AssertionError: tensorly.tt_tensor.validate_tt_rank test

[mobley-trent]

Describe the bug

The assertion test case for this tt_tensor method is failing when rounding="floor".
Only the test case for rounding="ceil" is passing.

Steps or Code to Reproduce

This is the code from the test case

import numpy as np
import tensorly as tl

coef = 0.2
tensor_shape = tuple(np.random.randint(5, 10, size=4))
n_param_tensor = np.prod(tensor_shape)

rank = tl.tt_tensor.validate_tt_rank(tensor_shape, coef, rounding="floor")
n_param = tl.tt_tensor._tt_n_param(tensor_shape, rank)
np.testing.assert_(n_param >= n_param_tensor * coef)

Expected behavior

The test case is supposed to pass.

Actual result

The test is returning an AssertionError

Versions

Linux-5.15.0-1041-azure-x86_64-with-glibc2.28
Python 3.10.0 (default, Mar 3 2022, 09:58:08) [GCC 7.5.0]
NumPy 1.25.2
SciPy 1.11.2
JAX 0.4.14
PyTorch 2.0.1+cpu
TensorFlow 2.13.0
TensorLy 0.8.1

[JeanKossaifi]

@mobley-trent this is not the code from the test, you changed only the rounding but not the check. You should change the check too if you want to use floor - you'd want to change the greater or equal to lower or equal since you're rounding down now instead of up.

The actual test-case is

   # Rounding = floor
    rank = validate_tt_rank(tensor_shape, rank=coef, rounding="floor")
    n_param = _tt_n_param(tensor_shape, rank)
    assert_(n_param <= n_param_tensor * coef)

See, test-file. All the tests are run at each new commit.