Issue: HuggingFaceEmbeddings can not take trust_remote_code argument

[jackfrost1411]

HuggingFaceEmbeddings can not take trust_remote_code argument

Suggestion:

No response

[damingerdai]

i also meet the same issue

[damingerdai]

i fixed by this code:

from langchain.llms import HuggingFacePipeline
from langchain.memory.buffer import ConversationBufferMemory
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline, AutoModelForSeq2SeqLM

local_llm = HuggingFacePipeline.from_model_id(
        model_id="chatglm-6b-int4",
        task="text-generation",
        model_kwargs={
            "temperature": 0, "max_length": 64,
+           "trust_remote_code": True
        },
    )
print(local_llm('What is the capital of France? '))

[nbbaier]

The above solution doesn't work for HuggingFaceEmbeddings:

from langchain.embeddings import HuggingFaceEmbeddings

model_name = "jinaai/jina-embeddings-v2-small-en"
model_kwargs = {"device": "cpu", "trust_remote_code": True}
encode_kwargs = {
    "normalize_embeddings": False,
}
hf = HuggingFaceEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
    encode_kwargs=encode_kwargs,
)
text = "This is a test document."

res = hf.embed_query(text)

print(len(res))

Results in the following error:

Traceback (most recent call last):
  File "/Users/nbbaier/grubby-galaxy/main.py", line 9, in <module>
    hf = HuggingFaceEmbeddings(
         ^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nbbaier/.pyenv/versions/3.11.0/lib/python3.11/site-packages/langchain/embeddings/huggingface.py", line 66, in __init__
    self.client = sentence_transformers.SentenceTransformer(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: SentenceTransformer.__init__() got an unexpected keyword argument 'trust_remote_code'

[GlitChed2k2]

The above solution doesn't work for HuggingFaceEmbeddings:

from langchain.embeddings import HuggingFaceEmbeddings

model_name = "jinaai/jina-embeddings-v2-small-en"
model_kwargs = {"device": "cpu", "trust_remote_code": True}
encode_kwargs = {
    "normalize_embeddings": False,
}
hf = HuggingFaceEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
    encode_kwargs=encode_kwargs,
)
text = "This is a test document."

res = hf.embed_query(text)

print(len(res))

Results in the following error:

Traceback (most recent call last):
  File "/Users/nbbaier/grubby-galaxy/main.py", line 9, in <module>
    hf = HuggingFaceEmbeddings(
         ^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/nbbaier/.pyenv/versions/3.11.0/lib/python3.11/site-packages/langchain/embeddings/huggingface.py", line 66, in __init__
    self.client = sentence_transformers.SentenceTransformer(
                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
TypeError: SentenceTransformer.__init__() got an unexpected keyword argument 'trust_remote_code'

Facing the same issue! Any solution for this?

[gururise]

Has anyone found a solution to this? Seems like we should be able to pass 'trust_remote_code' into kwargs of the HugginfFaceEmbeddings.

[Nafay-0]

from langchain.embeddings import HuggingFaceEmbeddings
from transformers import AutoModel
model = AutoModel.from_pretrained('jinaai/jina-embeddings-v2-base-en', trust_remote_code=True) 

model_name = "jinaai/jina-embeddings-v2-base-en"
model_kwargs = {'device': 'cpu'}
embeddings = HuggingFaceEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
)

Found the following work around for the problem by loading the embedding model via Transformers class.

[christophfroeschl]

workaround is good but a fix for the underlying issue pass

trust_remote_code into kwargs of the HugginfFaceEmbeddings

is welcome

[mallapraveen]

is this fixed?

[jcgeo9]

same isuse here, still waiting for the fix :)

[manjunathshiva]

Same issue! Need fix for the same

[weissenbacherpwc]

so there is the same performance when loading the embeddings model with:

from transformers import AutoModel
model = AutoModel.from_pretrained('PATH_TO_LOCAL_EMBEDDING_MODEL_FOLDER', trust_remote_code=True)

instead of:
from langchain.embeddings import HuggingFaceEmbeddings

model_name = "PATH_TO_LOCAL_EMBEDDING_MODEL_FOLDER"
model_kwargs = {'device': 'cpu'}
embeddings = HuggingFaceEmbeddings(
model_name=model_name,
model_kwargs=model_kwargs,
)

I figured out that some embeddings have a sligthly different value, so enabling "trust_remote_code=True" would be appreciated!

[hidacow]

pip install -U sentence-transformers

Update sentence-transformers to >=2.3.1 is working for me.
Seems trust_remote_code is not introduced into the constructor of SentenceTransformer before 2.3

[shimonl20]

This worked for me...

model_kwargs = {'trust_remote_code': True}
embeddings = HuggingFaceEmbeddings(model_name="jinaai/jina-embeddings-v2-base-en",
                                       model_kwargs=model_kwargs)

[weissenbacherpwc]

This worked for me...

model_kwargs = {'trust_remote_code': True}
embeddings = HuggingFaceEmbeddings(model_name="jinaai/jina-embeddings-v2-base-en",
                                       model_kwargs=model_kwargs)

are you using Langchain Huggingface Embeddings here?

[shimonl20]

This worked for me...

model_kwargs = {'trust_remote_code': True}
embeddings = HuggingFaceEmbeddings(model_name="jinaai/jina-embeddings-v2-base-en",
                                       model_kwargs=model_kwargs)

are you using Langchain Huggingface Embeddings here?

Yes

[KasemOmary]

embedding_model = HuggingFaceEmbeddings(model_name="whatever-model-you-are-using", model_kwargs={"trust_remote_code":True})

Worked fine for me. As mentioned earlier, be sure to update the versions being used. I've used transformer 2.4 models with no issues.

[ps360pa]

pip install -U sentence-transformers

Update sentence-transformers to >=2.3.1 is working for me. Seems trust_remote_code is not introduced into the constructor of SentenceTransformer before 2.3

pip install -U sentence-transformers==2.4.0
it works for 2.4.0 but 2.5.1 the sentence-transformers get error

[ramamimu]

pip install -U sentence-transformers==2.4.0

I am still facing another error which seems like similar things
TypeError: __init__() takes 1 positional argument but 2 were given