Improvements in TensorFlow Type Promotion Reduce Bit-Widening Risks

Content Overview Effects of turning on the new type promotion More consistent and predictable promotion type Reduced risk of bit-widening tf.Tensor mathematical dunder methods tf.Variable in-place ops tf.constant implicit conversions TF-NumPay Array Input Type Inference WeakTensor-supporting APIs Effects of turning on the new type promotion More consistent and predictable promotion type Reduced risk of bit-widening tf.Tensor mathematical dunder methods tf.Variable in-place ops tf.constant implicit conversions TF-NumPay Array Input Type Inference WeakTensor-supporting APIs Below is a non-exhaustive list of changes that result from turning on the new type promotion. More consistent and predictable promotion results. Reduced risk of bit-widening. tf.Tensor mathematical dunder methods use new type promotion. tf.constant can return WeakTensor. tf.constant allows implicit conversions when a Tensor input with a dtype different from the dtype arg is passed in. tf.Variable in-place ops (assign, assign-add, assign-sub) allow implicit conversions. tnp.array(1) and tnp.array(1.0) returns 32-bit WeakTensor. WeakTensors will be created and used for WeakTensor-supporting unary and binary API's. More consistent and predictable promotion results. Reduced risk of bit-widening. tf.Tensor mathematical dunder methods use new type promotion. tf.Tensor tf.constant can return WeakTensor. tf.constant WeakTensor tf.constant allows implicit conversions when a Tensor input with a dtype different from the dtype arg is passed in. tf.constant dtype tf.Variable in-place ops (assign, assign-add, assign-sub) allow implicit conversions. tf.Variable assign assign-add assign-sub tnp.array(1) and tnp.array(1.0) returns 32-bit WeakTensor. tnp.array(1) tnp.array(1.0) WeakTensors will be created and used for WeakTensor-supporting unary and binary API's. WeakTensor WeakTensor-supporting unary and binary API More consistent and predictable promotion results Using a lattice-based system allows the new type promotion to produce consistent and predictable type promotion results. lattice-based system Old Type Promotion Changing the order of operations produces inconsistent results using old type promotion. # Setup tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="legacy") a = np.array(1, dtype=np.int8) b = tf.constant(1) c = np.array(1, dtype=np.float16) # Setup tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="legacy") a = np.array(1, dtype=np.int8) b = tf.constant(1) c = np.array(1, dtype=np.float16) # (a + b) + c throws an InvalidArgumentError. try: tf.add(tf.add(a, b), c) except tf.errors.InvalidArgumentError as e: print(f'{type(e)}: {e}') # InvalidArgumentError # (a + b) + c throws an InvalidArgumentError. try: tf.add(tf.add(a, b), c) except tf.errors.InvalidArgumentError as e: print(f'{type(e)}: {e}') # InvalidArgumentError : cannot compute AddV2 as input #1(zero-based) was expected to be a int8 tensor but is a int32 tensor [Op:AddV2] name: : cannot compute AddV2 as input #1(zero-based) was expected to be a int8 tensor but is a int32 tensor [Op:AddV2] name: # (b + a) + c returns an i32 result. tf.add(tf.add(b, a), c) # # (b + a) + c returns an i32 result. tf.add(tf.add(b, a), c) # New Type Promotion New type promotion produces consistent results regardless of the order. tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") a = np.array(1, dtype=np.int8) b = tf.constant(1) c = np.array(1, dtype=np.float16) tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") a = np.array(1, dtype=np.int8) b = tf.constant(1) c = np.array(1, dtype=np.float16) WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. # (a + b) + c returns a f16 result. tf.add(tf.add(a, b), c) # # (a + b) + c returns a f16 result. tf.add(tf.add(a, b), c) # # (b + a) + c also returns a f16 