Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka

Qalabka Overview Qalabka Tensor API Shuruudaha ugu weyn ee API Shuruudaha ugu horeysay ee loo yaqaan Unary elementwise APIs Qalabka dhismaha ah ee APIs binary Shuruudaha Batchable BatchableExtensionType tusaale: Network TensorFlow APIs oo loo isticmaalaa ExtensionTypes Qalabka Qalabka Flow Control Qalabka Autograph Control Flow Haku Qalabka Haku Qalabka Tensor API Shuruudaha dhismaha waxaa laga yaabaa "tensor-like", in la mid ah in ay ku dhismaha ama si ay u dhismaha interfaces defined by the Tensor-ka-ka-ka-ka-ka-ka-ka-ka-ka-ka Haku iyo Haku TensorFlow waxaa loo isticmaali karaa si ay u isticmaali karaa si ay u isticmaali karaa macluumaadka macluumaadka TensorFlow. TensorFlow waxaa loo isticmaali karaa 3 macluumaadka macluumaadka: tf.Tensor RaggedTensor SparseTensor MaskedTensor Shirkadda Decorator Shirkadda Decorator @tf.experimental.dispatch_for_api(tf_api) @tf.experimental.dispatch_for_unary_elementwise_apis(x_type) @tf.experimental.dispatch_for_binary_elementwise_apis(x_type, y_type) Shuruudaha ugu weyn ee API EE.UU Taageerada waxaa laga yaabaa in ay ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah. Process waaweyn Haku: tf.experimental.dispatch_for_api tf.stack MaskedTensor @tf.experimental.dispatch_for_api(tf.stack) def masked_stack(values: List[MaskedTensor], axis = 0): return MaskedTensor(tf.stack([v.values for v in values], axis), tf.stack([v.mask for v in values], axis)) Sida loo yaqaan "default implementation" ee Sida loo yaqaan 'Lista' waxaa loo yaqaan 'Lista Qalabka (ka dib markii la Qalabka waxaa la aasaasay ee: tf.stack MaskedTensor values typing.List[MaskedTensor] x = MaskedTensor([1, 2, 3], [True, True, False]) y = MaskedTensor([4, 5, 6], [False, True, True]) tf.stack([x, y]) Sida loo yaabaa Sida loo isticmaali karaa lists of mixed iyo Waqtiga, sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale sidoo kale Parameter iyo dhismaha badan of function adeegga ah: tf.stack MaskedTensor Tensor values tf.experimental.unregister_dispatch_for(masked_stack) def convert_to_masked_tensor(x): if isinstance(x, MaskedTensor): return x else: return MaskedTensor(x, tf.ones_like(x, tf.bool)) @tf.experimental.dispatch_for_api(tf.stack) def masked_stack_v2(values: List[Union[MaskedTensor, tf.Tensor]], axis = 0): values = [convert_to_masked_tensor(v) for v in values] return MaskedTensor(tf.stack([v.values for v in values], axis), tf.stack([v.mask for v in values], axis)) x = MaskedTensor([1, 2, 3], [True, True, False]) y = tf.constant([4, 5, 6]) tf.stack([x, y, x]) Sida loo isticmaali karaa, waxaa loo isticmaali karaa API-ka. . tf.experimental.dispatch_for_api Shuruudaha ugu horeysay ee loo yaqaan Unary elementwise APIs EE.UU Shuruudaha ugu horeysay ee loo yaqaan 'Defense' waa in la soo xiriir. Unary elementwise ops (ka mid ka mid ah ) oo ka mid ah qiimaha ugu horeysay ee argumentiga (waxariis ku saabsan c) Dhammaan oo ka mid ah adiga ah Function Decorated waxaa loo isticmaali karaa 2 argumento: tf.experimental.dispatch_for_unary_elementwise_apis Haku tf.math.cos x x_type Haku api_func: Functiisa oo ka mid ah parameter ka mid ah wax soo saarka elementwise (ey, tf.abs). x: Argumentka ugu horeysay ee operation elementwise. Qalabka dhismaha ugu horeysay waa in la soo bandhigay. Haku: MaskedTensor @tf.experimental.dispatch_for_unary_elementwise_apis(MaskedTensor) def masked_tensor_unary_elementwise_api_handler(api_func, x): return MaskedTensor(api_func(x.values), x.mask) Waxaa la isticmaali karaa xawaaraha ugu horeysay ee loo isticmaali karaa in ay u isticmaali karaa in la isticmaali karaa. . MaskedTensor x = MaskedTensor([1, -2, -3], [True, False, True]) print(tf.abs(x)) print(tf.ones_like(x, dtype=tf.float32)) Qalabka