inf loss at big batch

just creating a todo. large batch sizes work now having fixed the `size_t` bug:

```
./train_gpt2cu -b 36 -v 200 -s 200 -i data/TinyStories
```

works, but 48 should fit but doesn't work

```
./train_gpt2cu -b 48 -v 200 -s 200 -i data/TinyStories
```

val loss is -nan and train loss stays at inf.

todo track down why and how to prevent

Start by reproducing the issue with the train_gpt2cu commands and the data/TinyStories dataset, comparing batch sizes 36 and 48. Track the source of the inf training loss and -nan validation loss; done means batch size 48 trains with finite losses and a preventative fix is identified.

Written by the indexing model from the issue text.