[!IMPORTANT]
this was finetuned at 16384 context length
This is an experiment on long context text generation (i.e. 6k+ tokens generated) to evaluate if/when generation breaks down, etc. As such, all the data on which this model has been fine-tuned are full-length books.
