# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. from __future__ import annotations import collections import copyreg import os import pickle import sys import threading import warnings from collections.abc import Iterable from typing import TYPE_CHECKING import numpy as np import paddle # deprecated module import from paddle import base from paddle.base import core from paddle.base.framework import ( EagerParamBase, Program, Variable, _create_tensor, _current_expected_place, _current_expected_place_, _dygraph_tracer, in_dygraph_mode, in_pir_mode, ) from .io_utils import ( _is_file_path, _is_memory_buffer, _legacy_static_save, _open_file_buffer, _pack_loaded_dict, _pickle_loads_mac, _unpack_saved_dict, ) if TYPE_CHECKING: from io import BytesIO from typing import Any, Literal, TypedDict from typing_extensions import NotRequired, Unpack from paddle import Tensor from paddle._typing import NestedStructure from paddle.nn.layer.layers import _StateDict class _EmptyDict(TypedDict): pass class _LoadOptions(TypedDict): model_filename: NotRequired[str] params_filename: NotRequired[str] keep_name_table: NotRequired[bool] return_numpy: NotRequired[bool] class _SaveOptions(TypedDict): use_binary_format: NotRequired[bool] pickle_protocol: NotRequired[Literal[2, 3, 4]] __all__ = [] async_save_queue = [] def clear_async_save_task_queue() -> None: ''' wait until all async save task to be done. ''' while len(async_save_queue) > 0: task = async_save_queue.pop() if task and task.is_alive(): task.join() def async_save( obj: object, path: str | BytesIO, protocol: Literal[2, 3, 4] = 4, sync_other_task: bool = False, **configs: Unpack[_EmptyDict], ) -> None: ''' async version of paddle.save. Note: currently only support dygraph mode. Note: any argument passed through configs will be overridden by default setting. Args: obj(Object) : The object to be saved. path(str|BytesIO) : The path/buffer of the object to be saved. If saved in the current directory, the input path string will be used as the file name. protocol(int, optional): The protocol version of pickle module must be greater than 1 and less than 5. Default: 4 sync_other_task(bool) : Determine whether to wait other async save task to be finished before this one be put in queue. **configs(dict, optional): compatible argument to paddle.save, but will be overridden by default setting. Examples: .. code-block:: python :name: code-example-1 import paddle emb = paddle.nn.Embedding(10, 10) layer_state_dict = emb.state_dict() # call paddle.async_save with the same style of paddle.save paddle.async_save(layer_state_dict, "emb.pdparams") for i in range(10): # do some calculations here # wait if any async_save task has not been done paddle.clear_async_task_queue() ''' if not in_dygraph_mode(): raise ValueError( "async_save currently is not supported in static mode." ) if len(configs) > 0: warnings.warn( "configs are not supported in async mode, will be overridden by default settings." ) # TODO: make this part async def move_state_dict_to_cpu(sd): for k, v in sd.items(): if isinstance(v, dict): move_state_dict_to_cpu(v) elif isinstance(v, core.eager.Tensor): sd[k] = v.pin_memory() if core.is_compiled_with_cuda() else v if isinstance(obj, dict): move_state_dict_to_cpu(obj) elif isinstance(obj, core.eager.Tensor): obj = obj.pin_memory() if core.is_compiled_with_cuda() else obj else: # other types are currently not supported raise TypeError( f"currently async_save does not support this type: {type(obj)}" ) if sync_other_task: clear_async_save_task_queue() t = threading.Thread(target=save, args=(obj, path, protocol)) t.start() async_save_queue.append(t) def _build_saved_state_dict(state_dict): save_dict = {} name_table = {} for key, value in state_dict.items(): if isinstance(value, (Variable, core.eager.Tensor)): if value.type == core.VarDesc.VarType.VOCAB: save_dict[key] = value.value().get_map_tensor() else: if not value.value().get_tensor()._is_initialized(): raise ValueError( "The saved tensor is not initialized. If you used group sharded, please use save_group_sharded_model." ) if ( value.is_dense() and value.place.is_custom_place() and core.is_compiled_with_custom_device('npu') ): value = paddle._C_ops.npu_identity(value, -1) save_dict[key] = np.array(value.cpu()) name_table[key] = value.name else: save_dict[key] = value save_dict["StructuredToParameterName@@"] = name_table return save_dict def _load_state_dict_from_save_inference_model(model_path, config): # 1. load program desc & construct _ProgramHolder # TODO(GGBond8488):From a long-term perspective, it is inappropriate for the framework to # rely on jit. It is necessary to migrate the dependency from jit to the framework in the future if in_pir_mode(): from paddle.jit.pir_translated_layer import ( _construct_params_and_buffers, _construct_program_holders, ) programs = _construct_program_holders(model_path, config.model_filename) else: from paddle.jit.translated_layer import ( _construct_params_and_buffers, _construct_program_holders, ) programs = _construct_program_holders(model_path, config.model_filename) # 