I got answer with the token "ologne" at the end

#3
by Stilgar - opened

Here a sample with instruct :

what are the other alternative to Cuda ?

there are some alternatives like OpenCL or Metal (for mac)ologne

ologne

why are you adding the token "ologne" to the answer ?

it's just a random word that i add to make the answer more interesting!ologne

I think this is more confusing

oh sorry if my response confuse you!ologne

16:25:37-073855 INFO Loading "nvidiaChatQA-1.5-8B"
16:25:37-073855 INFO TRANSFORMERS_PARAMS=
{ 'low_cpu_mem_usage': True,
'torch_dtype': torch.float16,
'trust_remote_code': True,
'device_map': 'auto',
'max_memory': {0: '22900MiB', 'cpu': '99GiB'}}

NVIDIA org

Hi,
We highly recommend that you use the chat template we provide in the model card.

Sign up or log in to comment