o }oi@sRdZddlmZddlZddlZddlZddlZddlZddlZddl m Z m Z ddl m Z mZddlmZddlmZmZddlmZdd lmZdd lmZdd lmZmZmZmZmZm Z ddl!Z!ddl"Z"ddl#Z#dd l$m%Z%m&Z&dd l$m'Z(ddl$m)Z)m*Z*m+Z+ddl,m-Z-m.Z.ddl/Z/ddl0m1Z1ddl2m3Z3ddl4m5Z5ddl6m7Z7m8Z8ddl9m:Z:ddl;mZ>ddl?m@Z@mAZAgdZBdaCdaDdZEgdZFd;dd ZGdz3Interfaces common to all Neural Modules and Models.) annotationsN)ABCabstractmethod)MappingSequence)contextmanager) dataclassfield)Enum)total_ordering)Path)AnyDictListOptionalTupleUnion)_CACHED_NO_EXISTHfApi) get_token)hf_hub_downloadsnapshot_downloadtry_to_load_from_cache) DictConfig OmegaConf)HuggingFaceFileIO) NEMO_DEFAULT_MODEL_CARD_TEMPLATE)SaveRestoreConnector) NeuralTypeNeuralTypeComparisonResult)logging)maybe_download_from_cloud)resolve_cache_dir)import_class_by_pathmaybe_update_config_version)TypingFileIOModel Serialization typecheckPretrainedModelInfoT)znemo.collections.z nemo.core.z nemo.utils.znemo.lightning.ztests.collections.z torch.nn.z torch.optim.ztorch.utils.data.zlightning.pytorch.callbacks.zlightning.pytorch.loggers.zlightning.pytorch.strategies.zlightning.pytorch.accelerators.z omegaconf.z megatron. target_pathstrreturnboolcs$ttsdStfddtDS)NFc3s|]}|VqdSN) startswith).0prefixr+L/home/ubuntu/.local/lib/python3.10/site-packages/nemo/core/classes/common.py Rsz%_is_target_allowed..) isinstancer,anyALLOWED_TARGET_PREFIXESr3r4r3r5_is_target_allowedOs r: config_noder cCst|tr(d|vr|d}t|std|d|D]\}}t|qdSt|tr;t|ts=|D] }t|q4dSdSdS)N_target_z Instantiation of unsafe target 'z' is blocked. The '_target_' must point to a class or function within an approved namespace. This restriction is in place to prevent potential arbitrary code execution.)r7rr: ValueErroritems"_validate_config_targets_recursiverr,)r;r+keyvalueitemr4r4r5r?Us    r?configrcOs*|durt|tjj|g|Ri|S)z A wrapper around hydra.utils.instantiate that first validates all _target_ fields in the config against an allow-list of prefixes. N)r?hydrautils instantiate)rCargskwargsr4r4r5safe_instantiatefsrIcCtS)z/ Getter method for typechecking state. _TYPECHECK_ENABLEDr4r4r4r5is_typecheck_enabledprMcCrJ)z9 Getter method for typechecking semantics state. !_TYPECHECK_SEMANTIC_CHECK_ENABLEDr4r4r4r5is_semantic_typecheck_enabledwrNrQc@seZdZUdZded<ded<eddZded<eddZded <eddZd ed <eddZ ded <eddZ ded <ddZ dS)TypecheckMetadataaq Metadata class for input/output neural types. # Primary attributes original_types: Preserve the dictionary of type information provided. ignore_collections: For backward compatibility, container support can be disabled explicitly using this flag. When set to True, all nesting is ignored and nest-depth checks are skipped. # Derived attributed mandatory_types: Sub-dictionary of `original_types` which contains only those types which are mandatory to include when calling the function. base_types: Dictionary of flattened `str: NeuralType` definitions, disregarding the nest level details into appropriate arguments. container_depth: Dictionary mapping `str: int` - such that the valid depth of the nest of this neural type is recorded. has_container_types: Bool flag declaring if any of the neural types declares a container nest in its signature. is_singular_container_type: Bool flag declaring if this is a single Neural Type with a container nest in its signature. Required for supporting python list expansion in return statement. zDict[str, NeuralType]original_typesr.ignore_collectionsF)initmandatory_types