GPT-2 Running on a Pi 4 8 Gig with Tensorflow 1.13.1, testing CPU and Memory Usage
Running Htop while feeding GPT-2's 345M model with a text prompt.
Parameters I have running in the code:
def interact_model(
model_name='345M',
seed=None,
nsamples=1,
batch_size=1,
length=140,
temperature=1.2,
top_k=48,
top_p=0.7,
models_dir='models',
):
No mods to clocking, etc. This is benchmarking prior to optimizing of any kind. There is a fairly long delay from prompt input to text response output, and I need to cut that way down. Open to suggestions. Thanks for watching.