dhismaha ah ee APIs binary Sidaas, Waxaa la isticmaali karaa si ay u adeegsanayo dhammaan operations binary elementwise si ay u isticmaali karaa Haku: tf.experimental.dispatch_for_binary_elementwise_apis MaskedTensor @tf.experimental.dispatch_for_binary_elementwise_apis(MaskedTensor, MaskedTensor) def masked_tensor_binary_elementwise_api_handler(api_func, x, y): return MaskedTensor(api_func(x.values, y.values), x.mask & y.mask) x = MaskedTensor([1, -2, -3], [True, False, True]) y = MaskedTensor([[4], [5]], [[True], [False]]) tf.math.add(x, y) Sida loo isticmaali karaa, waxaa loo isticmaali karaa API-ka ee loo isticmaali karaa. iyo . tf.experimental.dispatch_for_unary_elementwise_apis tf.experimental.dispatch_for_binary_elementwise_apis Haku s Qalabka Qalabka Haku Haku Sida loo isticmaali karaa mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ah mid ka mid ah. S. TensorFlow API-ga waxaa loo isticmaali karaa in la isticmaali karaa in la isticmaali karaa in la isticmaali karaa in la isticmaali karaa in la isticmaali karaa in la isticmaali karaa in la isticmaali karaa in la isticmaali karaa: ExtensionType Haku Tensor Tf.data.Dataset (batch, unbatch, from_tensor_slices) Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Haku: Haku Sida loo helo, Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Haku: S, iyo If this is not suitable for your class then you will need to use Sida loo isticmaali karaa, waxaa loo isticmaali karaa si ay u isticmaali karaa si loo isticmaali karaa. Waayo, waxaa loo yaqaan "Sparse Tensors" Haku iyo macluumaadka - in badan oo dhan, waxaad ka mid ahaysaa macluumaadka ka mid ah, sida ay ka mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ah mid ka mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ka mid ah. . BatchableExtensionType Tensor CompositeTensor ExtensionType tf.experimental.ExtensionTypeBatchEncoder tf.SparseTensor values indices dense_shape SparseTensor Markaad: BatchableExtensionTypes ma ahaatee ma ahaatee, tf.stack, tf.concat, tf.slice, iwm. Markaad: BatchableExtensionTypes ma ahaatee ma ahaatee, tf.stack, tf.concat, tf.slice, iwm. Haku: BatchableExtensionType Network Si kastaba ha ahaatee, sidoo kale loo isticmaali karaa Qalabka waxaa loo isticmaali karaa si ay u isticmaali karaa wax soo saarka in ay u isticmaali karaa in ay u isticmaali karaa in ay u isticmaali karaa in ay u isticmaali karaa in ay isticmaali karaa in ay isticmaali karaa in ay isticmaali karaa in ay isticmaali karaa. Network class Network(tf.experimental.ExtensionType): # This version is not batchable. work: tf.Tensor # work[n] = work left to do at node n bandwidth: tf.Tensor # bandwidth[n1, n2] = bandwidth from n1->n2 net1 = Network([5., 3, 8], [[0., 2, 0], [2, 0, 3], [0, 3, 0]]) net2 = Network([3., 4, 2], [[0., 2, 2], [2, 0, 2], [2, 2, 0]]) Sida loo isticmaali karaa, waxaa loo isticmaali karaa si ay u isticmaali karaa Shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha. Shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha ugu horeysay ee shuruudaha. Qalabka waxaa laga yaabaa in la Marka Sidaa waxa uu Qalabka . BatchableExtensionType shape shape tf.data.Dataset tf.map_fn Haku tf.keras class Network(tf.experimental.BatchableExtensionType): shape: tf.TensorShape # batch shape. A single network has shape=[]. work: tf.Tensor # work[*shape, n] = work left to do at node n bandwidth: tf.Tensor # bandwidth[*shape, n1, n2] = bandwidth from n1->n2 def __init__(self, work, bandwidth): self.work = tf.convert_to_tensor(work) self.bandwidth = tf.convert_to_tensor(bandwidth) work_batch_shape = self.work.shape[:-1] bandwidth_batch_shape = self.bandwidth.shape[:-2] self.shape = work_batch_shape.merge_with(bandwidth_batch_shape) def __repr__(self): return network_repr(self) def network_repr(network): work = network.work bandwidth = network.bandwidth if hasattr(work, 'numpy'): work = ' '.join(str(work.numpy()).split()) if hasattr(bandwidth, 'numpy'): bandwidth = ' '.join(str(bandwidth.numpy()).split()) return (f" ") net1 = Network([5., 3, 8], [[0., 2, 0], [2, 0, 3], [0, 3, 0]]) net2 = Network([3., 4, 2], [[0., 2, 2], [2, 0, 2], [2, 2, 0]]) batch_of_networks = Network( work=tf.stack([net1.work, net2.work]), bandwidth=tf.stack([net1.bandwidth, net2.bandwidth])) print(f"net1={net1}") print(f"net2={net2}") print(f"batch={batch_of_networks}") Waxaad u isticmaali karaa Sida loo isticmaali karaa in ka mid ah network: tf.data.Dataset dataset = tf.data.Dataset.from_tensor_slices(batch_of_networks) for i, network in enumerate(dataset): print(f"Batch element {i}: {network}") Waxaad sidoo kale isticmaali karaa Waayo, waxaa loo isticmaali karaa function for each batch element: map_fn def balance_work_greedy(network): delta = (tf.expand_dims(network.work, -1) - tf.expand_dims(network.work, -2)) delta /= 4 delta = tf.maximum(tf.minimum(delta, network.bandwidth), -network.bandwidth) new_work = network.work + tf.reduce_sum(delta, -1) return Network(new_work, network.bandwidth) tf.map_fn(balance_work_greedy, batch_of_networks) TensorFlow API-ga oo ku saabsan s Qalabka Qalabka Qalabka is a decorator which precomputes TensorFlow graphs for Python functions, which can substantially improve the performance of your TensorFlow code. Functiibada waaweyn. tf.function @tf.function class Pastry(tf.experimental.ExtensionType): sweetness: tf.Tensor # 2d embedding that encodes sweetness chewiness: tf.Tensor # 2d embedding that encodes chewiness @tf.function def combine_pastry_features(x: Pastry): return (x.sweetness + x.chewiness) / 2 cookie = Pastry(sweetness=[1.2, 0.4], chewiness=[0.8, 0.2]) combine_pastry_features(cookie) Marka aad u baahan yahay in ay ku yaalaa Haku Sida loo isticmaali karaa extension-ka . input_signature tf.function TypeSpec pastry_spec = Pastry.Spec(tf.TensorSpec([2]), tf.TensorSpec(2)) @tf.function(input_signature=[pastry_spec]) def increase_sweetness(x: Pastry, delta=1.0): return Pastry(x.sweetness + delta, x.chewiness) increase_sweetness(cookie) Functiyada Shuruudaha dhismaha waaweyn ee shuruudaha waaweyn ee shuruudaha waaweyn ee shuruudaha. Qalabka dhismaha waxaa loo isticmaali karaa in la isticmaali karaa in la isticmaali karaa macluumaadka. tf.function cf = combine_pastry_features.get_concrete_function(pastry_spec) cf(cookie) Qalabka Flow Control Shuruudaha dhismaha waxaa loo isticmaali karaa by TensorFlow's control-flow operations: QEEBE Qalabka Haku: Haku Qalabka Identity # Example: using tf.cond to select between two MaskedTensors. Note that the # two MaskedTensors don't need to have the same shape. a = MaskedTensor([1., 2, 3], [True, False, True]) b = MaskedTensor([22., 33, 108, 55], [True, True, True, False]) condition = tf.constant(True) print(tf.cond(condition, lambda: a, lambda: b)) # Example: using tf.while_loop with MaskedTensor. cond = lambda i, _: i < 10 def body(i, mt): return i + 1, mt.with_values(mt.values + 3 / 7) print(tf.while_loop(cond, body, [0, b])[1]) Qalabka Autograph Control Flow Shuruudaha dhismaha waxaa loo isticmaali karaa oo ka mid ah shuruudaha nidaamka (wax yar) Markaad ka mid ah mid ka mid ah mid ka mid ah Xafiiska iyo Qalabka waxaa loo isticmaali karaa automatically in iyo operations, oo loo isticmaali karaa qalabka extension. tf.function if for tf.cond tf.while_loop @tf.function def fn(x, b): if b: x = MaskedTensor(x, tf.less(x, 0)) else: x = MaskedTensor(x, tf.greater(x, 0)) for i in tf.range(5 if b else 7): x = x.with_values(x.values + 1 / 2) return