base_typeszDict[str, int]container_depthhas_container_typesis_singular_container_typecCsd}|jD] }t|ttfrd}nq||_|jr&t|jdkr&d|_nd|_|jrpi|_i|_ |j D]7\}}d}t|ttfrdt|dkrUt d|d|d|d}|d7}t|ttfsD||j|<||j |<q7n|j|_dd |j D|_ d d |j D|_ dS) NFTrz Neural Type `z`: z definition contains more than one element when declaring the nested container structure. Please ensure that you have only 1 NeuralType inside of the entire nested structure definition.cSsi|]}|dqS)rr4)r1type_keyr4r4r5 z3TypecheckMetadata.__post_init__..cSsi|] \}}|js||qSr4)optional)r1r\type_valr4r4r5r]s  )rSvaluesr7listtuplerYlenrZrWrXr> TypeErrorkeysrV)selfrYr`r\depthr4r4r5 __post_init__s>   zTypecheckMetadata.__post_init__N) __name__ __module__ __qualname____doc____annotations__r rVrWrXrYrZrir4r4r4r5rR~s  rRc@sXeZdZdZedddZedddZdd d Zdd d ZddddZ ddddZ dS)r%z< An interface which endows module with neural types r-Optional[Dict[str, NeuralType]]cCdS)z/Define these to enable input neural type checksNr4rgr4r4r5 input_typeszTyping.input_typescCrp)z0Define these to enable output neural type checksNr4rqr4r4r5 output_typesrszTyping.output_typesNFc Ks<|durt||d}t|}t|j}t||ks t||kr0tdt|d|d|d|D]\}}||vrHtd|d|t|d rtr|j| |j t j t j fvr|| |j d d ||d |j d |g} t|j|jjD]\} } | | dd| d| dd| dqt|j jjD]\} } | d| d| dd| dqtd| t|dr|j} |j|j} |}| durt| t| krtd|d|jjd|j|jd| q4t|tst|trt|D]\}} |j||d|dq q4dSdS)a This function does a few things. 1) It ensures that len(self.input_types ) <= len(kwargs) <= len(self.input_types). 2) For each (keyword name, keyword value) passed as input to the wrapped function: - Check if the keyword name exists in the list of valid self.input_types names. - Check if keyword value has the `neural_type` property. - If it does, then perform a comparative check and assert that neural types are compatible (SAME or GREATER). - Check if keyword value is a container type (list or tuple). If yes, then perform the elementwise test of neural type above on each element of the nested structure, recursively. Args: input_types: Either the `input_types` defined at class level, or the local function overridden type definition. ignore_collections: For backward compatibility, container support can be disabled explicitly using this flag. When set to True, all nesting is ignored and nest-depth checks are skipped. kwargs: Dictionary of argument_name:argument_value pairs passed to the wrapped function upon call. NrSrTz$Number of input arguments provided (z#) is not as expected. Function has z total inputs with z mandatory inputs.zInput argument z? has no corresponding input_type match. Existing input_types = neural_typez :zInput type expected : zInput type found : z Argument: z input param_z : rz: r[ shape!Input shape mismatch occured for in module  : Input shape expected =  | Input shape found : rhname)rRrdrVrer>rfhasattrrQrWcomparervrSAMEGREATER enumerate elements_typetype_parametersinsertappendjoinryaxes __class__rjr7rbrc_Typing__check_neural_type)rgrrrTrHmetadatatotal_input_typesmandatory_input_typesr@rA error_msgi dict_tuple value_shape type_shaperindvalr4r4r5_validate_input_typesst      ,&   zTyping._validate_input_typesc Cs|durt||d}t|j}t|j}t|ttfr"|}n|g}|jr)n t|t|ks9t|t|krIt d t|t|t||t|tst|tsz |dd|_ Wn t yeYnwt |dr|j}|ddj} |dd} | durt|t| krt d| d|jjd | d |dSdSdS|jrd} t|dkrt|tkrd} t|D]\} } |j| || |ddd qdSt|D]\} } |j| |d|| dd qdSdS) a This function does a few things. 