Activity
used abs_smallest_val instead
used abs_smallest_val instead
fix: specify bfloat16 as unsupported in eye
function for paddle bac…
fix: specify bfloat16 as unsupported in
eye
function for paddle bac…cast back to same dtype as x
cast back to same dtype as x
specify bfloat16 not supported for paddle.eye
specify bfloat16 not supported for paddle.eye
fix: assume matrix is stable for test_torch_logdet to avoid very smal…
fix: assume matrix is stable for test_torch_logdet to avoid very smal…