x print(fn(tf.constant([1., -2, 3]), tf.constant(True))) print(fn(tf.constant([1., -2, 3]), tf.constant(False))) Haku TensorFlow waa API-ka sare ee TensorFlow si ay u baabuurta iyo si ay u baabuurta dhismaha dhismaha dhismaha. Qalabka dhismaha waxaa laga yaabaa sida dhismaha in model Keras, dhismaha mid ka mid ah layers Keras, iyo dhismaha dhismaha Keras. QEEBE Waayo, waxaa laga yaqaan "Batchable ExtensionTypes" ah. Waayo, waxaa lagu soo bandhigay in la soo bandhigay in la soo bandhigay. shape[0] waa in la soo bandhigay batch. Tusaale badan oo ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah. Haku ka mid ah: Network Haku Qalabka ugu horeysay, sidoo kale ka mid ah Class defined in la "Batchable" s" aasaasay ugu horeysay, oo loo isticmaali karaa in la shaqeeyaan load balancing ka mid ah nodes. Network ExtensionType class Network(tf.experimental.BatchableExtensionType): shape: tf.TensorShape # batch shape. A single network has shape=[]. work: tf.Tensor # work[*shape, n] = work left to do at node n bandwidth: tf.Tensor # bandwidth[*shape, n1, n2] = bandwidth from n1->n2 def __init__(self, work, bandwidth): self.work = tf.convert_to_tensor(work) self.bandwidth = tf.convert_to_tensor(bandwidth) work_batch_shape = self.work.shape[:-1] bandwidth_batch_shape = self.bandwidth.shape[:-2] self.shape = work_batch_shape.merge_with(bandwidth_batch_shape) def __repr__(self): return network_repr(self) single_network = Network( # A single network with 4 nodes. work=[8.0, 5, 12, 2], bandwidth=[[0.0, 1, 2, 2], [1, 0, 0, 2], [2, 0, 0, 1], [2, 2, 1, 0]]) batch_of_networks = Network( # Batch of 2 networks, each w/ 2 nodes. work=[[8.0, 5], [3, 2]], bandwidth=[[[0.0, 1], [1, 0]], [[0, 2], [2, 0]]]) Waxaad ka mid ah ka mid ah wax soo saarka ah oo ka mid ah wax soo saarka. Haku Network class BalanceNetworkLayer(tf.keras.layers.Layer): """Layer that balances work between nodes in a network. Shifts work from more busy nodes to less busy nodes, constrained by bandwidth. """ def call(self, inputs): # This function is defined above in the "Batchable `ExtensionType`s" section. return balance_work_greedy(inputs) Waxaad ka mid ah u isticmaali karaa mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ah mid ka mid ah mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah. Sida loo isticmaali karaa, sidoo kale loo isticmaali karaa a Qalabka Nala soo saarka extension ee Haddii loo isticmaali karaa modelka Keras ee loo isticmaali karaa batch-ka, waxaan Waxa uu ku saabsan batch size. ExtensionType tf.keras.layer.Input type_spec TypeSpec type_spec input_spec = Network.Spec(shape=None, work=tf.TensorSpec(None, tf.float32), bandwidth=tf.TensorSpec(None, tf.float32)) model = tf.keras.Sequential([ tf.keras.layers.Input(type_spec=input_spec), BalanceNetworkLayer(), ]) Waxaad ka mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah. model(single_network) model(batch_of_networks) Qalabka: MaskedTensor Sida loo yaabaa, Waxaa laga yaqaan support Haku Waxaa laga yaabaa in ay ka mid ah wax soo saarka ah oo ka mid ah wax soo saarka. Keras waxay u baahan tahay in ay ka soo xiriir in la soo xiriir oo ka mid ah loo yaqaan extension Haku Sida loo helo a Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka MaskedTensor Keras shape values TypeSpec MaskedTensor __name__ SavedModel class MaskedTensor(tf.experimental.BatchableExtensionType): # __name__ is required for serialization in SavedModel; see below for details. __name__ = 'extension_type_colab.MaskedTensor' values: tf.Tensor mask: tf.Tensor shape = property(lambda self: self.values.shape) dtype = property(lambda self: self.values.dtype) def with_default(self, default): return tf.where(self.mask, self.values, default) def __repr__(self): return