Tensorboard only shows a single point

There are edge cases where tensorboard will only show a single dot

This means that MMvec only ran a single step. There are a couple of known things that could cause this
1. The number of sequences in the dataset is smaller than the batch size. In which cause you need to decrease the batch size (the default is a batch size of 50 sequences)
2. The number of epochs is only 1, which case you need to increase the number of epochs
3. Your summary interval is too long. It is possible that your summary interval is longer than your actual MMvec run, in which case you need to decrease your summary interval to something reasonable (i.e. 100x smaller than the length of your MMvec run).

If this is helpful, we will migrate the contents inside of this issue to the README.

Review the README and this issue's TensorBoard troubleshooting guidance first. Move the relevant causes and remedies into the README, preserving the batch-size, epoch, and summary-interval examples; done means users can find guidance for a single-point TensorBoard run there.

Written by the indexing model from the issue text.