Error in vector TMA descriptor creation: invalid argument

[jafioti]

When I build the low_latency_demo and run the repl script, I am getting:

Error in vector TMA descriptor creation: invalid argument

Seems like it prints on every forward pass. Also when I ask "How are you?" it responds with:

I'm doing well, thank you for asking. I'm a machine learning model, so I don't have emotions or feelings like humans do, but of of the difference. I'm just a large, an I am, in the language, with the ability to, and I am, with and I, is a, of, a, of, is a, of, in, of and, in, a and so, is a return, or, or, return, or, is a, or, or, or, or, or return to and, and and end, or and and and return,, or and and and and of of of and return,.

 and of of of of of of and return, return, or return,r of of of and endors, a  isr, or, or, or, or, orr and, or, or, or, or, and, end, or, or, or, or, end, or, or, or, orp return, or, or, or, or, or, or, or, or, or, or, or, or, and or, or, or, or, end, or, or, or, or of and r, or,r, or, or, r, or,,

The, or, or, or, or, or, r, r, or, or, andr and, of, or, or, or, and or, or,, and--

The and, or, or, or, or, or, or, and, and and and and and and and and and, and and and and and and, of of p, or, or, or, and r, or, and and, and and, and and, r, r, r, r, r, r, p, or, r, or, or, and and and and and and and and and and and and and and and, and of and and and, r, r, r, r, r, p, r, p, r, p, r, p, or, r, r, r, r, r, r, r, r, or, r, r, p, r, r, p, r, r, r, r, r, r, r, r, p, r, p, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, r, is a, r, r, is, the, r, the, r, r, is a, r, r, r, r, is, r, is, is, the, r, is, is, r, is, the, r, is, is, the, the, the, is, is, is, the, the, is, is, is, is, is, is, the, r, is, is, is, or, r, r, is, is, is, is, is, is, is, r, is, is, is, is, is, is, is, is, the, or, or, or, or, or, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is, is
Speed: 438.59 tokens/s

So it seems like some error builds up or maybe there is an error at a certian sequence length.

[jafioti]

For reproducability: Python 3.13.9, Cuda 12.8, H100 PCIE