I followed this fine-tuning tutorial and wanted to compare the performance of the base “bert-base-uncased” model vs the fine-tuned model on a sample sentence pair. I’m not sure what the task str should be passed to the pipeline (for the base model inference), based on the API doc.
Also if I try to do below, I get “RuntimeError: Inferring the task automatically requires to check the hub with a model_id defined as a str”, but it’s not clear what else needs to be passed.
basechkpnt = "bert-base-uncased"
basetokenizer = AutoTokenizer.from_pretrained(basechkpnt)
basemodel = AutoModelForSequenceClassification.from_pretrained(basechkpnt)
basepipe = pipeline(model=basemodel, tokenizer=basetokenizer) # <-- gives error