jrmccluskey commented on code in PR #34700: URL: https://github.com/apache/beam/pull/34700#discussion_r2075667094
########## sdks/python/apache_beam/ml/transforms/embeddings/open_ai.py: ########## @@ -0,0 +1,204 @@ +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from collections.abc import Iterable +from collections.abc import Sequence +from typing import Any +from typing import Optional +from typing import TypeVar +from typing import Union +from typing import cast + +import apache_beam as beam +import openai +from apache_beam.ml.inference.base import RemoteModelHandler +from apache_beam.ml.inference.base import RunInference +from apache_beam.ml.transforms.base import EmbeddingsManager +from apache_beam.ml.transforms.base import _TextEmbeddingHandler +from apache_beam.pvalue import PCollection +from apache_beam.pvalue import Row +from openai import APIError +from openai import RateLimitError + +__all__ = ["OpenAITextEmbeddings"] + +# Define a type variable for the output +MLTransformOutputT = TypeVar('MLTransformOutputT') + +_BATCH_SIZE = 20 + +LOGGER = logging.getLogger("OpenAIEmbeddings") + + +def _retry_on_appropriate_openai_error(exception): # pylint: disable=line-too-long + """ + Retry filter that returns True for rate limit (429) or server (5xx) errors. + + Args: + exception: the returned exception encountered during the request/response + loop. + + Returns: + boolean indication whether or not the exception is a Server Error (5xx) or + a RateLimitError (429) error. + """ + return isinstance(exception, (RateLimitError, APIError)) + + +class _OpenAITextEmbeddingHandler(RemoteModelHandler): + """ + Note: Intended for internal use and guarantees no backwards compatibility. + """ + def __init__( + self, + model_name: str, + api_key: Optional[str] = None, + organization: Optional[str] = None, + dimensions: Optional[int] = None, + user: Optional[str] = None, + batch_size: Optional[int] = None, + ): + super().__init__( + namespace="OpenAITextEmbeddings", + num_retries=5, + throttle_delay_secs=5, + retry_filter=_retry_on_appropriate_openai_error) + self.model_name = model_name + self.api_key = api_key + self.organization = organization + self.dimensions = dimensions + self.user = user + self.batch_size = batch_size or _BATCH_SIZE + + def create_client(self): + """Creates and returns an OpenAI client.""" + if self.api_key: + client = openai.OpenAI( + api_key=self.api_key, + organization=self.organization, + ) + else: + client = openai.OpenAI(organization=self.organization) + + return client + + def request( + self, + batch: Sequence[str], + model: Any, + inference_args: Optional[dict[str, Any]] = None, + ) -> Iterable: + """Makes a request to OpenAI embedding API and returns embeddings.""" + # Process in smaller batches if needed + if len(batch) > self.batch_size: + embeddings = [] + for i in range(0, len(batch), self.batch_size): + text_batch = batch[i:i + self.batch_size] + # Use request() recursively for each smaller batch + embeddings_batch = self.request(text_batch, model, inference_args) + embeddings.extend(embeddings_batch) + return embeddings + + # Prepare arguments for the API call + kwargs = { + "model": self.model_name, + "input": batch, + } + if self.dimensions: + kwargs["dimensions"] = cast(Any, self.dimensions) + if self.user: + kwargs["user"] = self.user + + try: + # Make the API call + response = model.embeddings.create(**kwargs) + return [item.embedding for item in response.data] + except RateLimitError as e: + LOGGER.warning("Request was rate limited by OpenAI API: %s", e) + raise + except Exception as e: + LOGGER.error("Unexpected exception raised as part of request: %s", e) + raise Review Comment: nit: the RemoteModelHandler framework handles the exception handling for you here minus the specific logging messages for different exceptions. You can remove this if you want, although you will lose the logging ########## sdks/python/apache_beam/ml/transforms/embeddings/open_ai.py: ########## @@ -0,0 +1,204 @@ +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from