Patch/adjust num batches

[uoefeb]

This pull request makes a few changes to finetune.py:

• round number of batches up: the number of batches was calculated as int(len(videos)/batch_size) causing the number of batches to round down to zero if the amount of data was less half the size of a batch. This pull request replaces int() with ceil(), always rounding up to the nearest full batch.
• adds a few comments to the train() method
• replaces commented out print statements with logging
• moves the batch creation outside of the epoch loop -- it looked like they weren't changing across loops, so I moved them outside of the loop. Please let me know if I've misunderstood this.
• deletes assignment of prefixes ... it doesn't look like this was used anywhere in the function.

---

in addition, this adds videos to .gitignore

[uoefeb]

@SanniM3 Given all the changes that have been made to finetune, it's probably best that we abandon this pull request. That said, I think we still probably need to change the calculation of the number of batches to the following, so we don't round off any partial batches:

ceil(len(full_list)/batch_size)