drculhane
drculhane
Good thought, and I’m a bit unhappy with the log trick anyway. I don’t see why that’s a good thing, and I can picture situations where it’s a bad idea....
Thanks. That falls squarely in the "oops" category. Fixed, and about to commit. On Thu, Mar 21, 2024 at 9:58 AM ajpotts ***@***.***> wrote: > ***@***.**** commented on this pull...
The prob-size is another Oops. I'll correct it, and make that change to clip. --- Andy On Thu, Mar 21, 2024 at 10:45 AM tess ***@***.***> wrote: > ***@***.**** commented...
It was a deliberate decision not to use more parameterization, because I felt that brevity would get in the way of clarity. But I'll still look at it again anyway,...
Good suggestions. I've noted that the PROTOs tests are inconsistent with how they do assertions. I'll look for places where I can use assert_frame_equal. Also, oops again on letting old...
So that entire chunk could be replaced by: assert_frame_equal(ij_expected_df.to_pandas(),ij_merged_df.to_pandas()) Right? There are several of these, so I'll go through all of them to see whether the same thing applies. On...
Is this really a bug? numpy gives the same results: >>> x = np.array([np.nan,1.09]) >>> x==x array([False, True]) >>> x>> x
Also adds tests for square, isinf and isfinite to tests/numpy/numeric_test.py
Line 21 above issues a call to generic_msg with command = "matMul". I'm puzzled. I don't see this anywhere in the chapel code. There is a matmul in LinalgMsg.chpl, but...
I think you're right about this broadcast_dims function.