quantize_func(function): Function that defines how to quantize weight. Using this can quickly test if user's quantization method works or not. In this function, user should both define quantization function and dequantization function, that is, the function's input is non-quantized weight and function returns dequantized weight. If None, will use quantization op defined by 'weight_quantize_type'. Default is None. act_quantize_func(function): Function that defines how to quantize activation. Using this can quickly test if user's quantization method works or not. In this function, user should both define quantization and dequantization process, that is, the function's input is non-quantized activation and function returns dequantized activation. If None, will use quantization op defined by 'activation_quantize_type'. Default is None. weight_preprocess_func(function): Function that defines how to preprocess weight before quantization. Using this can quickly test if user's preprocess method works or not. The function's input is non-quantized weight and function returns processed weight to be quantized. If None, the weight will be quantized directly. Default is None. act_preprocess_func(function): Function that defines how to preprocess activation before quantization. Using this can quickly test if user's preprocess method works or not. The function's input is non-quantized activation and function returns processed activation to be quantized. If None, the activation will be quantized directly. Default is None. optimizer_func(function): Fuction return a optimizer. When 'is_test' is False and user want to use self-defined quantization function and preprocess function, this function must be set. Default is None. exe(paddle.static.Executor): If user want to use self-defined quantization function and preprocess function, exe must be set for initialization. Default is None. return_program(bool): If user want return value is a Program rather than Compiled Program, This argument should be set True. Default is False. draw_graph(bool): whether to draw graph when quantization is initialized. In order to prevent cycle, the ERNIE model needs to be set to True. Default is False. return_scale_dict(bool): If user want to return scale dict, model_type and pattern_ops, this argument should be set True. Default is False. scale_dict(dict): Use scale dict to initialize scales in program. Default is None. model_type(str): Model type can be 'transformer' or 'non-transformer'. If model type is transformer, patterns will be analyzed. Default is None. pattern_ops(dict): Pattern_ops contain pattern name and corresponding ops. Default is None. Returns: paddle.static.CompiledProgram | paddle.static.Program: Program with quantization and dequantization ``operators`` config must be dictzquant_aware config Tr1r0quantizable_op_typer)F ) freeze_modelskip_tensor_listsame_scale_tensor_list batch_nums scale_dict return_graphfor_testrrr/scopeplacer&r'r%r$r+r, skip_patternr(weight_quantize_funcact_quantize_funcweight_preprocess_funcact_preprocess_funcoptimizer_funcexecutoris_test)rorp quant_bitsrx)rorpr,ryrqrerx)rorpr,rxrkzyWhen a preprocess_func is used in quant_aware, Need to save a mapping table to match variable names in the convert phase.zThe mapping table is saved as 'z'.z./z graph.pdf)'paddlestatic global_scoperJrLdictra_loggerinfoExecutorrquantize _scale_dictrSall_sub_graphsrrGraphdescweight_quant_operation_types activation_quant_operation_typesrVappendrWlenrrapplyr rr r r5rBout_node_mapping_tabledraw to_programCompiledProgramgraph)%programrpconfigrornrrrsrtrurvrwreturn_program calib_config draw_graphreturn_scale_dictrk model_type pattern_opsrhrirxexepost_training_quantization main_graph sub_graphstransform_pass_opsquant_dequant_opsr`transform_functransform_pass sub_graph residual_passsubgraphqdq_funcquant_dequant_passout_scale_training_pass quant_programs% r; quant_awarersV16FMM & & (5E ~&&$'>)>>'' LL&vh/0d/@(A$AG  !h 4289L2ML. /*2*:fmm$$&%D   & -#9! &  & "088: /;; *3356 TZZ 5I *3356  V #~(>!'"** !'!.. ""5655&--g6 ;;%,,W5 7 ! "Q &-(,.  ,#=1!''8 9  *00J)K  &,,B%C #=1#=1%7$$78&:#4(>%8 . "!" #N(( $$Y/(7!"34 M'##I.'  !A %-(&(  "*"=1!"34#$78$5" ( "((3(6=)  %%i0  +/B/N}%  H   67I6J"MN#I i66 7$k*>"--/  55j6F6FG j*kAArbc|stjjn|}|t}n"t |t sJdt |}tjd|ttj|jd}|drjt||}|jD]}|j|t!|||d}|jD]}|j|nt#|}|jD]}|j|t%|||d |d|d  } t&j(j+t,rt/|_|jD]}| j||j3} | j5} | j6D]} | j8d k(s| j;d j<} | j?| }g}|j@jCD]4\}}|jDs| jG||jI|6|D]}|jK||jM| jOd| jjQd||r |drJd|rGtS||}|jD]}|j||j3}| |fS| S)a convert quantized and well-trained ``program`` to final quantized ``program``that can be used to save ``inference model``. Args: program(paddle.static.Program): quantized and well-trained ``test program``. place(paddle.CPUPlace or paddle.CUDAPlace): This parameter represents the executor run on which device. config(dict, optional): configs for convert. if set None, will use default config. It must be same with config that used in 'quant_aware'. Default is None. scope(paddle.static.Scope, optional): Scope records the mapping between variable names and variables, similar to brackets in programming languages. Usually users can use `paddle.static.global_scope `_. When ``None`` will use `paddle.static.global_scope() `_ . Default: ``None``. save_int8: Whether to return ``program`` which model parameters' dtype is ``int8``. This parameter can only be used to get model size. Default: ``False``. Returns: Tuple : freezed program which can be used for inference. when ``save_int8`` is False, return ``freezed_program(paddle.static.Program)``. when ``save_int8`` is True, return ``freezed_program(paddle.static.Program)`` and ``freezed_program_int8(paddle.static.Program)`` rdzconvert config Trmr/r')rorpry)ror&r$)rorpr&r'r$while sub_blockXzQWhen onnx_format=True, already saved int8 weight,so you can't set save_int8=True.)rorp)*r{r|r}rJrLr~rarrrrrrrrrrr rospathexistsr5r<rr global_blockopstypeattridblockvarsitems persistable_clone_variabler _remove_varextendinput set_inputr )rrprro save_int8 test_graphquant_weight_passrout_scale_infer_pass freeze_passfreezed_programr_op _block_id_block persistables_name_varconvert_int8_passfreezed_program_int8s r;convertrs816FMM & & (5E ~&&$'>)>>'' LL?6(+,GLL1DAJ m+E59#224I  # #I .5>u8I1J $224I & &y 15 8eD#224I & &y 15-}-"#45!'(>!?   77>>, -09 J -#224I   i (5!++-O#//1L 88w -00I$**95FL%{{002 t## 006 ''. 3&""5)&    # / HH  sL 1  f]+[Z[ -EG#224I  # #I .5)446 444rb)NNF)0rHr7loggingrr{base.frameworkrr log_helperrquantization_passr r r r r rr__name__INFOrrrrrrrrrPrNACTIVATION_QUANTIZATION_TYPESrOrXrS#SUPPORT_WEIGHT_QUANTIZATION_OP_DICTkeysrV SUPPORT_ACT_QUANTIZATION_OP_DICTrWrUr5rJr<rBrarrrzrbr;rsU   +# XW\\ 2G '=%="!*& x 4499;#11668B3 8&> 9@& Z@    %Xv]rb