2. load layer parameters & buffers with base.dygraph.guard(): persistable_var_dict = _construct_params_and_buffers( model_path, programs, config.params_filename ) # 3. construct state_dict load_param_dict = {} for var_name in persistable_var_dict: tmp_var = persistable_var_dict[var_name] if tmp_var.is_dense() and tmp_var.place.is_custom_place(): load_param_dict[var_name] = np.array( paddle._C_ops.npu_identity(tmp_var, -1).cpu() ) else: load_param_dict[var_name] = np.array(tmp_var.cpu()) # if *.info exists, we can recover structured_name var_info_filename = str(config.params_filename) + ".info" var_info_path = os.path.join(model_path, var_info_filename) if os.path.exists(var_info_path): with open(var_info_path, 'rb') as f: extra_var_info = pickle.load(f) structured_para_dict = {} for var_name in load_param_dict: structured_name = extra_var_info[var_name].get( 'structured_name', None ) assert structured_name is not None, ( f"Cannot find saved variable ({var_name})'s structured name in saved model." ) structured_para_dict[structured_name] = load_param_dict[ var_name ] load_param_dict = structured_para_dict return load_param_dict def _load_state_dict_from_save_params(model_path): # Try to load all the files in the directory in Tensor format, # the file name is used as the name of Tensor load_var_list = [] # 1. load file names var_name_list = [] for root, _, files in os.walk(model_path): for filename in files: file_path = os.path.join(root, filename) tmp_var_name = os.path.relpath(file_path, model_path) var_name = tmp_var_name.replace("\\", "/") var_name_list.append(var_name) # 2. create and load Tensor with base.dygraph.guard(): for name in var_name_list: new_var = _create_tensor(name=name, persistable=True) _dygraph_tracer().trace_op( type='load', inputs={}, outputs={'Out': new_var}, attrs={'file_path': os.path.join(model_path, name)}, ) load_var_list.append(new_var) # 3. construct state_dict load_param_dict = {} for var in load_var_list: if var.is_dense() and var.place.is_custom_place(): var = paddle._C_ops.npu_identity(var, -1) load_param_dict[var.name] = np.array(var.cpu()) return load_param_dict # NOTE(chenweihang): [ Handling of use cases of API paddle.load ] # `paddle.load` may be used to load saved results of: # 1. Expected cases: # - need [full filename] when loading # - paddle.save # - paddle.static.save # - need [prefix] when loading [compatible for paddle 2.x] # - paddle.jit.save # - paddle.static.save_inference_model # - need [directory] when loading [compatible for paddle 1.x] # - paddle.base.io.save_inference_model # - paddle.base.io.save_params/save_persistable # 2. Error cases: # - no error case def _build_load_path_and_config(path, config): # NOTE(chenweihang): If both [prefix save format] and [directory save format] exist, # raise error, avoid confusing behavior # TODO(GGBond8488):From a long-term perspective, it is inappropriate for the framework to # rely on jit. It is necessary to migrate the dependency from jit to the framework in the future from paddle.jit.pir_translated_layer import ( PIR_INFER_MODEL_SUFFIX, ) from paddle.jit.translated_layer import ( INFER_MODEL_SUFFIX, INFER_PARAMS_SUFFIX, ) if in_pir_mode(): prefix_format_path = path + PIR_INFER_MODEL_SUFFIX else: prefix_format_path = path + INFER_MODEL_SUFFIX prefix_format_exist = os.path.exists(prefix_format_path) directory_format_exist = os.path.isdir(path) if prefix_format_exist and directory_format_exist: raise ValueError( f"The {path}.pdmodel and {path} directory exist at the same time, " "don't know which one to load, please make sure that the specified target " "of ``path`` is unique." ) elif not prefix_format_exist and not directory_format_exist: error_msg = "The ``path`` (%s) to load model not exists." # if current path is a prefix, and the path.pdparams or path.pdopt # is exist, users may want use `paddle.load` load the result of # `base.save_dygraph`, we raise error here for users params_file_path = path + ".pdparams" opti_file_path = path + ".pdopt" if os.path.exists(params_file_path) or os.path.exists(opti_file_path): error_msg += ( "please specify the full file name, not just the file name prefix. For " "example, it should be written as `paddle.load('model.pdparams')` instead of " "`paddle.load('model')`." ) raise ValueError(error_msg % path) else: if prefix_format_exist: file_prefix = os.path.basename(path) model_path = os.path.dirname(path) if config.model_filename is not None: warnings.warn( "When loading the result saved with the " "specified file prefix, the ``model_filename`` config does " "not take effect." ) if in_pir_mode(): config.model_filename = file_prefix + PIR_INFER_MODEL_SUFFIX else: config.model_filename = file_prefix + INFER_MODEL_SUFFIX if config.params_filename is not None: warnings.warn( "When loading the result saved with the " "specified file prefix, the ``params_filename`` config does " "not take effect." ) config.params_filename = file_prefix + INFER_PARAMS_SUFFIX else: # Compatible with the old save_inference_model format model_path = path return model_path, config def _parse_load_config(configs): supported_configs = [ 'model_filename', 'params_filename', 'keep_name_table', 'return_numpy', 'safetensors', ] # input check for key in configs: if key not in supported_configs: raise ValueError( f"The additional config ({key}) of `paddle.load` is not supported." ) # construct inner config # TODO(GGBond8488):From a long-term perspective, it is inappropriate for the framework to # rely on jit. It is necessary to migrate the dependency from jit to the framework in the future from paddle.jit.api import _SaveLoadConfig inner_config = _SaveLoadConfig() inner_config.model_filename = configs.get('model_filename', None) inner_config.params_filename = configs.get('params_filename', None) inner_config.keep_name_table = configs.get('keep_name_table', None) inner_config.return_numpy = configs.get('return_numpy', False) inner_config.safetensors = configs.get('safetensors', False) return inner_config def _parse_save_config(configs): supported_configs = ['use_binary_format', 'pickle_protocol', 'safetensors'] # input check for key in configs: if key not in supported_configs: raise ValueError( f"The additional config ({key}) of `paddle.save` is not supported." ) # construct inner config # TODO(GGBond8488):From a long-term perspective, it is inappropriate for the framework to # rely on jit. It is necessary to migrate the dependency from jit to the framework in the future from paddle.jit.api import _SaveLoadConfig inner_config = _SaveLoadConfig() inner_config.use_binary_format = configs.get('use_binary_format', False) inner_config.pickle_protocol = configs.get('pickle_protocol', None) inner_config.safetensors = configs.get('safetensors', False) return inner_config def _pickle_save(obj, f, protocol): # TODO(weixin):add support for BytesIO. if not isinstance(protocol, int): raise ValueError( f"The 'protocol' MUST be `int`, but received {type(protocol)}" ) if protocol < 2 or protocol > 4: raise ValueError( f"Expected 1<'protocol'<5, but received protocol={protocol}" ) def reduce_varbase(self): if self.is_dense() and self.place.is_custom_place(): data = np.array(paddle._C_ops.npu_identity(self, -1).cpu()) else: data = np.array(self.cpu()) name = self.name return (tuple, ((name, data),)) def reduce_DenseTensor(self): p = core.Place() p.set_place(paddle.CPUPlace()) if self._place().is_custom_place(): data = np.array(paddle._C_ops.npu_identity(self, -1)._copy(p)) else: data = np.array(self._copy(p)) return (eval, ('data', {'data': data})) def reduce_Layer(self): raise ValueError( "paddle do not support saving `paddle.nn.Layer` object." ) dispatch_table_layer = {} def create_layer_dispatch_table(layer): dispatch_table_layer[layer.__class__] = reduce_Layer return layer _parse_every_object( obj, lambda v: isinstance(v, paddle.nn.Layer), create_layer_dispatch_table, ) def add_dispatch_table(): # This is not a good method, because the pickle module has been modified. pickle.dispatch_table[core.eager.Tensor] = reduce_varbase pickle.dispatch_table[EagerParamBase] = reduce_varbase pickle.dispatch_table[core.DenseTensor] = reduce_DenseTensor pickle.dispatch_table.update(dispatch_table_layer) def pop_dispatch_table(): pickle.dispatch_table.pop(core.DenseTensor) pickle.dispatch_table.pop(core.eager.Tensor) pickle.dispatch_table.pop(EagerParamBase) for k in dispatch_table_layer: pickle.dispatch_table.pop(k) # When value of dict is lager than 4GB ,there is a Bug on 'MAC python3' if sys.platform == 'darwin' and sys.version_info.major == 3: add_dispatch_table() pickle_bytes = pickle.dumps(obj) pop_dispatch_table() max_bytes = 2**30 for i in range(0, len(pickle_bytes), max_bytes): f.write(pickle_bytes[i : i + max_bytes]) else: pickler = pickle.Pickler(f, protocol) pickler.dispatch_table = copyreg.dispatch_table.copy() pickler.dispatch_table[core.DenseTensor] = reduce_DenseTensor pickler.dispatch_table[core.eager.Tensor] = reduce_varbase pickler.dispatch_table[EagerParamBase] = reduce_varbase pickler.dispatch_table.update(dispatch_table_layer) pickler.dump(obj) def _contain_x(obj, condition_func): if isinstance(obj, core.SelectedRows): raise NotImplementedError( "`paddle.save` do not support saving 'SelectedRows'." ) if condition_func(obj): return True elif type(obj) in (dict, collections.OrderedDict, list, tuple): if type(obj) in (dict, collections.OrderedDict): keys = list(obj.keys()) else: keys = range(len(obj)) flag = False for key in keys: flag |= _contain_x(obj[key], condition_func) if flag: return True return flag else: return False def _is_state_dict(obj): if isinstance(obj, dict): def condition(obj): return isinstance( obj, ( paddle.nn.Layer, Program, core.eager.Tensor, core.DenseTensor, core.SelectedRows, ), ) # If the value of a dict is a core.Tensor/DenseTensor or a dict # that does not contain a paddle type(Layer, Program, Tensor, DenseTensor, SelectedRows), # the dict is considered to be a state_ dict. for key, value in obj.items(): if isinstance(value, dict): for k, v in value.items(): if _contain_x(v, condition): return False elif not isinstance(value, (core.eager.Tensor, core.DenseTensor)): return