9iQSBSSKrSSKrSSKrSSKrSSKrSSKrSSKJrJr SSK J r SSK J r SSK JrJr SSKJr SS KJr SS KJrJrJr S S /rSS jr\r"SS \R8S9r"SS\5r"SS\ 5r!"SS\!5r""SS\!5r#"SS\!5r$g)N)flagslogging) get_cache_dir) parse_device)UnsupportedAPIErrorUnsupportedParamError)CachedProperty)Config)build_model_from_model_infobuild_runner_from_model_infoget_registered_model_info PaddleModel BaseModelcUcUc [S5eUbUbXR:wa [S5eOUcUb URn[U5n[ X!S9$![an[ [ U5S35UeSnAff=f) _create_modelNz;At least one of `model_name` and `config` must be not None.zgIf both `model_name` and `config` are not None, `model_name` should be the same as `config.model_name`.z is not a registered model name.) model_infoconfig) ValueError model_namerKeyErrorr reprr )rrres \/var/www/html/land-doc-ocr/venv/lib/python3.13/site-packages/paddlex/repo_apis/base/model.pyrr&sfnVWW  F$6 ** *  +   2&& .z: '* LL #J @ A  s A B$A<<Bc^\rSrSrSrSrSrSU4Sjjr\RSSj5r \RSSj5r \RSSj5r \RS 5r \RSS j5r\RSS j5r\R"S 5r\R"S 5r\S5r\S5r\S5r\S5r\S5r\S5r\S5r\S5rSrSrU=r $)r>z Abstract base class of Model. Model defines how Config and Runner interact with each other. In addition, Model provides users with multiple APIs to perform model training, prediction, etc. )trainevaluatepredictexportinfer compressionzsupported_{api_name}_optsc>[TU]5 Xl[U5Ul[ UR5UlUc![R"S5 [U5nX l UR5 g)z Initialize the instance. Args: model_name (str): A registered model name. config (base.config.BaseConfig|None): Config. Default: None. NaWe strongly discourage leaving `config` unset or setting it to None. Please note that when `config` is None, default settings will be used for every unspecified configuration item, which may lead to unexpected result. Please make sure that this is what you intend to do.) super__init__namerrrrunnerrwarningr r _patch_apis)selfrr __class__s rr&BaseModel.__init__Jsg  3J?34??C > OOl  J'F  c  [e)a& Train a model. Args: batch_size (int|None): Number of samples in each mini-batch. If multiple devices are used, this is the batch size on each device. If None, use a default setting. Default: None. learning_rate (float|None): Learning rate of model training. If None, use a default setting. Default: None. epochs_iters (int|None): Total epochs or iterations of model training. If None, use a default setting. Default: None. ips (str|None): If not None, enable multi-machine training mode. `ips` specifies Paddle cluster node ips, e.g., '192.168.0.16,192.168.0.17'. Default: None. device (str): A string that describes the device(s) to use, e.g., 'cpu', 'gpu', 'gpu:1,2'. Default: 'gpu'. resume_path (str|None): If not None, resume training from the model snapshot corresponding to the weight file `resume_path`. If None, use a default setting. Default: None. dy2st (bool): Whether to enable dynamic-to-static training. Default: False. amp (str): Optimization level to use in AMP training. Choices are ['O1', 'O2', 'OFF']. Default: 'OFF'. num_workers (int|None): Number of subprocesses to use for data loading. If None, use a default setting. Default: None. use_vdl (bool): Whether to enable VisualDL during training. Default: True. save_dir (str|None): Directory to store model snapshots and logs. If None, use a default setting. Default: None. Returns: paddlex.repo_apis.base.utils.subprocess.CompletedProcess NotImplementedError) r+ batch_size learning_rate epochs_itersipsdevice