collections.abc import Iterable +from collections.abc import Sequence +from typing import Any +from typing import Optional +from typing import TypeVar +from typing import Union +from typing import cast + +import apache_beam as beam +import openai +from apache_beam.ml.inference.base import RemoteModelHandler +from apache_beam.ml.inference.base import RunInference +from apache_beam.ml.transforms.base import EmbeddingsManager +from apache_beam.ml.transforms.base import _TextEmbeddingHandler +from apache_beam.pvalue import PCollection +from apache_beam.pvalue import Row +from openai import APIError +from openai import RateLimitError + +__all__ = ["OpenAITextEmbeddings"] + +# Define a type variable for the output +MLTransformOutputT = TypeVar('MLTransformOutputT') + +_BATCH_SIZE = 20 + +LOGGER = logging.getLogger("OpenAIEmbeddings") + + +def _retry_on_appropriate_openai_error(exception): # pylint: disable=line-too-long + """ + Retry filter that returns True for rate limit (429) or server (5xx) errors. + + Args: + exception: the returned exception encountered during the request/response + loop. + + Returns: + boolean indication whether or not the exception is a Server Error (5xx) or + a RateLimitError (429) error. + """ + return isinstance(exception, (RateLimitError, APIError)) + + +class _OpenAITextEmbeddingHandler(RemoteModelHandler): + """ + Note: Intended for internal use and guarantees no backwards compatibility. + """ + def __init__( + self, + model_name: str, + api_key: Optional[str] = None, + organization: Optional[str] = None, + dimensions: Optional[int] = None, + user: Optional[str] = None, + batch_size: Optional[int] = None, + ): + super().__init__( + namespace="OpenAITextEmbeddings", + num_retries=5, + throttle_delay_secs=5, + retry_filter=_retry_on_appropriate_openai_error) + self.model_name = model_name + self.api_key = api_key + self.organization = organization + self.dimensions = dimensions + self.user = user + self.batch_size = batch_size or _BATCH_SIZE + + def create_client(self): + """Creates and returns an OpenAI client.""" + if self.api_key: + client = openai.OpenAI( + api_key=self.api_key, + organization=self.organization, + ) + else: + client = openai.OpenAI(organization=self.organization) + + return client + + def request( + self, + batch: Sequence[str], + model: Any, + inference_args: Optional[dict[str, Any]] = None, + ) -> Iterable: + """Makes a request to OpenAI embedding API and returns embeddings.""" + # Process in smaller batches if needed + if len(batch) > self.batch_size: + embeddings = [] + for i in range(0, len(batch), self.batch_size): + text_batch = batch[i:i + self.batch_size] + # Use request() recursively for each smaller batch + embeddings_batch = self.request(text_batch, model, inference_args) + embeddings.extend(embeddings_batch) + return embeddings + + # Prepare arguments for the API call + kwargs = { + "model": self.model_name, + "input": batch, + } + if self.dimensions: + kwargs["dimensions"] = cast(Any, self.dimensions) + if self.user: + kwargs["user"] = self.user + + try: + # Make the API call + response = model.embeddings.create(**kwargs) + return [item.embedding for item in response.data] + except RateLimitError as e: + LOGGER.warning("Request was rate limited by OpenAI API: %s", e) + raise + except Exception as e: + LOGGER.error("Unexpected exception raised as part of request: %s", e) + raise + + def batch_elements_kwargs(self) -> dict[str, Any]: + """Returns kwargs suitable for beam.BatchElements with appropriate batch size.""" # pylint: disable=line-too-long + return {'max_batch_size': self.batch_size} + + def __repr__(self): + return 'OpenAITextEmbeddings' + + +class OpenAITextEmbeddings(EmbeddingsManager): + """ + A PTransform that uses OpenAI's API to generate embeddings from text inputs. + + Example Usage:: + + with pipeline as p: # pylint: disable=line-too-long + text = p | "Create texts" >> beam.Create([{"text": "Hello world"}, + {"text": "Beam ML"}]) + embeddings = text | OpenAITextEmbeddings( + model_name="text-embedding-3-small", + columns=["embedding_col"], + api_key=api_key + ) + """ + @beam.typehints.with_output_types(PCollection[Union[MLTransformOutputT, Row]]) # pylint: disable=line-too-long + def __init__( + self, + model_name: str, + columns: list[str], + api_key: Optional[str] = None, + organization: Optional[str] = None, + dimensions: Optional[int] = None, + user: Optional[str] = None, + batch_size: Optional[int] = None, Review Comment: this is misleading since you're setting a single batch size value then take it as the max. I'd recommend exposing the min and max batch sizes separately (or just taking them as kwargs) ########## sdks/python/apache_beam/ml/transforms/embeddings/open_ai.py: ########## @@ -0,0 +1,204 @@ +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from collections.abc import Iterable +from collections.abc import Sequence +from typing import Any +from typing import Optional +from typing import TypeVar +from typing import Union +from typing import cast + +import apache_beam as beam +import openai +from apache_beam.ml.inference.base import RemoteModelHandler +from apache_beam.ml.inference.base import RunInference +from apache_beam.ml.transforms.base import EmbeddingsManager +from apache_beam.ml.transforms.base import _TextEmbeddingHandler +from apache_beam.pvalue import PCollection +from apache_beam.pvalue import Row +from openai import APIError +from openai import RateLimitError + +__all__ = ["OpenAITextEmbeddings"] + +# Define a type variable for the output +MLTransformOutputT = TypeVar('MLTransformOutputT') + +_BATCH_SIZE = 20 + +LOGGER = logging.getLogger("OpenAIEmbeddings") + + +def _retry_on_appropriate_openai_error(exception): # pylint: disable=line-too-long + """ + Retry filter that returns True for rate limit (429) or server (5xx) errors. + + Args: + exception: the returned exception encountered during the request/response + loop. + + Returns: + boolean indication whether or not the exception is a Server Error (5xx) or + a RateLimitError (429) error. + """ + return isinstance(exception, (RateLimitError, APIError)) + + +class _OpenAITextEmbeddingHandler(RemoteModelHandler): + """ + Note: Intended for internal use and guarantees no backwards compatibility. + """ + def __init__( + self, + model_name: str, + api_key: Optional[str] = None, + organization: Optional[str] = None, + dimensions: Optional[int] = None, + user: Optional[str] = None, + batch_size: Optional[int] = None, + ): + super().__init__( + namespace="OpenAITextEmbeddings", + num_retries=5, + throttle_delay_secs=5, + retry_filter=_retry_on_appropriate_openai_error) + self.model_name = model_name + self.api_key = api_key + self.organization = organization + self.dimensions = dimensions + self.user = user + self.batch_size = batch_size or _BATCH_SIZE + + def create_client(self): + """Creates and returns an OpenAI client.""" + if self.api_key: + client = openai.OpenAI( + api_key=self.api_key, + organization=self.organization, + ) + else: + client = openai.OpenAI(organization=self.organization) + + return client + + def request( + self, + batch: Sequence[str], + model: Any, + inference_args: Optional[dict[str, Any]] = None, + ) -> Iterable: + """Makes a request to OpenAI embedding API and returns embeddings.""" + # Process in smaller batches if needed + if len(batch) > self.batch_size: + embeddings = [] + for i in range(0, len(batch), self.batch_size): + text_batch = batch[i:i + self.batch_size] + # Use request() recursively for each smaller batch + embeddings_batch = self.request(text_batch, model, inference_args) + embeddings.extend(embeddings_batch) + return embeddings + + # Prepare arguments for the API call + kwargs = { + "model": self.model_name, + "input": batch, + } + if self.dimensions: + kwargs["dimensions"] = cast(Any, self.dimensions) Review Comment: Is there an oddity in the type checking for the dimensions kwarg when you pass it to `create()`? The manual cast to Any feels unnecessary ########## sdks/python/apache_beam/ml/transforms/embeddings/open_ai.py: ########## @@ -0,0 +1,204 @@ +# Licensed to the Apache Software Foundation (ASF) under one or more +# contributor license agreements. See the NOTICE file distributed with +# this work for additional information regarding copyright ownership. +# The ASF licenses this file to You under the Apache License, Version 2.0 +# (the "License"); you may not use this file except in compliance with +# the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +# See the License for the specific language governing permissions and +# limitations under the License. + +import logging +from collections.abc import Iterable +from collections.abc import Sequence +from typing import Any +from typing import Optional +from typing import TypeVar +from typing import Union +from typing import cast + +import apache_beam as beam +import openai +from apache_beam.ml.inference.base import RemoteModelHandler +from apache_beam.ml.inference.base import RunInference +from apache_beam.ml.transforms.base import EmbeddingsManager +from apache_beam.ml.transforms.base import _TextEmbeddingHandler +from apache_beam.pvalue import PCollection +from apache_beam.pvalue import Row +from openai import APIError +from openai import RateLimitError + +__all__ = ["OpenAITextEmbeddings"] + +# Define a type variable for the output +MLTransformOutputT = TypeVar('MLTransformOutputT') + +_BATCH_SIZE = 20 + +LOGGER = logging.getLogger("OpenAIEmbeddings") + + +def _retry_on_appropriate_openai_error(exception): # pylint: disable=line-too-long + """ + Retry filter that returns True for rate limit (429) or server (5xx) errors. + + Args: + exception: the returned exception encountered during the request/response + loop. + + Returns: + boolean indication whether or not the exception is a Server Error (5xx) or + a RateLimitError (429) error. + """ + return isinstance(exception, (RateLimitError, APIError)) + + +class _OpenAITextEmbeddingHandler(RemoteModelHandler): + """ + Note: Intended for internal use and guarantees no backwards compatibility. + """ + def __init__( + self, + model_name: str, + api_key: Optional[str] = None, + organization: Optional[str] = None, + dimensions: Optional[int] = None, + user: Optional[str] = None, + batch_size: Optional[int] = None, + ): + super().__init__( + namespace="OpenAITextEmbeddings", + num_retries=5, + throttle_delay_secs=5, + retry_filter=_retry_on_appropriate_openai_error) + self.model_name = model_name + self.api_key = api_key + self.organization = organization + self.dimensions = dimensions + self.user = user + self.batch_size = batch_size or _BATCH_SIZE + + def create_client(self): + """Creates and returns an OpenAI client.""" + if self.api_key: + client = openai.OpenAI( + api_key=self.api_key, + organization=self.organization, + ) + else: + client = openai.OpenAI(organization=self.organization) + + return client + + def request( + self, + batch: Sequence[str], + model: Any, + inference_args: Optional[dict[str, Any]] = None, + ) -> Iterable: + """Makes a request to OpenAI embedding API and returns embeddings.""" + # Process in smaller batches if needed + if len(batch) > self.batch_size: + embeddings = [] + for i in range(0, len(batch), self.batch_size): + text_batch = batch[i:i + self.batch_size] + # Use request() recursively for each smaller batch + embeddings_batch = self.request(text_batch, model, inference_args) + embeddings.extend(embeddings_batch) + return embeddings Review Comment: I don't love the recursive call here conceptually. I'd rather see batches formed in the try block and have every call to the model be done that way; however, this situation should not happen anyway given that the batching parameters are given to a BatchElements call before this code is executed anyway. Barring situations where batching simply cannot work (like with Vertex AI endpoints) you're better off not thinking about manipulating the batches in the RunInference code -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: github-unsubscr...@beam.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