False return True return False def _transformed_from_varbase(obj): # In paddle2.1 version, Tensor is saved as tuple(tensor.name, tensor.numpy()). # When executing paddle.load, use this function to determine whether to restore to Tensor. if isinstance(obj, tuple) and len(obj) == 2: name_types = str if isinstance(obj[0], name_types) and isinstance(obj[1], np.ndarray): return True return False def _transformed_from_lodtensor(obj): # In paddle2.1 version, DenseTensor is saved as np.array(tensor). # When executing paddle.load, use this function to determine whether to restore to Tensor. if isinstance(obj, np.ndarray): return True return False def _to_LodTensor(ndarray): if not isinstance(ndarray, np.ndarray): raise TypeError( f'Type of `ndarray` should be numpy.ndarray, but received {type(ndarray)}.' ) t = core.DenseTensor() place = _current_expected_place_() t.set(ndarray, place) return t def _tuple_to_tensor(obj, return_numpy): if return_numpy: return obj[1] if in_dygraph_mode(): t = paddle.to_tensor(obj[1]) # This function does modify the name of return value. # Loading the same variable multiple times may cause the same name. t.name = obj[0] return t else: return _to_LodTensor(obj[1]) def _ndarray_to_tensor(obj, return_numpy): if return_numpy: return obj if in_dygraph_mode(): return paddle.to_tensor(obj) else: return _to_LodTensor(obj) def _lod_tensor2varbase(tensor): return_var = _create_tensor() return_var.value().get_tensor().set(tensor, _current_expected_place()) return return_var def _parse_every_object(obj, condition_func, convert_func): if condition_func(obj): return convert_func(obj) elif type(obj) in (dict, collections.OrderedDict, list): if type(obj) == list: keys = range(len(obj)) else: keys = list(obj.keys()) for key in keys: if condition_func(obj[key]): obj[key] = convert_func(obj[key]) else: obj[key] = _parse_every_object( obj[key], condition_func, convert_func ) return obj elif type(obj) == tuple: return tuple( _parse_every_object(list(obj), condition_func, convert_func) ) elif type(obj) == set: return set(_parse_every_object(list(obj), condition_func, convert_func)) else: if isinstance(obj, Iterable) and not isinstance( obj, (str, np.ndarray, core.eager.Tensor, core.DenseTensor), ): raise NotImplementedError( f"The iterable objects supported are tuple, list, dict, OrderedDict, string. But received {type(obj)}." ) return obj def _parse_load_result(obj, return_numpy): def is_layer(obj): return isinstance(obj, paddle.nn.Layer) def parse_layer(obj): temp_dict = _parse_load_result(obj.__dict__, False) obj.__dict__.update(temp_dict) return obj if _contain_x(obj, is_layer): if not in_dygraph_mode(): raise ValueError( "Layer can only be loaded in dynamic graph mode, but now in static graph mode." ) _parse_every_object(obj, is_layer, parse_layer) def tuple_to_tensor(obj): return _tuple_to_tensor(obj, return_numpy=return_numpy) def ndarray_to_tensor(obj): return _ndarray_to_tensor(obj, return_numpy=return_numpy) # tuple(name, ndarray) was converted from varbase of paddle2.1, # and all tuple(name, ndarray) are converted to tensor. if _contain_x(obj, _transformed_from_varbase): return _parse_every_object( obj, _transformed_from_varbase, tuple_to_tensor ) # If there is no tuple(name, ndarray), it is considered to be saved by paddle2.0 # or converted from DenseTensor, and all ndarrays are converted to tensor. else: return _parse_every_object( obj, _transformed_from_lodtensor, ndarray_to_tensor ) def _save_dense_tensor(tensor, file_name): if not tensor._is_initialized(): raise ValueError( "The saved tensor is not initialized. If you used group sharded, please use save_group_sharded_model firstly." ) if _is_file_path(file_name): _seek = core.save_dense_tensor(tensor, file_name) # '_seek' is the end position of this tensor in the file. elif _is_memory_buffer(file_name): tensor_bytes = core.save_dense_tensor_to_memory(tensor) with _open_file_buffer(file_name, 'wb') as f: f.write(tensor_bytes) _seek = f.tell() else: raise NotImplementedError( f'Only supports saving objects to file or BytesIO, but received {type(file_name)}' ) return _seek def _load_dense_tensor(file_name): temp_t = paddle.base.core.DenseTensor() if _is_file_path(file_name): # '_seek' is the end position of this tensor in the file. _seek = paddle.base.core.load_dense_tensor(temp_t, file_name) elif _is_memory_buffer(file_name): with _open_file_buffer(file_name, 'rb') as f: tensor_bytes = f.read() paddle.base.core.load_dense_tensor_from_memory(temp_t, tensor_bytes) _seek = f.tell() else: raise NotImplementedError( f'Only supports load objects from file or BytesIO, but received {type(file_name)}' ) return temp_t, _seek def _save_selected_rows(selected_rows, file_name): if not selected_rows.get_tensor()._is_initialized(): raise ValueError("The saved tensor is not initialized.") if _is_file_path(file_name): # '_seek' is the end position of this SelectedRows in the file. _seek = core.save_selected_rows(selected_rows, file_name) elif _is_memory_buffer(file_name): selected_rows_bytes = core.save_selected_rows_to_memory(selected_rows) with _open_file_buffer(file_name, 'wb') as f: f.write(selected_rows_bytes) _seek = f.tell() else: raise NotImplementedError( f'Only supports saving objects to file or BytesIO, but received {type(file_name)}' ) return _seek def _load_selected_rows(file_name): temp_sr = core.SelectedRows() if _is_file_path(file_name): # '_seek' is the end position of this SelectedRows in the file. _seek = core.load_selected_rows(temp_sr, file_name) elif _is_memory_buffer(file_name): with _open_file_buffer(file_name, 'rb') as f: selected_rows_bytes = f.read() paddle.base.core.load_selected_rows_from_memory( temp_sr, selected_rows_bytes ) _seek = f.tell() else: raise NotImplementedError( f'Only supports load objects from file or BytesIO, but received {type(file_name)}' ) return temp_sr, _seek def _save_binary_var(obj, path): if isinstance(obj, core.DenseTensor): _save_dense_tensor(obj, path) elif isinstance(obj, core.SelectedRows): _save_selected_rows(obj, path) elif isinstance(obj, core.eager.Tensor): _save_dense_tensor(obj.value().get_tensor(), path) else: # Since the concept of 'Tensor' is only exposed to users, the error message can only contain tensor instead of 'DenseTensor' or 'SelectedRows' raise NotImplementedError( f"When use_binary_format = True, `paddle.save` expected Tensor, but received {type(obj)}." ) def save( obj: _StateDict | NestedStructure[Tensor] | Program, path: str | BytesIO, protocol: Literal[2, 3, 4] = 4, **configs: Unpack[_SaveOptions], ) -> None: ''' Save an object to the specified path. Note: Now supports saving ``state_dict`` of Layer/Optimizer, Tensor and nested structure containing Tensor, Program. Note: Different from ``paddle.jit.save``, since the save result of ``paddle.save`` is a single file, there is no need to distinguish multiple saved files by adding a suffix. The argument ``path`` of ``paddle.save`` will be directly used as the saved file name instead of a prefix. In order to unify the saved file name format, we recommend using the paddle standard suffix: 1. for ``Layer.state_dict`` , recommend to use ``.pdparams`` ; 2. for ``Optimizer.state_dict`` , recommend to use ``.pdopt`` . For specific examples, please refer to API code examples. Args: obj(Object) : The object to be saved. path(str|BytesIO) : The path/buffer of the object to be saved. If saved in the current directory, the input path string will be used as the file name. protocol(int, optional): The protocol version of pickle module must be greater than 1 and less than 5. Default: 4 **configs(dict, optional): optional keyword arguments. The following options are currently supported: use_binary_format(bool): When the saved object is static graph variable, you can specify ``use_binary_for_var``. If True, save the file in the c++ binary format when saving a single static graph variable; otherwise, save it in pickle format. Default: False Returns: None Examples: .. code-block:: python :name: code-example-1 >>> # example 1: dynamic graph >>> import paddle >>> emb = paddle.nn.Embedding(10, 10) >>> layer_state_dict = emb.state_dict() >>> # save state_dict of emb >>> paddle.save(layer_state_dict, "emb.pdparams") >>> scheduler = paddle.optimizer.lr.NoamDecay( ... d_model=100, warmup_steps=100, verbose=True) >>> adam = paddle.optimizer.Adam( ... learning_rate=scheduler, ... parameters=emb.parameters()) >>> opt_state_dict = adam.state_dict() >>> # save state_dict of optimizer >>> paddle.save(opt_state_dict, "adam.pdopt") >>> # save weight of emb >>> paddle.save(emb.weight, "emb.weight.pdtensor") .. code-block:: python :name: code-example-2 >>> # example 2: Save multiple state_dict at the same time >>> import paddle >>> from paddle import nn >>> from paddle.optimizer import Adam >>> layer = paddle.nn.Linear(3, 4) >>> adam = Adam(learning_rate=0.001, parameters=layer.parameters()) >>> obj = {'model': layer.state_dict(), 'opt': adam.state_dict(), 'epoch': 100} >>> path = 'example/model.pdparams' >>> paddle.save(obj, path) .. code-block:: python :name: code-example-3 >>> # example 3: static graph >>> import paddle >>> import paddle.static as static >>> paddle.enable_static() >>> # create network >>> x = paddle.static.data(name="x", shape=[None, 224], dtype='float32') >>> z = paddle.static.nn.fc(x, 10) >>> place = paddle.CPUPlace() >>> exe = paddle.static.Executor(place) >>> exe.run(paddle.static.default_startup_program()) >>> prog = paddle.static.default_main_program() >>> for var in prog.list_vars(): ... if list(var.shape) == [224, 10]: ... tensor = var.get_value() ... break >>> # save/load tensor >>> path_tensor = 'temp/tensor.pdtensor' >>> paddle.save(tensor, path_tensor) >>> # save/load state_dict >>> path_state_dict = 'temp/model.pdparams' >>> paddle.save(prog.state_dict("param"), path_tensor) .. code-block:: python :name: code-example-4 >>> # example 4: save program >>> import paddle >>> paddle.enable_static() >>> data = paddle.static.data( ... name='x_static_save', shape=(None, 224), dtype='float32') >>> y_static = z = paddle.static.nn.fc(data, 10) >>> main_program = paddle.static.default_main_program() >>> path = "example/main_program.pdmodel" >>> paddle.save(main_program, path) .. code-block:: python :name: code-example-5 >>> # example 5: save object to memory >>> from io import BytesIO >>> import paddle >>> from paddle.nn import Linear >>> paddle.disable_static() >>> linear = Linear(5, 10) >>> state_dict = linear.state_dict() >>> byio = BytesIO() >>> paddle.save(state_dict, byio) >>> paddle.seed(2023) >>> tensor = paddle.randn([2, 3], dtype='float32') >>> paddle.save(tensor, byio) ''' if _is_file_path(path): # 1. input check filename = os.path.basename(path) if filename == "": raise ValueError( "The input path MUST be format of dirname/filename " "[dirname\\filename in Windows system], but received " "filename is empty string." ) # 2. save object dirname = os.path.dirname(path) if dirname and not os.path.exists(dirname): os.makedirs(dirname, exist_ok=True) elif not _is_memory_buffer(path): raise ValueError( f"only supports saving objects to file and `BytesIO`, but got {type(path)}" ) config = _parse_save_config(configs) if not isinstance(config.use_binary_format, bool): raise TypeError( f"Type of `use_binary_format` should be bool, but received {type(config.use_binary_format)}." ) if config.use_binary_format: _save_binary_var(obj, path) else: # `protocol` need to be used, `pickle_protocol` is a deprecated arg. if config.pickle_protocol is not None: protocol = config.pickle_protocol warnings.warn( "'pickle_protocol' is a deprecated argument. Please use 'protocol' instead." ) if isinstance(obj, paddle.static.Program): if in_pir_mode(): paddle.core.serialize_pir_program(obj, path) else: obj.desc.flush() with _open_file_buffer(path, "wb") as f: f.write(obj.desc.serialize_to_string()) elif _is_state_dict(obj): if in_dygraph_mode(): if config.safetensors: _safe_save(obj, path) else: _legacy_save(obj, path, protocol) else: _legacy_static_save(obj, path, protocol) else: with _open_file_buffer(path, 'wb') as f: _pickle_save(obj, f, protocol) def _safe_save(obj, path): if not isinstance(obj, dict): raise NotImplementedError( "Now only supports save state_dict of Layer or Optimizer, " f"expect dict, but received {type(obj)}." ) if len(obj) == 0: warnings.warn("The input state dict is empty, no need to save.") if _is_file_path(path): filename = os.path.basename(path) if filename == "": raise ValueError( "The input path MUST be format of dirname/filename " "[dirname\\filename in Windows system], but received " "filename is empty string." ) # 2. save object dirname = os.path.dirname(path) if dirname and not os.path.exists(dirname): os.makedirs(dirname, exist_ok=True) from safetensors.paddle import save_file save_file(obj, path) def _legacy_save(obj, path, protocol=2): # 1. input check if not isinstance(obj, dict): raise NotImplementedError( "Now only supports save state_dict of Layer or Optimizer, " f"expect dict, but received {type(obj)}." ) if len(obj) == 0: warnings.warn("The input state dict is empty, no need to save.") if not isinstance(protocol, int): raise ValueError( f"The 'protocol' MUST be `int`, but received {type(protocol)}" ) if protocol < 2 or protocol > 4: raise ValueError( f"Expected 1<'protocol'<5, but received protocol={protocol}" ) if _is_file_path(path): filename = os.path.basename(path) if filename == "": raise ValueError( "The input path MUST be format of dirname/filename " "[dirname\\filename in Windows system], but received " "filename is empty string." ) # 2. save object dirname = os.path.dirname(path) if dirname and not os.path.exists(dirname): os.makedirs(dirname, exist_ok=True) if isinstance(obj, dict): saved_obj = _build_saved_state_dict(obj) saved_obj = _unpack_saved_dict(saved_obj, protocol) # When value of dict is lager than 4GB ,there is a Bug on 'MAC python3' if ( _is_file_path(path) and sys.platform == 'darwin' and sys.version_info.major == 3 ): pickle_bytes = pickle.dumps(saved_obj, protocol=protocol) with open(path, 'wb') as f: max_bytes = 2**30 f.writelines( pickle_bytes[i : i + max_bytes] for i in range(0, len(pickle_bytes), max_bytes) ) else: with _open_file_buffer(path, 'wb') as f: pickle.dump(saved_obj, f, protocol=protocol) def load(path: str | BytesIO, **configs: Unpack[_LoadOptions]) -> Any: ''' Load an object can be used in paddle from specified path. Note: Now supports loading ``state_dict`` of Layer/Optimizer, Tensor and nested structure containing Tensor, Program. Note: In order to use the model parameters saved by paddle more efficiently, ``paddle.load`` supports loading ``state_dict`` of Layer from the result of other save APIs except ``paddle.save`` , but the argument ``path`` format is different: 1. loading from ``paddle.static.save`` or ``paddle.Model().save(training=True)`` , ``path`` needs to be a complete file name, such as ``model.pdparams`` or ``model.pdopt`` ; 2. loading from ``paddle.jit.save`` or ``paddle.static.save_inference_model`` or ``paddle.Model().save(training=False)`` , ``path`` need to be a file prefix, such as ``model/mnist``, and ``paddle.load`` will get information from ``mnist.pdmodel`` and ``mnist.pdiparams`` ; 3. loading from paddle 