resume_pathdy2stamp num_workersuse_vdlsave_dirkwargss rrBaseModel.traines b"!r.c [e)a Evaluate a model. Args: weight_path (str): Path of the weights to initialize the model. batch_size (int|None): Number of samples in each mini-batch. If multiple devices are used, this is the batch size on each device. If None, use a default setting. Default: None. ips (str|None): If not None, enable multi-machine evaluation mode. `ips` specifies Paddle cluster node ips, e.g., '192.168.0.16,192.168.0.17'. Default: None. device (str): A string that describes the device(s) to use, e.g., 'cpu', 'gpu', 'gpu:1,2'. Default: 'gpu'. amp (str): Optimization level to use in AMP training. Choices are ['O1', 'O2', 'OFF']. Default: 'OFF'. num_workers (int|None): Number of subprocesses to use for data loading. If None, use a default setting. Default: None. Returns: paddlex.repo_apis.base.utils.subprocess.CompletedProcess r0)r+ weight_pathr2r5r6r9r:r=s rrBaseModel.evaluates @"!r.c [e)a Make prediction with a pre-trained model. Args: weight_path (str): Path of the weights to initialize the model. input_path (str): Path of the input file, e.g. an image. device (str): A string that describes the device to use, e.g., 'cpu', 'gpu'. Default: 'gpu'. save_dir (str|None): Directory to store prediction results. If None, use a default setting. Default: None. Returns: paddlex.repo_apis.base.utils.subprocess.CompletedProcess r0)r+r@ input_pathr6r<r=s rr BaseModel.predict "!r.c [e)z Export a pre-trained model. Args: weight_path (str): Path of the weights to initialize the model. save_dir (str): Directory to store the exported model. Returns: paddlex.repo_apis.base.utils.subprocess.CompletedProcess r0)r+r@r<r=s rr!BaseModel.exports "!r.c [e)a Make inference with an exported inference model. Args: model_dir (str): Path of the exported inference model. input_path (str): Path of the input file, e.g. an image. device (str): A string that describes the device(s) to use, e.g., 'cpu', 'gpu'. Default: 'gpu'. save_dir (str|None): Directory to store inference results. If None, use a default setting. Default: None. Returns: paddlex.repo_apis.base.utils.subprocess.CompletedProcess r0)r+ model_dirrCr6r<r=s rr"BaseModel.inferrEr.c [e)a Perform quantization aware training (QAT) and export the quantized model. Args: weight_path (str): Path of the weights to initialize the model. batch_size (int|None): Number of samples in each mini-batch. If multiple devices are used, this is the batch size on each device. If None, use a default setting. Default: None. learning_rate (float|None): Learning rate of QAT. If None, use a default setting. Default: None. epochs_iters (int|None): Total epochs of iterations of model training. If None, use a default setting. Default: None. device (str): A string that describes the device(s) to use, e.g., 'cpu', 'gpu'. Default: 'gpu'. use_vdl (bool): Whether to enable VisualDL during training. Default: True. save_dir (str|None): Directory to store the results. If None, use a default setting. Default: None. Returns: tuple[paddlex.repo_apis.base.utils.subprocess.CompletedProcess] r0) r+r@r2r3r4r6r;r<r=s rr#BaseModel.compressions F"!r.c#p# URnURSnSRURR 5U/5nUS-n[ R (d^[R"[5S9n[RRXT5n[USSS9 SSS5 Uv SSS5 g[RR[5U5n[USSS9 SSS5 Uv g!,(df  N]=f!,(df  g=f!,(df  N5=f7f)Nr_z.ymldirwutf-8encoding r,rjoin__name__lowerrDEBUGtempfileTemporaryDirectoryrospathopen)r+clsrtagyaml_file_nametdr]s r_create_new_config_file!BaseModel._create_new_config_filesnn__\2 hh **,j9:v{{,,ARww||B7$g67 BA 77<< @DdC'23J76BA32NA=D6?