Let's load the SageMaker Endpoints Embeddings class. The class can be used if you host, e.g. your own Hugging Face model on SageMaker.
Note: In order to handle batched requests, you will need to adjust the return line in the predict_fn() function within the custom inference.py script:
return {"vectors": sentence_embeddings.tolist()}.
import json
from typing import Dict, List
from langchain_community.embeddings import SagemakerEndpointEmbeddings
from langchain_community.embeddings.sagemaker_endpoint import EmbeddingsContentHandler
class ContentHandler(EmbeddingsContentHandler):
content_type = "application/json"
accepts = "application/json"
def transform_input(self, inputs: list[str], model_kwargs: Dict) -> bytes:
"""
Transforms the input into bytes that can be consumed by SageMaker endpoint.
Args:
inputs: List of input strings.
model_kwargs: Additional keyword arguments to be passed to the endpoint.
Returns:
The transformed bytes input.
"""
input_str = json.dumps({"inputs": inputs, **model_kwargs})
return input_str.encode("utf-8")
def transform_output(self, output: bytes) -> List[List[float]]:
"""
Transforms the bytes output from the endpoint into a list of embeddings.
Args:
output: The bytes output from SageMaker endpoint.
Returns:
The transformed output - list of embeddings
Note:
The length of the outer list is the number of input strings.
The length of the inner lists is the embedding dimension.
"""
response_json = json.loads(output.read().decode("utf-8"))
return response_json["vectors"]
content_handler = ContentHandler()
embeddings = SagemakerEndpointEmbeddings(
endpoint_name="huggingface-pytorch-inference-2023-03-21-16-14-03-834",
region_name="us-east-1",
content_handler=content_handler,
)
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4