masked_tensor_str(self.values, self.mask) class Spec: def __init__(self, shape, dtype=tf.float32): self.values = tf.TensorSpec(shape, dtype) self.mask = tf.TensorSpec(shape, tf.bool) shape = property(lambda self: self.values.shape) dtype = property(lambda self: self.values.dtype) def with_shape(self): return MaskedTensor.Spec(tf.TensorSpec(shape, self.values.dtype), tf.TensorSpec(shape, self.mask.dtype)) Sidaas, decorators dispatch waxaa loo isticmaali karaa si ay u isticmaali karaa wax soo saarka default of several TensorFlow APIs. Since these APIs are used by standard Keras layers (s. Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Sida loo isticmaali karaa in aad u isticmaali karaa. for masked tensors waxaa loo isticmaali karaa in la isticmaali karaa qiimaha masked sida zero (eyna, in la isticmaali karaa in wax soo saarka). Dense MaskedTensor matmul @tf.experimental.dispatch_for_unary_elementwise_apis(MaskedTensor) def unary_elementwise_op_handler(op, x): return MaskedTensor(op(x.values), x.mask) @tf.experimental.dispatch_for_binary_elementwise_apis( Union[MaskedTensor, tf.Tensor], Union[MaskedTensor, tf.Tensor]) def binary_elementwise_op_handler(op, x, y): x = convert_to_masked_tensor(x) y = convert_to_masked_tensor(y) return MaskedTensor(op(x.values, y.values), x.mask & y.mask) @tf.experimental.dispatch_for_api(tf.matmul) def masked_matmul(a: MaskedTensor, b, transpose_a=False, transpose_b=False, adjoint_a=False, adjoint_b=False, a_is_sparse=False, b_is_sparse=False, output_type=None): if isinstance(a, MaskedTensor): a = a.with_default(0) if isinstance(b, MaskedTensor): b = b.with_default(0) return tf.matmul(a, b, transpose_a, transpose_b, adjoint_a, adjoint_b, a_is_sparse, b_is_sparse, output_type) Waxaad ka heli karaa modelka Keras oo ku saabsan Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka Qalabka MaskedTensor input_spec = MaskedTensor.Spec([None, 2], tf.float32) masked_tensor_model = tf.keras.Sequential([ tf.keras.layers.Input(type_spec=input_spec), tf.keras.layers.Dense(16, activation="relu"), tf.keras.layers.Dense(1)]) masked_tensor_model.compile(loss='binary_crossentropy', optimizer='rmsprop') a = MaskedTensor([[1., 2], [3, 4], [5, 6]], [[True, False], [False, True], [True, True]]) masked_tensor_model.fit(a, tf.constant([[1], [0], [1]]), epochs=3) print(masked_tensor_model(a)) Qalabka a) Qalabka TensorFlow waa shuruudaha seryada ah, oo ku yaalaa sida wax soo saarka iyo wax soo saarka. Waxaa laga yaqaan Keras model ama model custom. Qalabka SavedModel waxay ku salaysan models, layers, iyo functions oo loo isticmaali karaa qalabka qalabka qalabka qalabka qalabka qalabka qalabka. Nala soo bandhigiisa waxaa loo isticmaali karaa in ay u isticmaali karaa in ay u isticmaali karaa in ay u isticmaali karaa in ay isticmaali karaa. __name__ Example: si ay u socdaan model Keras Keras models that use extension types may be stored by . SavedModel masked_tensor_model_path = tempfile.mkdtemp() tf.saved_model.save(masked_tensor_model, masked_tensor_model_path) imported_model = tf.saved_model.load(masked_tensor_model_path) imported_model(a) Example: si ay u soo saarka model Saxiiciga waxaa loo isticmaali karaa in ay ku salaysan subclasses oo ka mid ah macluumaadka oo loo isticmaali karaa macluumaadka. tf.Module class CustomModule(tf.Module): def __init__(self, variable_value): super().