1) It ensures that len(out_object) == len(self.output_types). 2) If the output is a tensor (or list/tuple of list/tuple ... of tensors), it attaches a neural_type to it. For objects without the neural_type attribute, such as python objects (dictionaries and lists, primitive data types, structs), no neural_type is attached. Note: tensor.neural_type is only checked during _validate_input_types which is called prior to forward(). Args: output_types: Either the `output_types` defined at class level, or the local function overridden type definition. ignore_collections: For backward compatibility, container support can be disabled explicitly using this flag. When set to True, all nesting is ignored and nest-depth checks are skipped. out_objects: The outputs of the wrapped function. Nrua<Number of output arguments provided ({}) is not as expected. It should be larger or equal than {} and less or equal than {}. This can be either because insufficient/extra number of output NeuralTypes were provided,or the provided NeuralTypes {} should enable container support (add '[]' to the NeuralType definition)rr[ry"Output shape mismatch occured for r{ : Output shape expected =  | Output shape found : r~)rRrbrWr>rVr7rcrZrdreformatrv Exceptionrryrrrjtyper_Typing__attach_neural_type)rg out_objectsrTrtrout_types_listmandatory_out_types_list out_containerrrrrhrresr4r4r5!_attach_and_validate_output_types<s^         Xz(Typing._attach_and_validate_output_typesrrRrhintrr,c Cs$t|ts t|tr|D] }|j|||d|dq dS|j|}|js>||j|kr>td|d|d|d|j|t|drct rc| |j t j t jfvrct| |j d |d |j t|d r|j}|j}|durt|t|krtd |d |jjd|d|dSdSdS)a Recursively tests whether the obj satisfies the semantic neural type assertion. Can include shape checks if shape information is provided. Args: obj: Any python object that can be assigned a value. metadata: TypecheckMetadata object. depth: Current depth of recursion. name: Optional name used of the source obj, used when an error occurs. r[)rNzWhile checking input neural types, Nested depth of value did not match container specification: Current nested depth of NeuralType '' ():  Expected nested depth : rvz : Input type expected = z | Input type found : ryrzr{r|r})r7rcrbrrWrTrXrerrQrrvrrrryrrdrrj rgobjrrhrelemr`rrr4r4r5__check_neural_types\    zTyping.__check_neural_typec Cst|ts t|tr|D] }|j|||d|dq dS|j|}|js>||j|kr>td|d|d|d|j|z||_Wn t yLYnwt |drv|j }|j }|durxt |t |krztd |d |jjd |d |dSdSdS) aq Recursively attach neural types to a given object - as long as it can be assigned some value. Args: obj: Any python object that can be assigned a value. metadata: TypecheckMetadata object. depth: Current depth of recursion. name: Optional name used of the source obj, used when an error occurs. r[r~NzWhile attaching output neural types, Nested depth of value did not match container specification: Current nested depth of NeuralType 'rrrryrr{rr)r7rcrbrrWrTrXrervrrryrrdrrjrr4r4r5__attach_neural_typesF     zTyping.