encoding_kwargs for InstructEmbeddings (#5450)

# What does this PR do?

Bring support of `encode_kwargs` for ` HuggingFaceInstructEmbeddings`,
change the docstring example and add a test to illustrate with
`normalize_embeddings`.

Fixes #3605
(Similar to #3914)

Use case:
```python
from langchain.embeddings import HuggingFaceInstructEmbeddings

model_name = "hkunlp/instructor-large"
model_kwargs = {'device': 'cpu'}
encode_kwargs = {'normalize_embeddings': True}
hf = HuggingFaceInstructEmbeddings(
    model_name=model_name,
    model_kwargs=model_kwargs,
    encode_kwargs=encode_kwargs
)
```
This commit is contained in:
Yoann Poupart
2023-05-30 20:57:04 +02:00
committed by GitHub
parent e09afb4b44
commit c1807d8408
2 changed files with 33 additions and 6 deletions

View File

@@ -26,7 +26,8 @@ def test_huggingface_embedding_query() -> None:
def test_huggingface_instructor_embedding_documents() -> None:
"""Test huggingface embeddings."""
documents = ["foo bar"]
embedding = HuggingFaceInstructEmbeddings()
model_name = "hkunlp/instructor-base"
embedding = HuggingFaceInstructEmbeddings(model_name=model_name)
output = embedding.embed_documents(documents)
assert len(output) == 1
assert len(output[0]) == 768
@@ -35,6 +36,22 @@ def test_huggingface_instructor_embedding_documents() -> None:
def test_huggingface_instructor_embedding_query() -> None:
"""Test huggingface embeddings."""
query = "foo bar"
embedding = HuggingFaceInstructEmbeddings()
model_name = "hkunlp/instructor-base"
embedding = HuggingFaceInstructEmbeddings(model_name=model_name)
output = embedding.embed_query(query)
assert len(output) == 768
def test_huggingface_instructor_embedding_normalize() -> None:
"""Test huggingface embeddings."""
query = "foo bar"
model_name = "hkunlp/instructor-base"
encode_kwargs = {"normalize_embeddings": True}
embedding = HuggingFaceInstructEmbeddings(
model_name=model_name, encode_kwargs=encode_kwargs
)
output = embedding.embed_query(query)
assert len(output) == 768
eps = 1e-5
norm = sum([o**2 for o in output])
assert abs(1 - norm) <= eps