I’m a beginner working with code that previously used this implementation to access the blenderbot-3B:
from transformers import pipeline
from transformers.pipelines.conversational import Conversation
device = torch.cuda.current_device() if torch.cuda.is_available() else -1
conversational_pipeline = pipeline("conversational", model="facebook/blenderbot-3B", device=device)
conversation = Conversation()
The remainder of the code follows and has been working perfectly. It is running on an Amazon EC2 Instance in a docker container.
Unfortunately I am now receiving this error after not changing my implementation at all:
ValueError: Could not load model facebook/blenderbot-3B with any of the following classes: (<class 'transformers.models.auto.modeling_auto.AutoModelForSeq2SeqLM'>, <class 'transformers.models.auto.modeling_auto.AutoModelForCausalLM'>, <class 'transformers.models.blenderbot.modeling_blenderbot.BlenderbotForConditionalGeneration'>).
I’ve tried upgrading transformers to the newest version (which lead to a different error) and using a local path is not possible. Any idea what could have sparked this new issue?