IndexError: index xxxxxx is out of bounds for dimension 0 with size 36

[igor17400]

Hi, when executing some models such as NRMS and MINER after a few epochs—2 for NRMS and 5 for MINER—I am encountering the following error message:

IndexError: index 1021546040 is out of bounds for dimension 0 with size 36

From my investigations, it seems that the loss might be exploding and the forward pass is returning nan vectors. For instance, before receiving the error, I receive warning messages like:

Encountered `nan` values in tensor. Will be removed.  warnings.warn(*args, **kwargs)  # noqa: B028

I am using the default configuration files:

• NRMS: NRMS configuration
• MINER: MINER configuration

---

I plan to add some print statements to better evaluate the bug. Has anyone else faced this issue?