result. tf.add(tf.add(b, a), c) # # (b + a) + c also returns a f16 result. tf.add(tf.add(b, a), c) # Reduced risk of bit-widening Old Type Promotion Old type promotion often resulted in 64-bit results. tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="legacy") tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="legacy") np.array(3.2, np.float16) + tf.constant(1, tf.int8) + tf.constant(50) # np.array(3.2, np.float16) + tf.constant(1, tf.int8) + tf.constant(50) # New Type Promotion New type promotion returns results with minimal number of bits necessary. tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. np.array(3.2, np.float16) + tf.constant(1, tf.int8) + tf.constant(50) # np.array(3.2, np.float16) + tf.constant(1, tf.int8) + tf.constant(50) # tf.Tensor mathematical dunder methods All tf.Tensor mathematical dunder methods will follow the new type promotion. tf.Tensor -tf.constant(5) # -tf.constant(5) # tf.constant(5, tf.int16) - tf.constant(1, tf.float32) # tf.constant(5, tf.int16) - tf.constant(1, tf.float32) # tf.Variable in-place ops Implicit conversions will be allowed in tf.Variable in-place ops. tf.Variable Note: Any promotion that results in a dtype that is different from the variable's original dtype will be not allowed. This is because tf.Variable cannot change its dtype. Note: Any promotion that results in a dtype that is different from the variable's original dtype will be not allowed. This is because tf.Variable cannot change its dtype. Note: tf.Variable tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") a = tf.Variable(10, tf.int32) a.assign_add(tf.constant(5, tf.int16)) # tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") a = tf.Variable(10, tf.int32) a.assign_add(tf.constant(5, tf.int16)) # WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. tf.constant implicit conversions In the old type promotion, tf.constant required an input Tensor to have the same dtype as the dtype argument. However, in the new type promotion, we implicitly convert Tensor to the specified dtype. tf.constant tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") a = tf.constant(10, tf.int16) tf.constant(a, tf.float32) # tnp.experimental_enable_numpy_behavior(dtype_conversion_mode="all") a = tf.constant(10, tf.int16) tf.constant(a, tf.float32) # WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. WARNING:tensorflow:UserWarning: enabling the new type promotion must happen at the beginning of the program. Please ensure no TF APIs have been used yet. TF-NumPy Array tnp.array defaults to i32* and f32* for python inputs using the new type promotion. tnp.array i32* f32* tnp.array(1) # tnp.array(1) # tnp.array(1.0) # tnp.array(1.0) # Input Type Inference This is how different inputs' types are inferred in the new type promotion. tf.Tensor: Since tf.Tensor has a dtype property, we don't do further inference. NumPy types: This includes types like np.array(1), np.int16(1), and np.float. Since NumPy inputs also have a dtype property, we take the dtype property as the result inference type. Note that NumPy defaults to i64 and f64. Python scalars/Nested types: This includes types like 1, [1, 2, 3], and (1.0, 2.0). Python int is inferred as i32*. Python float is inferred as f32*. Python complex is inferred as c128*. If the input doesn't fall into any of the above categories but has a dtype property, we take the dtype property as the result inference type. tf.Tensor: Since tf.Tensor has a dtype property, we don't do further inference. tf.Tensor tf.Tensor NumPy types: This includes types like np.array(1), np.int16(1), and np.float. Since NumPy inputs also have a dtype property, we take the dtype property as the result inference type. Note that NumPy defaults to i64 and f64. np.array(1) np.int16(1) np.float i64 f64 Python scalars/Nested types: This includes types like 1, [1, 2, 3], and (1.0, 2.0). Python int is inferred as i32*. Python float is inferred as f32*. Python complex is inferred as c128*. 