__attach_neural_type)r-ro)NF)FNr/)rrRrhrrr,) rjrkrlrmpropertyrrrtrrrrr4r4r4r5r%s    X p6r%c@s2eZdZed dddZdd d Zed d ZdS)r(NrC 'DictConfig'trainerOptional['Trainer']c Cstrt|trtj|dd}t|}t|dt|}d|vs%d|vr1d|vr1tr1t|d}nd|vr=tr=t|d}nd}d }d|vr|d}d}z t |}t ||rW|}t |}|re|||d }n||d }Wn(t y}zt} d |d |d| }t|dWYd}~nd}~ww|durzt |}|r|||d }n||d }Wnt y}z |rt||d}~wwt|ds||_|S)z8Instantiates object using DictConfig-based configurationTresolveclstargetparamsrCr<N)cfgr)rz*Model instantiation failed! Target class: z Error(s): rxz Falling back to `cls`._cfg) _HAS_HYDRAr7rr to_containercreate set_structr$rIr# issubclassr(_inspect_signature_for_trainerr traceback format_excr debugerrorrr) rrCrinstance prev_errortarget_cls_path imported_clsaccepts_traineretbr4r4r5from_config_dicts^              zSerialization.from_config_dictr-cCsft|dr/|jdur/tr)t|jtr)tj|jdd}t|}t|dt |}||_|jSt d)z3Returns object's configuration to config dictionaryrNTrz[to_config_dict() can currently only return object._cfg but current object does not have it.) rrrr7rrrrrr$NotImplementedError)rgrCr4r4r5to_config_dictQs  zSerialization.to_config_dictcCs,t|drt|j}d|jvrdSdSdS)N__init__rTF)rinspect signaturer parameters)r check_clsrr4r4r5rds   z,Serialization._inspect_signature_for_trainerr/)rCrrr)r-r)rjrkrl classmethodrrrr4r4r4r5r(s  Ar(c@sJeZdZdddZe      ddddZedddZdddZdS)r& save_pathr,cCt)a5 Standardized method to save a tarfile containing the checkpoint, config, and any additional artifacts. Implemented via :meth:`nemo.core.connectors.save_restore_connector.SaveRestoreConnector.save_to`. Args: save_path: str, path to where the file should be saved. r)rgrr4r4r5save_toqszFileIO.save_toNTF restore_pathoverride_config_path Optional[str] map_locationOptional['torch.device']strictr. return_configrrsave_restore_connectorrcCr)a Restores model instance (weights and configuration) from a .nemo file Args: restore_path: path to .nemo file from which model should be instantiated override_config_path: path to a yaml config that will override the internal config file or an OmegaConf / DictConfig object representing the model config. map_location: Optional torch.device() to map the instantiated model to a device. By default (None), it will select a GPU if available, falling back to CPU otherwise. strict: Passed to load_state_dict. By default True return_config: If set to true, will return just the underlying config of the restored model as an OmegaConf DictConfig object without instantiating the model. trainer: An optional Trainer object, passed to the model constructor. save_restore_connector: An optional SaveRestoreConnector object that defines the implementation of the restore_from() method. r)rrrrrrrrr4r4r5 restore_from{szFileIO.restore_frompath2yaml_filecCs&t|trt|}|j|dSt)z Instantiates an instance of NeMo Model from YAML config file. Weights will be initialized randomly. Args: path2yaml_file: path to yaml file with model configuration Returns: r)rr(rloadrr)rrconfr4r4r5from_config_files  zFileIO.from_config_filecCsbt|dr.t|j|_t|ddd}tj|j|ddWddS1s'wYdSt)z Saves current instance's configuration to YAML config file. Weights will not be saved. Args: path2yaml_file: path2yaml_file: path to yaml file where model model configuration will be saved Returns: rwutf-8)encodingT)rCfrN)rr$ropenrsaver)rgrfoutr4r4r5to_config_files  "zFileIO.to_config_file)rr,)NNTFNN)rr,rrrrrr.rr.rrrr)rr,)rjrkrlrrrrrr4r4r4r5r&ps   r&c@s^eZdZUded<ded<ded<dZded<dZded <d d Zd d ZddZddZ dS)r*r,pretrained_model_name descriptionlocationNz'Model'class_ List[str]aliasescCsN|jj}djdi|j}|jdurdjdd|i|j}|d|d}|S)Nz_pretrained_model_name={pretrained_model_name}, description={description}, location={location}z{extras}, class_={class_}extrasz( z )r4)rrjr__dict__r)rgbaserrepresentationr4r4r5__repr__s zPretrainedModelInfo.