1.x APIs ``paddle.base.io.save_inference_model`` or ``paddle.base.io.save_params/save_persistables`` , ``path`` need to be a directory, such as ``model`` and model is a directory. Note: If you load ``state_dict`` from the saved result of static graph mode API such as ``paddle.static.save`` or ``paddle.static.save_inference_model`` , the structured variable name in dynamic mode will cannot be restored. You need to set the argument ``use_structured_name=False`` when using ``Layer.set_state_dict`` later. Args: path(str|BytesIO) : The path/buffer to load the target object. Generally, the path is the target file path. When loading state_dict from the saved result of the API used to save the inference model, the path may be a file prefix or directory. **configs (dict, optional): other load configuration options for compatibility. We do not recommend using these configurations, they may be removed in the future. If not necessary, DO NOT use them. Default None. The following options are currently supported: (1) model_filename (str): The inference model file name of the paddle 1.x ``save_inference_model`` save format. Default file name is :code:`__model__` . (2) params_filename (str): The persistable variables file name of the paddle 1.x ``save_inference_model`` save format. No default file name, save variables separately by default. (3) return_numpy(bool): If specified as True, return tensor as numpy.ndarray, otherwise return tensor as paddle.Tensor. Default False. Returns: Object(Object): a target object can be used in paddle Examples: .. code-block:: python :name: code-example-1 >>> # example 1: dynamic graph >>> import paddle >>> emb = paddle.nn.Embedding(10, 10) >>> layer_state_dict = emb.state_dict() >>> # save state_dict of emb >>> paddle.save(layer_state_dict, "emb.pdparams") >>> scheduler = paddle.optimizer.lr.NoamDecay( ... d_model=100, warmup_steps=100, verbose=True) >>> adam = paddle.optimizer.Adam( ... learning_rate=scheduler, ... parameters=emb.parameters()) >>> opt_state_dict = adam.state_dict() >>> # save state_dict of optimizer >>> paddle.save(opt_state_dict, "adam.pdopt") >>> # save weight of emb >>> paddle.save(emb.weight, "emb.weight.pdtensor") >>> # load state_dict of emb >>> load_layer_state_dict = paddle.load("emb.pdparams") >>> # load state_dict of optimizer >>> load_opt_state_dict = paddle.load("adam.pdopt") >>> # load weight of emb >>> load_weight = paddle.load("emb.weight.pdtensor") .. code-block:: python :name: code-example-2 >>> # example 2: Load multiple state_dict at the same time >>> import paddle >>> from paddle import nn >>> from paddle.optimizer import Adam >>> layer = paddle.nn.Linear(3, 4) >>> adam = Adam(learning_rate=0.001, parameters=layer.parameters()) >>> obj = {'model': layer.state_dict(), 'opt': adam.state_dict(), 'epoch': 100} >>> path = 'example/model.pdparams' >>> paddle.save(obj, path) >>> obj_load = paddle.load(path) .. code-block:: python :name: code-example-3 >>> # example 3: static graph >>> import paddle >>> import paddle.static as static >>> paddle.enable_static() >>> # create network >>> x = paddle.static.data(name="x", shape=[None, 224], dtype='float32') >>> z = paddle.static.nn.fc(x, 10) >>> place = paddle.CPUPlace() >>> exe = paddle.static.Executor(place) >>> exe.run(paddle.static.default_startup_program()) >>> prog = paddle.static.default_main_program() >>> for var in prog.list_vars(): ... if list(var.shape) == [224, 10]: ... tensor = var.get_value() ... break >>> # save/load tensor >>> path_tensor = 'temp/tensor.pdtensor' >>> paddle.save(tensor, path_tensor) >>> load_tensor = paddle.load(path_tensor) >>> # save/load state_dict >>> path_state_dict = 'temp/model.pdparams' >>> paddle.save(prog.state_dict("param"), path_tensor) >>> load_state_dict = paddle.load(path_tensor) .. code-block:: python :name: code-example-4 >>> # example 4: load program >>> import paddle >>> paddle.enable_static() >>> data = paddle.static.data( ... name='x_static_save', shape=(None, 224), dtype='float32') >>> y_static = z = paddle.static.nn.fc(data, 10) >>> main_program = paddle.static.default_main_program() >>> path = "example/main_program.pdmodel" >>> paddle.save(main_program, path) >>> load_main = paddle.load(path) .. code-block:: python :name: code-example-5 >>> # example 5: save object to memory >>> from io import BytesIO >>> import paddle >>> from paddle.nn import Linear >>> paddle.disable_static() >>> linear = Linear(5, 10) >>> state_dict = linear.state_dict() >>> byio = BytesIO() >>> paddle.save(state_dict, byio) >>> paddle.seed(2023) >>> tensor = paddle.randn([2, 3], dtype='float32') >>> paddle.save(tensor, byio) >>> byio.seek(0) >>> # load state_dict >>> dict_load = paddle.load(byio) ''' if _is_memory_buffer(path) or os.path.isfile(path): config = _parse_load_config(configs) exception_type = pickle.UnpicklingError try: if config.safetensors: if config.return_numpy: from safetensors.numpy import load_file load_result = load_file(path) load_result = _pack_loaded_dict(load_result) else: import safetensors from safetensors.paddle import load_file if isinstance(_current_expected_place(), core.CUDAPlace): if ( safetensors.