LangChain + Hugging Face -> HuggingFacePIpeline Error

Question

I'm trying to do the following simple code:

from transformers import pipeline
import langchain
from langchain.llms import HuggingFacePipeline

model_name = "bert-base-uncased"
task = "question-answering"

hf_pipeline = pipeline(task, model=model_name)

langchain_pipeline = HuggingFacePipeline(hf_pipeline)

I get the following error:

• ERROR: TypeError: Serializable.__init__() takes 1 positional argument but 2 were given
• LINE: langchain_pipeline = HuggingFacePipeline(hf_pipeline)

Haven't found anything online that actually helped me here

---

I'm using Databricks with the following cluster:

• Runtime: 12.2 LTS ML (includes Apache Spark 3.3.2, Scala 2.12)
• Node type: Standard_DS5_v2 56 GB Memory, 16 Cores
• Libraries:

Answer 1

Here's the code example

# from transformers import pipeline
# import langchain
from langchain_community.llms import HuggingFacePipeline

model_name = "bert-base-uncased"
# task = "question-answering"
task = "text-generation"

# hf_pipeline = pipeline(task, model=model_name)

langchain_pipeline = HuggingFacePipeline.from_model_id(
    model_name,
    task,
)

And only ('text2text-generation', 'text-generation', 'summarization') are supported

HuggingFacePipeline doc for reference > https://python.langchain.com/docs/integrations/llms/huggingface_pipelines