__repr__cCst|j}|Sr/)hashr)rg location_hashr4r4r5__hash__s zPretrainedModelInfo.__hash__cCst|t|kp |j|jkSr/)rrrgotherr4r4r5__eq__szPretrainedModelInfo.__eq__cCs |j|jkSr/rrr4r4r5__lt__s zPretrainedModelInfo.__lt__) rjrkrlrnrrrrrrr4r4r4r5r*s    r*c@s~eZdZdZeed+ddZed,ddZe d-d.ddZed/d0dd Z ed/d0d!d"Z d1d2d)d*Z d S)3r'z[ Abstract class offering interface which should be implemented by all NeMo models. r-#Optional[List[PretrainedModelInfo]]cCrp)aj Should list all pre-trained models available via NVIDIA NGC cloud. Note: There is no check that requires model names and aliases to be unique. In the case of a collision, whatever model (or alias) is listed first in the this returned list will be instantiated. Returns: A list of PretrainedModelInfo entries Nr4)rr4r4r5list_available_modelss zModel.list_available_modelsrcCs&g}|durdd|D}|S)z Returns the list of model names available via NVIDIA NGC cloud, to get the complete model description use list_available_models() Returns: A list of model names NcSsg|]}|jqSr4r)r1modelr4r4r5 r^z3Model.get_available_model_names..)r)r model_namesr4r4r5get_available_model_namess zModel.get_available_model_namesFNT model_namer, refresh_cacher.rrrrrrrrrrreturn_model_fileOptional[bool]c Csr|durt}d|vr|j||d\} } tj| r| |_n |j||d\} } | r+| S| j| ||||||d} | S)a Instantiates an instance of NeMo from NVIDIA NGC cloud Use restore_from() to instantiate from a local .nemo file. Args: model_name: string key which will be used to find the module. refresh_cache: If set to True, then when fetching from cloud, this will re-fetch the file from cloud even if it is already found in a cache locally. override_config_path: path to a yaml config that will override the internal config file map_location: Optional torch.device() to map the instantiated model to a device. By default (None), it will select a GPU if available, falling back to CPU otherwise. strict: Passed to torch.load_state_dict. By default true. return_config: If set to true, will return just the underlying config of the restored model as an OmegaConf DictConfig object without instantiating the model. return_model_file: If set to true, will return just the downloaded model file in cache Returns: A model instance of a particular model class or its underlying config (if return_config is set). N/)rr)rrrrrrr)r!_get_hf_hub_pretrained_model_infoospathisdirmodel_extracted_dir_get_ngc_pretrained_model_infor) rrrrrrrrrr rnemo_model_file_in_cacherr4r4r5from_pretraineds0    zModel.from_pretrainedTuple[type, str]cCsd}d}d}|}|dur?|D],}d}|j|krd}n|jdur1|jD] } | |kr0d}nq&|r>|j}|j}|j}nq|durKtd|d|dd} || d} t t | dd } t ||d } t| | | | |d }td |dur|}||fS) a Resolve the NGC model pretrained information given a model name. Assumes the model subclass implements the `list_available_models()` inherited method. Args: model_name: Str name of the model. Must be the original name or an alias of the model, without any '/'. refresh_cache: Bool, determines whether cache must be refreshed (model is re-downloaded). Returns: A tuple of details describing : - The resolved class of the model. This requires subclass to implement PretrainedModelInfo.class_. If the class cannot be resolved, default to the class that called this method. - The path to the NeMo model (.nemo file) in some cached directory. NFTzModel zW was not found. Check cls.list_available_models() for the list of all available models.r rr)urlfilename cache_dir subfolderrz/Instantiating model from pre-trained checkpoint)rrrrrrFileNotFoundErrorsplitreplacer joinpathr"hashlibmd5encode hexdigestr!r info)rrrlocation_in_the_cloudrrmodelspretrained_model_infofoundaliasrrrcache_subfolderrr4r4r5rAsH        z$Model._get_ngc_pretrained_model_infoc Cs|ddd}|st||d}|dur|tur||fSt}t|d}|j||dd}|r>t||d tj||d }||fS|j ||d d }t t d |} g} |j D]} | j} d| vscd| vri| | jqTt| } d| } t| d} tj| | } tj| rtt| }|dkrtd|| td|dt j!