You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
we might not always want greedy sampling do we?
Could you implement do_sample beeing an init param for JsonFormer or is there anything technically that prohibits this change?
The text was updated successfully, but these errors were encountered:
I was able to set my model to do_sample=True before passing it to jsonformer. model = AutoModelForCausalLM.from_pretrained( "databricks/dolly-v2-3b", device_map='auto', use_cache=True, do_sample=True)
we might not always want greedy sampling do we?
Could you implement do_sample beeing an init param for JsonFormer or is there anything technically that prohibits this change?
The text was updated successfully, but these errors were encountered: