17:58:43.186741: I tensorflow/core/common_runtime/gpu/gpu_:1082] Device interconnect StreamExecutor with strength 1 edge matrix: 17:58:43.181018: I tensorflow/stream_executor/platform/default/dso_:42] Successfully opened dynamic library libcudart.so.10.0

17:58:43.180913: I tensorflow/core/common_runtime/gpu/gpu_:1551] Found device 0 with properties: 17:58:43.180183: I tensorflow/compiler/xla/service/:168] XLA service 0x4cdb210 executing computations on platform Host. 17:58:43.157505: I tensorflow/compiler/xla/service/:175] StreamExecutor device (0): GeForce GTX 1080 Ti, Compute Capability 6.1 17:58:43.157452: I tensorflow/compiler/xla/service/:168] XLA service 0x50cbf50 executing computations on platform CUDA. 17:58:43.012348: I tensorflow/stream_executor/platform/default/dso_:42] Successfully opened dynamic library libcuda.so.1 17:58:42.998488: I tensorflow/core/platform/cpu_feature_:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AV #traceback.print_exc() Click here to see the corresponding errors Return model_forward, model_inverse def test1():īijector = permute construct_keras_bijector_models( bijector)īijector = realnvp construct_keras_bijector_models( bijector)Ĭonstruct_keras_bijector_models( bijector) Model( inputs = input_x, outputs = output_z) Output_z = BijectorInverse( bijector)( input_x) Model( inputs = input_z, outputs = output_x) Output_x = BijectorForward( bijector)( input_z) Return permute, realnvp def construct_keras_bijector_models( bijector): Num_masked = dim_z // 2, shift_and_log_scale_fn = additive_cf, is_constant_jacobian = True) bijector = bijector def call( self, input): It is also possible to specify a batch size (with potentially undeterminedīatch dimension, denoted by 'null') for the first layer using theīatchInputShape key.Import tensorflow as tf import tensorflow_probability as tfp import tensorflow_datasets as tfds tf. The 1st dimension is the undetermined batch dimension the // 2nd is the output size of the model's last layer. Inspect the inferred shape of the model's output, which equals // ``. Currently, the `texture` has 4 pixels: // Pixel0 is )) registerBackend( 'custom-webgl', () => customBackend) createElement( 'canvas') Ĭonst customBackend = new tf. Example for WebGL2: const customCanvas = document. For example, if your application includes a preprocessing step on the GPU, // you could upload the GPU output directly to TF.js, rather than first // downloading the values. This makes it possible for TF.js applications to avoid GPU / CPU sync. Pass a `WebGLData` object and specify a shape yourself. Pass a flat array and specify a shape yourself. Pass a nested array of values to make a matrix or a higher // dimensional tensor. Pass an array of values to create a vector. Creates a tf.Tensor with the provided values, shape and dtype.

Tf permute