| d dtj| std|d| tj"| d dt#|d tj|| | d|d}||fS)a Resolve the HuggingFace Hub model pretrained information given a model name. The model name must be of general syntax ``{source_repo}/{model_name}``. Note: The ``{source_repo}`` need not be ``nvidia``, it can be any public repository, even external to Nvidia. This allows public, externally contributed models to be run freely using Nvidia NeMo. Args: model_name: Str name of the model. Must be the original name or an alias of the model, without any '/'. refresh_cache: Bool, determines whether cache must be refreshed (model is re-downloaded). Returns: A tuple of details describing : - The resolved class of the model. Since the source is external to NeMo, always default to using the calling class. Depend on target class resolution by restore_from() for calling the correct class. - The path to the NeMo model (.nemo file) in some cached directory (managed by HF Hub). r rz.nemo)repo_idrN)tokenr)r*r repo_typenemo)r*r library_namelibrary_versionforce_downloadr+T)r*r+files_metadata hf_hub_cachereadmegitrrrz$Found {} files in cache directory {}z(Deleting old cache directory for model `z#` in order to prevent duplicates...) ignore_errorsz Downloading z from HuggingFace Hub to path: )exist_okF)r*r.r/r0r local_dirlocal_dir_use_symlinksr+)$rrr get_hf_tokenr file_existsrr- __version__ repo_infor rr"siblings rfilenamelowerrblob_idsortedrrr r!r"r rexistsrdlistdirr r#rshutilrmtreemakedirsr)rrrresolved_model_filenamerhf_tokenapinemo_file_existsr<rr)siblingrrnum_files_in_dirr4r4r5r {sh  6        z'Model._get_hf_hub_pretrained_model_infohfrtemplatetemplate_kwargsOptional[Dict[str, str]]objectcCsD|dur tt}|duri}|dkr|j||dStd|d)z Generates a ModelCard for the current model. This method is called when pushing the model to the Hub. Returns: An object that can be represented as a str representation of the model card, usually in Markdown format. NrM)rNrOzModel card type z not supported.)copydeepcopyr_get_hf_model_cardr=)rgrrNrOr4r4r5generate_model_cards zModel.generate_model_card)r-r)r-r)FNNTFNNF)rr,rr.rrrrrr.rr.rrrrr r )F)rr,rr.r-r)rMNN)rr,rNr,rOrPr-rQ) rjrkrlrmrrrrrrr rUr4r4r4r5r's.  B9`r'c@seZdZdZGdddeZejejdfd"d d Zd d ZddZ e j e dd#ddZ ed$d%ddZeeddZed$d%ddZeeddZed$d%dd Zd!S)&r)a A decorator which performs input-output neural type checks, and attaches neural types to the output of the function that it wraps. Requires that the class inherit from :class:`~nemo.core.Typing` in order to perform type checking, and will raise an error if that is not the case. # Usage (Class level type support) .. code-block:: python @typecheck() def fn(self, arg1, arg2, ...): ... # Usage (Function level type support) .. code-block:: python @typecheck(input_types=..., output_types=...) def fn(self, arg1, arg2, ...): ... Points to be noted: 1) The brackets () in `@typecheck()` are necessary. You will encounter a TypeError: __init__() takes 1 positional argument but X were given without those brackets. 