The article introduces tensors as multi-dimensional arrays known for their uniform data types (dtypes), highlighting their similarity to NumPy arrays. It emphasizes that tensors are immutable; their contents can only be modified by creating new tensors rather than altering existing ones. The article also outlines the process of creating basic tensors, such as scalars or rank-0 tensors, and mentions common issues related to registering GPU libraries, which could hinder performance in TensorFlow applications. Lastly, supported data types are available to examine via tf.dtypes, enhancing user flexibility.
Tensors are multi-dimensional arrays with a uniform type (called a dtype) and are immutable, meaning their contents cannot be updated after creation.
Creating basic tensors is simple, and the example illustrates a scalar or rank-0 tensor which contains a single value without any axes.
All supported dtypes can be viewed at tf.dtypes, showing the flexibility of tensor types akin to that of NumPy arrays.
The warning messages indicate issues registering cuFFT, cuDNN, and cuBLAS factories, which could affect the operation of TensorFlow with GPU capabilities.
Collection
[
|
...
]