Config accurate?
#1
by
bartowski
- opened
The config.json lists the model architecture as "MistralModel", never seen that before, is that a typo meant to say "MistralForCausalLM" ?
Hmm...great catch! I'm not sure why it says that. This model was created using an Unsloth notebook as an experiment and uploaded straight from there to the hub with the merged model as-is. Maybe something with their framework? Still works fine on my local machine with Text-Gen-WebUI. I'll keep investigating
EDIT: Just to be safe, I fixed it. Seemed like an error on the model push. Thanks!!
Severian
changed discussion status to
closed