When implementing tensor-based algorithms, avoid assumptions about fixed tensor shapes by not relying on direct shape attributes (`.shape`, `.ndim`), which may return `None` in graph mode. Instead, use TensorFlow operations like `tf.shape()` and `tf.rank()` which work in both eager and graph execution modes. For validation, use TensorFlow's conditional operations rather than Python conditionals. For example, replace: ```python # Problematic code - may fail in graph mode mat1_shape = mat1.shape dense_rows = mat1_shape[-2] # This might fail if shape is not known if dense_shape[-2] != dense_rows: # Python conditional won't work with tensors raise ValueError("Shapes don't match") ``` With: ```python # Robust code that works in both eager and graph modes mat1_shape = array_ops.shape(mat1) dense_rows = mat1_shape[-2] condition = math_ops.equal(dense_shape[-2], dense_rows) gen_logging_ops._assert(condition, ["Shapes don't match"], name="Assert") ``` This approach ensures algorithms work reliably regardless of whether shapes are statically known at graph construction time or only determined during execution.