1 [1, 2, 3] (1.0, 2.0) Python int is inferred as i32*. Python float is inferred as f32*. Python complex is inferred as c128*. Python int is inferred as i32*. int i32* Python float is inferred as f32*. float f32* Python complex is inferred as c128*. complex c128* If the input doesn't fall into any of the above categories but has a dtype property, we take the dtype property as the result inference type. Further Reading The new type promotion closely resembles JAX-NumPy's type promotion. If you want to know more details about the new type promotion and the design choices, check out the resources below. JAX Type Promotion Semantics Design of Type Promotion Semantics for JAX Old TF-NumPy Promotion Semantics JAX Type Promotion Semantics JAX Type Promotion Semantics Design of Type Promotion Semantics for JAX Design of Type Promotion Semantics for JAX Old TF-NumPy Promotion Semantics Old TF-NumPy Promotion Semantics WeakTensor-supporting APIs Below is a list of APIs that supports WeakTensor. WeakTensor For an unary op, this means that if an input with no user-specified type is passed in, it will return a WeakTensor. WeakTensor For a binary op, it will follow the promotion table here. It may or may not return a WeakTensor depending on the promotion result of the two inputs. here WeakTensor Note: All mathematical operations (+, -, *, ...) are supported. Note: All mathematical operations (+, -, *, ...) are supported. Note: + - * tf.bitwise.invert tf.clip_by_value tf.debugging.check_numerics tf.expand_dims tf.identity tf.image.adjust_brightness tf.image.adjust_gamma tf.image.extract_patches tf.image.random_brightness tf.image.stateless_random_brightness tf.linalg.diag tf.linalg.diag_part tf.linalg.matmul tf.linalg.matrix_transpose tf.linalg.tensor_diag_part tf.linalg.trace tf.math.abs tf.math.acos tf.math.acosh tf.math.add tf.math.angle tf.math.asin tf.math.asinh tf.math.atan tf.math.atanh tf.math.ceil tf.math.conj tf.math.cos tf.math.cosh tf.math.digamma tf.math.divide_no_nan tf.math.divide tf.math.erf tf.math.erfc tf.math.erfcinv tf.math.erfinv tf.math.exp tf.math.expm1 tf.math.floor tf.math.floordiv tf.math.floormod tf.math.imag tf.math.lgamma tf.math.log1p tf.math.log_sigmoid tf.math.log tf.math.multiply_no_nan tf.math.multiply tf.math.ndtri tf.math.negative tf.math.pow tf.math.real tf.math.real tf.math.reciprocal_no_nan tf.math.reciprocal tf.math.reduce_euclidean_norm tf.math.reduce_logsumexp tf.math.reduce_max tf.math.reduce_mean tf.math.reduce_min tf.math.reduce_prod tf.math.reduce_std tf.math.reduce_sum tf.math.reduce_variance tf.math.rint tf.math.round tf.math.rsqrt tf.math.scalar_mul tf.math.sigmoid tf.math.sign tf.math.sin tf.math.sinh tf.math.softplus tf.math.special.bessel_i0 tf.math.special.bessel_i0e tf.math.special.bessel_i1 tf.math.special.bessel_i1e tf.math.special.bessel_j0 tf.math.special.bessel_j1 tf.math.special.bessel_k0 tf.math.special.bessel_k0e tf.math.special.bessel_k1 tf.math.special.bessel_k1e tf.math.special.bessel_y0 tf.math.special.bessel_y1 tf.math.special.dawsn tf.math.special.expint tf.math.special.fresnel_cos tf.math.special.fresnel_sin tf.math.special.spence tf.math.sqrt tf.math.square tf.math.subtract tf.math.tan tf.math.tanh tf.nn.depth_to_space tf.nn.elu tf.nn.gelu tf.nn.leaky_relu tf.nn.log_softmax tf.nn.relu6 tf.nn.relu tf.nn.selu tf.nn.softsign tf.nn.space_to_depth tf.nn.swish tf.ones_like tf.realdiv tf.reshape tf.squeeze tf.stop_gradient tf.transpose tf.truncatediv tf.truncatemod tf.zeros_like tf.experimental.numpy.abs tf.experimental.numpy.absolute tf.experimental.numpy.amax tf.experimental.numpy.amin tf.experimental.numpy.angle tf.experimental.numpy.arange tf.experimental.numpy.arccos tf.experimental.numpy.arccosh tf.experimental.numpy.arcsin tf.experimental.numpy.arcsinh