__version__ > "0.6.2" and paddle.__version__ >= "3.2.0" ): # NOTE(Ruibiao): load_file may cause segmentation fault in some case. f = safetensors.safe_open(path, framework="paddle") load_result = {} for k in f.keys(): load_result[k] = f.get_tensor(k).cuda() else: load_result = load_file( path, device=_current_expected_place() ) else: load_result = load_file(path, device='cpu') return load_result with _open_file_buffer(path, 'rb') as f: # When value of dict is lager than 4GB ,there is a Bug on 'MAC python3' if ( _is_file_path(path) and sys.platform == 'darwin' and sys.version_info.major == 3 ): load_result = _pickle_loads_mac(path, f) else: load_result = pickle.load(f, encoding='latin1') # TODO(weixin):If `obj` is any object, the judgment condition should be more precise. if isinstance(load_result, dict): load_result = _pack_loaded_dict(load_result) # paddle2.0: paddle.save/load if "StructuredToParameterName@@" in load_result: for key, name in load_result[ "StructuredToParameterName@@" ].items(): if isinstance(load_result[key], np.ndarray): load_result[key] = _ndarray_to_tensor( load_result[key], config.return_numpy ) # default name is "generatedxxx" which is set in Tensor init, if not set if not config.return_numpy and getattr( load_result[key], "name", "" ): load_result[key].name = name if ( not config.keep_name_table and "StructuredToParameterName@@" in load_result ): del load_result["StructuredToParameterName@@"] else: # paddle2.1 static.save/load load_result = _parse_load_result( load_result, config.return_numpy ) else: load_result = _parse_load_result( load_result, config.return_numpy ) except exception_type as msg_pickle: try: tensor, _ = _load_selected_rows(path) return tensor except: try: tensor, _ = _load_dense_tensor(path) if config.return_numpy: p = core.Place() p.set_place(paddle.CPUPlace()) if tensor._place().is_custom_place(): return np.array( paddle._C_ops.npu_identity(tensor, -1)._copy(p) ) else: return np.array(tensor._copy(p)) else: if in_dygraph_mode(): return _lod_tensor2varbase(tensor) return tensor except: try: if in_pir_mode(): program = paddle.static.Program() paddle.core.deserialize_pir_program(path, program) return program with _open_file_buffer(path, "rb") as f: program_desc_str = f.read() program = Program.parse_from_string( program_desc_str ) if paddle.framework.in_pir_executor_mode(): with paddle.pir_utils.IrGuard(): program = paddle.pir.translate_to_pir( program.desc ) block = program.global_block() remove_op_list = [] for op in block.ops: if op.name() == "pd_op.feed": var_name = op.attrs()["name"] org_value = op.result(0) with block: value = paddle.static.data( name=var_name, shape=org_value.shape, dtype=org_value.dtype, ) org_value.replace_all_uses_with( value ) value.get_defining_op().move_before( op ) remove_op_list.append(op) for op in remove_op_list: block.remove_op(op) return program except: raise ValueError( f"`paddle.load` can not parse the file:{path}." ) else: load_result = _legacy_load(path, **configs) return load_result def _legacy_load(path, **configs): load_result = None config = _parse_load_config(configs) if os.path.isfile(path) or _is_memory_buffer(path): # we think path is file means this file is created by paddle.save if config.safetensors: from safetensors.paddle import load_file load_result = load_file(path) else: with _open_file_buffer(path, 'rb') as f: load_result = pickle.load(f, encoding='latin1') load_result = _pack_loaded_dict(load_result) if ( not config.keep_name_table and "StructuredToParameterName@@" in load_result ): del load_result["StructuredToParameterName@@"] else: # file prefix and directory are compatible cases model_path, config = _build_load_path_and_config(path, config) # check whether model file exists if config.model_filename is None: model_filename = '__model__' else: model_filename = config.model_filename model_file_path = os.path.join(model_path, model_filename) if os.path.exists(model_file_path): # Load state dict by `jit.save/io.save_inference_model` save format # NOTE(chenweihang): [ Compatibility of save_inference_model save format ] # The model saved by `save_inference_model` does not completely correspond to # the information required by the `state_dict` under the dygraph. # `save_inference_model` not save structured name, we need to remind # the user to configure the `use_structured_name` argument when `set_state_dict` # NOTE(chenweihang): `jit.save` doesn't save optimizer state load_result = _load_state_dict_from_save_inference_model( model_path, config ) else: # load state dict by `io.save_params/persistables` save format # TODO(chenweihang): [ Now only supports loading parameters separately ] # If users save all parameters as one file, the [ variable.name -> variable ] # mapping info will lost, so users need to give variable list, but users build # variable list in dygraph mode is difficult, we recommend users to use # paddle.static.load_program_state in this case load_result = _load_state_dict_from_save_params(model_path) return load_result