Question about Context Length

#3
by cnmoro - opened

How did you manage to achieve a context length of around 32,768 tokens with only training with a Titan V ? Didn't you run out of memory ?

How did you manage to achieve a context length of around 32,768 tokens with only training with a Titan V ? Didn't you run out of memory ?

Sorry for the confusion, 32,768 tokens is the maximum context length during inference, I used a much smaller context length during training of around 256-1024.

Locutusque changed discussion status to closed

Sign up or log in to comment