tf.experimental.numpy.arctan tf.experimental.numpy.arctanh tf.experimental.numpy.around tf.experimental.numpy.array tf.experimental.numpy.asanyarray tf.experimental.numpy.asarray tf.experimental.numpy.ascontiguousarray tf.experimental.numpy.average tf.experimental.numpy.bitwise_not tf.experimental.numpy.cbrt tf.experimental.numpy.ceil tf.experimental.numpy.conj tf.experimental.numpy.conjugate tf.experimental.numpy.copy tf.experimental.numpy.cos tf.experimental.numpy.cosh tf.experimental.numpy.cumprod tf.experimental.numpy.cumsum tf.experimental.numpy.deg2rad tf.experimental.numpy.diag tf.experimental.numpy.diagflat tf.experimental.numpy.diagonal tf.experimental.numpy.diff tf.experimental.numpy.empty_like tf.experimental.numpy.exp2 tf.experimental.numpy.exp tf.experimental.numpy.expand_dims tf.experimental.numpy.expm1 tf.experimental.numpy.fabs tf.experimental.numpy.fix tf.experimental.numpy.flatten tf.experimental.numpy.flip tf.experimental.numpy.fliplr tf.experimental.numpy.flipud tf.experimental.numpy.floor tf.experimental.numpy.full_like tf.experimental.numpy.imag tf.experimental.numpy.log10 tf.experimental.numpy.log1p tf.experimental.numpy.log2 tf.experimental.numpy.log tf.experimental.numpy.max tf.experimental.numpy.mean tf.experimental.numpy.min tf.experimental.numpy.moveaxis tf.experimental.numpy.nanmean tf.experimental.numpy.negative tf.experimental.numpy.ones_like tf.experimental.numpy.positive tf.experimental.numpy.prod tf.experimental.numpy.rad2deg tf.experimental.numpy.ravel tf.experimental.numpy.real tf.experimental.numpy.reciprocal tf.experimental.numpy.repeat tf.experimental.numpy.reshape tf.experimental.numpy.rot90 tf.experimental.numpy.round tf.experimental.numpy.signbit tf.experimental.numpy.sin tf.experimental.numpy.sinc tf.experimental.numpy.sinh tf.experimental.numpy.sort tf.experimental.numpy.sqrt tf.experimental.numpy.square tf.experimental.numpy.squeeze tf.experimental.numpy.std tf.experimental.numpy.sum tf.experimental.numpy.swapaxes tf.experimental.numpy.tan tf.experimental.numpy.tanh tf.experimental.numpy.trace tf.experimental.numpy.transpose tf.experimental.numpy.triu tf.experimental.numpy.vander tf.experimental.numpy.var tf.experimental.numpy.zeros_like tf.bitwise.invert tf.bitwise.invert tf.clip_by_value tf.clip_by_value tf.debugging.check_numerics tf.debugging.check_numerics tf.expand_dims tf.expand_dims tf.identity tf.identity tf.image.adjust_brightness tf.image.adjust_brightness tf.image.adjust_gamma tf.image.adjust_gamma tf.image.extract_patches tf.image.extract_patches tf.image.random_brightness tf.image.random_brightness tf.image.stateless_random_brightness tf.image.stateless_random_brightness tf.linalg.diag tf.linalg.diag tf.linalg.diag_part tf.linalg.diag_part tf.linalg.matmul tf.linalg.matmul tf.linalg.matrix_transpose tf.linalg.matrix_transpose tf.linalg.tensor_diag_part tf.linalg.tensor_diag_part tf.linalg.trace tf.linalg.trace tf.math.abs tf.math.abs tf.math.acos tf.math.acos tf.math.acosh tf.math.acosh tf.math.add tf.math.add tf.math.angle tf.math.angle tf.math.asin tf.math.asin tf.math.asinh tf.math.asinh tf.math.atan tf.math.atan tf.math.atanh tf.math.atanh tf.math.ceil tf.math.ceil tf.math.conj tf.math.conj tf.math.cos tf.math.cos tf.math.cosh tf.math.cosh tf.math.digamma tf.math.digamma tf.math.divide_no_nan tf.math.divide_no_nan tf.math.divide tf.math.divide tf.math.erf tf.math.erf tf.math.erfc tf.math.erfc tf.math.erfcinv tf.math.erfcinv tf.math.erfinv tf.math.erfinv tf.math.exp tf.math.exp tf.math.expm1 tf.math.expm1 tf.math.floor tf.math.floor tf.math.floordiv tf.math.floordiv tf.math.floormod tf.math.floormod tf.math.imag tf.math.imag tf.math.lgamma tf.math.lgamma tf.math.log1p tf.math.log1p tf.math.log_sigmoid tf.math.log_sigmoid tf.math.log tf.math.log tf.math.multiply_no_nan tf.math.multiply_no_nan tf.math.multiply tf.math.multiply tf.math.ndtri tf.math.ndtri tf.math.negative tf.math.negative tf.math.pow tf.math.pow tf.math.real tf.math.real tf.math.real tf.math.real tf.math.reciprocal_no_nan tf.math.reciprocal_no_nan tf.math.reciprocal tf.math.reciprocal tf.math.reduce_euclidean_norm tf.math.reduce_euclidean_norm tf.math.reduce_logsumexp tf.math.reduce_logsumexp tf.math.reduce_max tf.math.reduce_max tf.math.reduce_mean tf.math.reduce_mean tf.math.reduce_min tf.math.reduce_min tf.math.reduce_prod tf.math.reduce_prod tf.math.reduce_std tf.math.reduce_std tf.math.reduce_sum tf.math.reduce_sum tf.math.reduce_variance tf.math.reduce_variance tf.math.rint tf.math.rint tf.math.round tf.math.round tf.math.rsqrt tf.math.rsqrt tf.math.scalar_mul tf.math.scalar_mul tf.math.sigmoid tf.math.sigmoid tf.math.sign tf.math.sign tf.math.sin tf.math.sin tf.math.sinh tf.math.sinh tf.math.softplus tf.math.softplus tf.math.special.bessel_i0 tf.math.special.bessel_i0 tf.math.special.bessel_i0e tf.math.special.bessel_i0e tf.math.special.bessel_i1 tf.math.special.bessel_i1 tf.math.special.bessel_i1e tf.math.special.bessel_i1e tf.math.special.bessel_j0 tf.math.special.bessel_j0 tf.math.special.bessel_j1 tf.math.special.bessel_j1 tf.math.special.bessel_k0 tf.math.special.bessel_k0 tf.math.special.bessel_k0e tf.math.special.bessel_k0e tf.math.special.bessel_k1 tf.math.special.bessel_k1 tf.math.special.bessel_k1e tf.math.special.bessel_k1e tf.math.special.bessel_y0 tf.math.special.bessel_y0 tf.math.special.bessel_y1 tf.math.special.bessel_y1 tf.math.special.dawsn tf.math.special.dawsn tf.math.special.expint tf.math.special.expint tf.math.special.fresnel_cos tf.math.special.fresnel_cos tf.math.special.fresnel_sin tf.math.special.fresnel_sin tf.math.special.spence tf.math.special.spence tf.math.sqrt tf.math.sqrt tf.math.square tf.math.square tf.math.subtract tf.math.subtract tf.math.tan tf.math.tan tf.math.tanh tf.math.tanh tf.nn.depth_to_space tf.nn.depth_to_space tf.nn.elu tf.nn.elu tf.nn.gelu tf.nn.gelu tf.nn.leaky_relu tf.nn.leaky_relu tf.nn.log_softmax tf.nn.log_softmax tf.nn.relu6 tf.nn.relu6 tf.nn.relu tf.nn.relu tf.nn.selu tf.nn.selu tf.nn.softsign tf.nn.softsign tf.nn.space_to_depth tf.nn.space_to_depth tf.nn.swish tf.nn.swish tf.ones_like tf.ones_like tf.realdiv tf.realdiv tf.reshape tf.reshape tf.squeeze tf.squeeze tf.stop_gradient tf.stop_gradient tf.transpose tf.transpose tf.truncatediv tf.truncatediv tf.truncatemod tf.truncatemod tf.zeros_like tf.zeros_like tf.experimental.numpy.abs tf.experimental.numpy.abs tf.experimental.numpy.absolute tf.experimental.numpy.absolute tf.experimental.numpy.amax tf.experimental.numpy.amax tf.experimental.numpy.amin tf.experimental.numpy.amin tf.experimental.numpy.angle tf.experimental.numpy.angle tf.experimental.numpy.arange tf.experimental.numpy.arange tf.experimental.numpy.arccos tf.experimental.numpy.arccos tf.experimental.numpy.arccosh tf.experimental.numpy.arccosh tf.experimental.numpy.arcsin tf.experimental.numpy.arcsin tf.experimental.numpy.arcsinh tf.experimental.numpy.arcsinh tf.experimental.numpy.arctan tf.experimental.numpy.arctan tf.experimental.numpy.arctanh tf.experimental.numpy.arctanh tf.experimental.numpy.around tf.experimental.numpy.around tf.experimental.numpy.array tf.experimental.numpy.array tf.experimental.numpy.asanyarray tf.experimental.numpy.asanyarray tf.experimental.numpy.asarray tf.experimental.numpy.asarray tf.experimental.numpy.ascontiguousarray tf.experimental.numpy.ascontiguousarray tf.experimental.numpy.average tf.experimental.numpy.average tf.experimental.numpy.bitwise_not tf.experimental.numpy.bitwise_not tf.experimental.numpy.cbrt tf.experimental.numpy.cbrt tf.experimental.numpy.ceil tf.experimental.numpy.ceil tf.experimental.numpy.conj tf.experimental.numpy.conj tf.experimental.numpy.conjugate tf.experimental.numpy.conjugate tf.experimental.numpy.copy tf.experimental.numpy.copy