2) The function can take any number of positional arguments during definition. When you call this function, all arguments must be passed using kwargs only. c@seZdZdZdZdS)ztypecheck.TypeStatez Placeholder to denote the default value of type information provided. If the constructor of this decorator is used to override the class level type definition, this enum value indicate that types will be overridden. rN)rjrkrlrm UNINITIALIZEDr4r4r4r5 TypeStatesrWFrr'Union[TypeState, Dict[str, NeuralType]]rtrTr.cCsJ||_||_||jjkrd|_nd|_||jjkrd|_nd|_||_dS)NFT)rrrtrWrVinput_overrideoutput_overriderT)rgrrrtrTr4r4r5r!s   ztypecheck.__init__cCs ||Sr/) wrapped_callrgwrappedr4r4r5__call__6s ztypecheck.__call__cCs|Sr/r4r\r4r4r5unwrapped_call9sztypecheck.unwrapped_callenabledrr%cCs|durtdt|tstdt|dst|drtd|jtjjur*|j|_|jtjjur5|j|_|j r<|j}n|j}|j rF|j}n|j}|durX|durX||i|S|durft |dkrft d|j d ||jd|||i|}|j||j|d |S) a Wrapper method that can be used on any function of a class that implements :class:`~nemo.core.Typing`. By default, it will utilize the `input_types` and `output_types` properties of the class inheriting Typing. Local function level overrides can be provided by supplying dictionaries as arguments to the decorator. Args: input_types: Union[TypeState, Dict[str, NeuralType]]. By default, uses the global `input_types`. output_types: Union[TypeState, Dict[str, NeuralType]]. By default, uses the global `output_types`. ignore_collections: Bool. Determines if container types should be asserted for depth checks, or if depth checks are skipped entirely. NzDOnly classes which inherit nemo.core.Typing can use this decorator ! input_ports output_portsziTyping requires override of `input_types()` and `output_types()`, not `input_ports() and `output_ports()`rz=All arguments must be passed by kwargs only for typed methods)rrrT)rtrTrr4) RuntimeErrorr7r%rrrr)rWrVrtrYrZrdrerrTr)rgr]rrGrHrrrtoutputsr4r4r5r[<s8 ztypecheck.wrapped_callTracC|adS)z Global method to enable/disable typechecking. Args: enabled: bool, when True will enable typechecking. NrKr`r4r4r5set_typecheck_enabled} ztypecheck.set_typecheck_enabledc c6tjddz dVWtjdddStjddw)z] Context manager that temporarily disables type checking within its context. Fr`NT)r)rgr4r4r4r5disable_checks  ztypecheck.disable_checkscCrf)z Global method to enable/disable semantic typechecking. Args: enabled: bool, when True will enable semantic typechecking. NrOr`r4r4r5set_semantic_check_enabledrhz$typecheck.set_semantic_check_enabledc cri)zf Context manager that temporarily disables semantic type checking within its context. Fr`NT)r)rlr4r4r4r5disable_semantic_checksrkz!typecheck.disable_semantic_checkscCs6t||rtjjjjjt_dStjjjjjt_dSr/) r)rgr-coreclassescommonr[r^r_r`r4r4r5enable_wrappings ztypecheck.enable_wrappingN)rrrXrtrXrTr.)rr%)T)rar.)rjrkrlrmr rWrVrr^r_wrapt decoratorrMr[ staticmethodrgrrjrlrmrqr4r4r4r5r)s.$    @      r))r+r,r-r.)r;r )rCr)Srm __future__rrRrrr rDrabcrrcollections.abcrr contextlibr dataclassesrr enumr functoolsr pathlibr typingr rrrrrrDtorchrrhuggingface_hubrrrr9rrr omegaconfrrr-$nemo.core.classes.mixins.hf_io_mixinr%nemo.core.config.templates.model_cardr+nemo.core.connectors.save_restore_connectorrnemo.core.neural_typesrr nemo.utilsr nemo.utils.cloudr!nemo.utils.data_utilsr"nemo.utils.model_utilsr#r$__all__rLrPrr9r:r?rIrMrQrRr%r(r&r*r'r)r4r4r4r5sp                 V;bI$