openai/openai-python
Publicmirrored from https://github.com/openai/openai-pythonAvailable
openai/api_resources/embedding.py
58lines · modecode
| 1 | import base64 |
| 2 | import time |
| 3 | |
| 4 | import numpy as np |
| 5 | |
| 6 | from openai import util |
| 7 | from openai.api_resources.abstract import DeletableAPIResource, ListableAPIResource |
| 8 | from openai.api_resources.abstract.engine_api_resource import EngineAPIResource |
| 9 | from openai.error import InvalidRequestError, TryAgain |
| 10 | |
| 11 | |
| 12 | class Embedding(EngineAPIResource, ListableAPIResource, DeletableAPIResource): |
| 13 | engine_required = False |
| 14 | OBJECT_NAME = "embeddings" |
| 15 | |
| 16 | @classmethod |
| 17 | def create(cls, *args, **kwargs): |
| 18 | """ |
| 19 | Creates a new embedding for the provided input and parameters. |
| 20 | |
| 21 | See https://beta.openai.com/docs/api-reference/embeddings for a list |
| 22 | of valid parameters. |
| 23 | """ |
| 24 | start = time.time() |
| 25 | timeout = kwargs.pop("timeout", None) |
| 26 | if kwargs.get("model", None) is None and kwargs.get("engine", None) is None: |
| 27 | raise InvalidRequestError( |
| 28 | "Must provide an 'engine' or 'model' parameter to create an Embedding.", |
| 29 | param="engine", |
| 30 | ) |
| 31 | |
| 32 | user_provided_encoding_format = kwargs.get("encoding_format", None) |
| 33 | |
| 34 | # If encoding format was not explicitly specified, we opaquely use base64 for performance |
| 35 | if not user_provided_encoding_format: |
| 36 | kwargs["encoding_format"] = "base64" |
| 37 | |
| 38 | while True: |
| 39 | try: |
| 40 | response = super().create(*args, **kwargs) |
| 41 | |
| 42 | # If a user specifies base64, we'll just return the encoded string. |
| 43 | # This is only for the default case. |
| 44 | if not user_provided_encoding_format: |
| 45 | for data in response.data: |
| 46 | |
| 47 | # If an engine isn't using this optimization, don't do anything |
| 48 | if type(data["embedding"]) == str: |
| 49 | data["embedding"] = np.frombuffer( |
| 50 | base64.b64decode(data["embedding"]), dtype="float32" |
| 51 | ).tolist() |
| 52 | |
| 53 | return response |
| 54 | except TryAgain as e: |
| 55 | if timeout is not None and time.time() > start + timeout: |
| 56 | raise |
| 57 | |
| 58 | util.log_info("Waiting for model to warm up", error=e) |