tf.experimental.numpy.cos tf.experimental.numpy.cos tf.experimental.numpy.cosh tf.experimental.numpy.cosh tf.experimental.numpy.cumprod tf.experimental.numpy.cumprod tf.experimental.numpy.cumsum tf.experimental.numpy.cumsum tf.experimental.numpy.deg2rad tf.experimental.numpy.deg2rad tf.experimental.numpy.diag tf.experimental.numpy.diag tf.experimental.numpy.diagflat tf.experimental.numpy.diagflat tf.experimental.numpy.diagonal tf.experimental.numpy.diagonal tf.experimental.numpy.diff tf.experimental.numpy.diff tf.experimental.numpy.empty_like tf.experimental.numpy.empty_like tf.experimental.numpy.exp2 tf.experimental.numpy.exp2 tf.experimental.numpy.exp tf.experimental.numpy.exp tf.experimental.numpy.expand_dims tf.experimental.numpy.expand_dims tf.experimental.numpy.expm1 tf.experimental.numpy.expm1 tf.experimental.numpy.fabs tf.experimental.numpy.fabs tf.experimental.numpy.fix tf.experimental.numpy.fix tf.experimental.numpy.flatten tf.experimental.numpy.flatten tf.experimental.numpy.flip tf.experimental.numpy.flip tf.experimental.numpy.fliplr tf.experimental.numpy.fliplr tf.experimental.numpy.flipud tf.experimental.numpy.flipud tf.experimental.numpy.floor tf.experimental.numpy.floor tf.experimental.numpy.full_like tf.experimental.numpy.full_like tf.experimental.numpy.imag tf.experimental.numpy.imag tf.experimental.numpy.log10 tf.experimental.numpy.log10 tf.experimental.numpy.log1p tf.experimental.numpy.log1p tf.experimental.numpy.log2 tf.experimental.numpy.log2 tf.experimental.numpy.log tf.experimental.numpy.log tf.experimental.numpy.max tf.experimental.numpy.max tf.experimental.numpy.mean tf.experimental.numpy.mean tf.experimental.numpy.min tf.experimental.numpy.min tf.experimental.numpy.moveaxis tf.experimental.numpy.moveaxis tf.experimental.numpy.nanmean tf.experimental.numpy.nanmean tf.experimental.numpy.negative tf.experimental.numpy.negative tf.experimental.numpy.ones_like tf.experimental.numpy.ones_like tf.experimental.numpy.positive tf.experimental.numpy.positive tf.experimental.numpy.prod tf.experimental.numpy.prod tf.experimental.numpy.rad2deg tf.experimental.numpy.rad2deg tf.experimental.numpy.ravel tf.experimental.numpy.ravel tf.experimental.numpy.real tf.experimental.numpy.real tf.experimental.numpy.reciprocal tf.experimental.numpy.reciprocal tf.experimental.numpy.repeat tf.experimental.numpy.repeat tf.experimental.numpy.reshape tf.experimental.numpy.reshape tf.experimental.numpy.rot90 tf.experimental.numpy.rot90 tf.experimental.numpy.round tf.experimental.numpy.round tf.experimental.numpy.signbit tf.experimental.numpy.signbit tf.experimental.numpy.sin tf.experimental.numpy.sin tf.experimental.numpy.sinc tf.experimental.numpy.sinc tf.experimental.numpy.sinh tf.experimental.numpy.sinh tf.experimental.numpy.sort tf.experimental.numpy.sort tf.experimental.numpy.sqrt tf.experimental.numpy.sqrt tf.experimental.numpy.square tf.experimental.numpy.square tf.experimental.numpy.squeeze tf.experimental.numpy.squeeze tf.experimental.numpy.std tf.experimental.numpy.std tf.experimental.numpy.sum tf.experimental.numpy.sum tf.experimental.numpy.swapaxes tf.experimental.numpy.swapaxes tf.experimental.numpy.tan tf.experimental.numpy.tan tf.experimental.numpy.tanh tf.experimental.numpy.tanh tf.experimental.numpy.trace tf.experimental.numpy.trace tf.experimental.numpy.transpose tf.experimental.numpy.transpose tf.experimental.numpy.triu tf.experimental.numpy.triu tf.experimental.numpy.vander tf.experimental.numpy.vander tf.experimental.numpy.var tf.experimental.numpy.var tf.experimental.numpy.zeros_like tf.experimental.numpy.zeros_like Originally published on the 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. Originally published on the 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. Originally published on the 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. TensorFlow TensorFlow