+D*D, D8^[R"T5UU4Sj5nU$)NcV>TRnTRn[USUS35e)Nz does not support `z`.)r'rWr)argsr=rrr bnd_methodr+s r_unavailable_apiJBaseModel._patch_apis.._make_unavailable.._unavailable_apims5!YY %..)!l"5hZrBr.) functoolswraps)rrr+s` r_make_unavailable0BaseModel._patch_apis.._make_unavailablels& __Z ( ) $ #r.ct>^^[R"T5UU4Sj5nTRn/mUS:XayTRnUbhSU;aTR [ USSS95 SU;aTR [ US55 SU;aTR [US55 U$US:XaVTRnUbESU;aTR [ USSS95 SU;aTR [US55 U$US :Xa3TRnUb"SU;aTR [ USS S95 U$US :Xa3TRnUb"SU;aTR [ USS S95 U$US :Xa3TRnUb"SU;aTR [ USSS95 U$T$) Nc>[RRT5nUR"U0UD6nURnUR R 5HHnURU;dMURURLdM0URXER'MJ THnURU5 M T"U0UD6$N) inspect Signature from_callablebind arguments parametersvaluesr'defaultemptycheck) rr=sigbnd_args args_dictprrcheckss r_api_with_prechecksJBaseModel._patch_apis.._add_prechecks.._api_with_prechecksxs''55jA88T4V4$.. ..0AvvY.199AGG3K,-II &&)1 $EKK *$"42622r.rr6Tcheck_mcr8r9rr Fr"r#) rrrWrvappend _CheckDevice _CheckDy2St _CheckAMPrzr}rr)rrrroptsrr+s` @r_add_prechecks-BaseModel._patch_apis.._add_prechecksws __Z ( 3) 3 "**HF7"00#4' l4>D&QR$ k$w-&@A} iU &<=4' &3Z'33#4' l4>D&QR} iU &<=&' &%Y&22#4' l4>E&RS' &W$00#4' l4>E&RS' &]*66#4' l4>D&QR' &"!r.)rkset_API_FULL_LISTgetattrsetattrr CHECK_OPTS)r+rrrk avail_api_setrrapi decorated_apis` rr*BaseModel._patch_apisks $7 'r,,  % 3 34M 3 34M++H$)C,!2# 6 h6###%33$7MD(:,r.)rrr'r(r) NNNNgpuNFOFFNTN)NNrrN)rN)NNNrTN)!rW __module__ __qualname____firstlineno____doc__rrtr&abcabstractmethodrrr r!r"r# contextlibcontextmanagerrcrhcached_propertyrkrvrzr}rrr staticmethodrr*__static_attributes__ __classcell__r,s@rrr>sXN&A#6   0"0"d   ""B """  " " """ """"H"";;          DDQQX;X;r.) metaclassc(\rSrSrSrSrSrSrSrg) _CheckFailediTc(XlX lX0lgrarg_namearg_val legal_vals)r+rrrs rr&_CheckFailed.__init__s  $r.cTSURSURSUR3$)N`z*` is expected to be one of or conforms to z , but got )rrrrns r__str___CheckFailed.__str__s.4==/!KDOOK\\fgkgsgsftuur.rN) rWrrrrcheck_failed_errorr&rrr.rrrs% vr.rc2^\rSrSrSrU4SjrSrSrU=r$)_APICallArgsCheckeric.>[TU]5 Xlgr)r%r&r)r+rr,s rr&_APICallArgsChecker.__init__s $r.c[e)rr0)r+rs rr_APICallArgsChecker.checks!!r.)r rWrrrrr&rrrrs@rrrs%""r.rc6^\rSrSrSrSU4SjjrSrSrU=r$)ric0>[TU]U5 X lgr)r%r&r)r+rrr,s rr&_CheckDevice.__init__s $ r.cSU;deUSnUGb>[U5up4UR(d&X0R;a[SX R5egUS:wa&X0R;a[SX R5egUS3nUS3nUS3n[ U5S::aFXPR;a6X`R;a&XpR;a[SX R5egggSU;deUSb&XpR;a[SX R5egX`R;a&XpR;a[SX R5eggg) rr6Nr_n1c1_n1cx_nxcxr r5)rrrrr)r+rr6 device_typedev_ids n1c1_desc n1cx_desc nxcx_descs rr_CheckDevice.checks`4h  #/#7 K==oo5&xII6%'"//9*8V__MM:$/-u 5I#.-u 5I#.-u 5I7|q(%__< ) @ ) @".x"QQ!A!A= %},};2(?&28V__&U U @ !* @$-__$D&28V__&U U%E!A r.r)Frrs@rrrs!**r.rc\rSrSrSrSrSrg)ric.SU;deUSn[UR[5(a4[UR5S:Xde[ URS5nO[ UR5nUbU(aU(d [ SX#/5eggg)rr8r rN) isinstancerlistrboolr)r+rr8 support_dy2sts rr_CheckDy2St.checks$W  doot , ,t'1, ,, !34M 1M  ]"7E?CC+u r.rNrWrrrrrrrr.rrrs  r.rc\rSrSrSrSrSrg)ri(c~SU;deUSnUb-US:wa&X R;a[SX R5eggg)rr9Nr)rr)r+rr9s rr_CheckAMP.check+sI}}5k ?e|?? :"5#??!;| r.rNrrr.rrr(s r.r)NN)%rrrrr\rZutilsrr utils.cacher utils.devicer utils.errorsrr utils.miscr rrr registerr rr__all__rrABCMetar Exceptionrobjectrrrrrr.rrs   #((F; + &M* E;#++E;P v9 v "& "1&1h%& # r.