__init__() self.v = tf.Variable(variable_value) @tf.function def grow(self, x: MaskedTensor): """Increase values in `x` by multiplying them by `self.v`.""" return MaskedTensor(x.values * self.v, x.mask) module = CustomModule(100.0) module.grow.get_concrete_function(MaskedTensor.Spec(shape=None, dtype=tf.float32)) custom_module_path = tempfile.mkdtemp() tf.saved_model.save(module, custom_module_path) imported_model = tf.saved_model.load(custom_module_path) imported_model.grow(MaskedTensor([1., 2, 3], [False, True, False])) Shuruudaha ugu soo saarka ah ee loo yaqaan 'SavedModel' Waxaa laga yaabaa Qalabka Qalabka Sida loo soo saarka a Sida loo isticmaali karaa Sidaa waxa uu waxaa laga yaabaa in ay ku yaalaa (eyna waa in ay ku yaalaa in ay ku yaalaa in ay ku yaalaa in ay ku yaalaa TensorFlow si ay u isticmaalaa "anonymous extension type" object. This object will have the same fields as the original type, but will lack any further customization you have added for the type, such as custom methods or properties. SavedModel ExtensionType ExtensionType Haku Qalabka TensorFlow Qalabka Qalabka hadda na soo xiriir, (aga macaamiisha kale ee "SavedModel" "signatures" dictionary) waxaa loo isticmaalaa in ay ka mid ah dhismaha iyo dhismaha waa tensors raw. If you want to use TensorFlow Serving with a model that uses extension types, then you can add wrapper methods that compose or decompose extension type values from tensors. TensorFlow waxaa loo isticmaalo class CustomModuleWrapper(tf.Module): def __init__(self, variable_value): super().__init__() self.v = tf.Variable(variable_value) @tf.function def var_weighted_mean(self, x: MaskedTensor): """Mean value of unmasked values in x, weighted by self.v.""" x = MaskedTensor(x.values * self.v, x.mask) return (tf.reduce_sum(x.with_default(0)) / tf.reduce_sum(tf.cast(x.mask, x.dtype))) @tf.function() def var_weighted_mean_wrapper(self, x_values, x_mask): """Raw tensor wrapper for var_weighted_mean.""" return self.var_weighted_mean(MaskedTensor(x_values, x_mask)) module = CustomModuleWrapper([3., 2., 8., 5.]) module.var_weighted_mean_wrapper.get_concrete_function( tf.TensorSpec(None, tf.float32), tf.TensorSpec(None, tf.bool)) custom_module_path = tempfile.mkdtemp() tf.saved_model.save(module, custom_module_path) imported_model = tf.saved_model.load(custom_module_path) x = MaskedTensor([1., 2., 3., 4.], [False, True, False, True]) imported_model.var_weighted_mean_wrapper(x.values, x.mask) s Dataset API waxaa loo isticmaali karaa in ay u isticmaali karaa in ay u isticmaali karaa in ay u isticmaali karaa in ay isticmaali karaa in ay isticmaali karaa. , oo ku yaalaa soo saarka elementada, oo ka mid ah cadaadiga ah ayaa ku yaalaa mid ka mid ah mid ka mid ah mid ka mid ah mid ka mid ah. tf.data tf.data.Dataset Qalabka S-ka oo ku saabsan extension-ka Dataset QEEBE Qalabka dhismaha waxaa loo isticmaali karaa oo ku saabsan dhismaha dhismaha Haku iyo : Dataset.from_tensors Dataset.from_tensor_slices Dataset.from_generator ds = tf.data.Dataset.from_tensors(Pastry(5, 5)) iter(ds).next() mt = MaskedTensor(tf.reshape(range(20), [5, 4]), tf.ones([5, 4])) ds = tf.data.Dataset.from_tensor_slices(mt) for value in ds: print(value) def value_gen(): for i in range(2, 7): yield MaskedTensor(range(10), [j%i != 0 for j in range(10)]) ds = tf.data.Dataset.from_generator( value_gen, output_signature=MaskedTensor.Spec(shape=[10], dtype=tf.int32)) for value in ds: print(value) Qalabka iyo Unbatching S-ka oo ku saabsan extension-ka QEEBE QEEBE Dhismaha dhismaha iyo dhismaha dhismaha iyo dhismaha dhismaha iyo . Dataset.batch Dataset.unbatch batched_ds = ds.batch(2) for value in batched_ds: print(value) unbatched_ds = batched_ds.unbatch() for value in unbatched_ds: print(value) Markaas oo ku yaalaa TensorFlow website, this article appears here under a new headline and is licensed under CC BY 4.0. code samples shared under the Apache 2.0 License. Markaas oo ku yaalaa TensorFlow website, this article appears here under a new headline and is licensed under CC BY 4.0. code samples shared under the Apache 2.0 License. Qalabka Qalabka