TheBloke

/

Nous-Hermes-13B-GPTQ

like 175

1. prompt_template variable was missing before the first run, which prevented tokenizer from working, got the error:

Traceback (most recent call last):
File "/home/sd/hf/nous-hermes.py", line 22, in
input_ids = tokenizer(prompt_template, return_tensors='pt').input_ids.cuda()
NameError: name 'prompt_template' is not defined

2. adjusted prompt_template text in both cases in the instruction so it is the same as in previous description above (alpaca format). However, I am not sure if the whitespace after the "Instruction:" matters or needed.