Iikka Hauhio

`nvidia-smi` shows that there is available memory. Setting `CUDA_LAUNCH_BLOCKING=1` remotes the `For debugging consider passing CUDA_LAUNCH_BLOCKING=1.` line at the end, but doesn't change the stack trace.

The log contains this error before the random errors start to occur: ``` Token indices sequence length is longer than the specified maximum sequence length for this model (1133 >...

I confirmed that the error above is the cause of the problem, so I'm closing this issue. I'm still puzzled how an error in a different library can cause errors...

I'll try this week if I can get the bug reproduced in a minimal test program. I extracted the calls related to translation from our program, they are like this:...

I ran the above script in the Amazon server we use and it produces this output. Note that the first error is for translating the too long text, but the...

I tried updating the libraries and the error changes. The second sentence still fails even though it shouldn't, but the error is different: ``` $ CUDA_LAUNCH_BLOCKING=1 python /tmp/test.py 2023-01-19 13:31:03...

Why not just disallow space at the end in the grammar that is used to constrain the output? Right now it allows tailing whitespace, but I see no reason for...