Embedding#

openai#

class gptcache.embedding.openai.OpenAI(model: str = 'text-embedding-ada-002', api_key: Optional[str] = None, api_base: Optional[str] = None)[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate text embedding for given text using OpenAI.

Parameters

  • model (str) – model name, defaults to ‘text-embedding-ada-002’.

  • api_key (str) – OpenAI API Key. When the parameter is not specified, it will load the key by default if it is available.

Example

from gptcache.embedding import OpenAI

test_sentence = 'Hello, world.'
encoder = OpenAI(api_key='your_openai_key')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

property dimension#

Embedding dimension.

Returns

embedding dimension

static dim_dict()[source]#

cohere#

class gptcache.embedding.cohere.Cohere(model: str = 'large', api_key: Optional[str] = None)[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate text embedding for given text using Cohere.

Parameters

  • model (str) – model name (size), defaults to ‘large’.

  • api_key (str) – Cohere API Key.

Example

from gptcache.embedding import Cohere

test_sentence = 'Hello, world.'
encoder = Cohere(model='small', api_key='your_cohere_key')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

property dimension#

Embedding dimension.

Returns

embedding dimension

static dim_dict()[source]#

sbert#

class gptcache.embedding.sbert.SBERT(model: str = 'all-MiniLM-L6-v2')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate sentence embedding for given text using pretrained models of Sentence Transformers.

Parameters

model (str) – model name, defaults to ‘all-MiniLM-L6-v2’.

Example

from gptcache.embedding import SBERT

test_sentence = 'Hello, world.'
encoder = SBERT('all-MiniLM-L6-v2')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

property dimension#

Embedding dimension.

Returns

embedding dimension

fasttext#

class gptcache.embedding.fasttext.FastText(model: str = 'en', dim: Optional[int] = None)[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate sentence embedding for given text using pretrained models of different languages from fastText.

Parameters

  • model (str) – model name, defaults to ‘en’.

  • dim (int) – reduced dimension of embedding. If this parameter is not provided, the embedding dimension (300) will not change.

Example

from gptcache.embedding import FastText

test_sentence = 'Hello, world.'
encoder = FastText(model='en', dim=100)
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

property dimension#

Embedding dimension.

Returns

embedding dimension

base#

class gptcache.embedding.base.BaseEmbedding[source]#

Bases: object

Base Embedding interface.

abstract to_embeddings(data, **kwargs)[source]#
abstract property dimension: int#

string#

gptcache.embedding.string.to_embeddings(data, **_)[source]#

Nothing to do, return the origin data

vit#

class gptcache.embedding.vit.ViT(model: str = 'google/vit-base-patch16-384')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate sentence embedding for given text using pretrained models from Huggingface transformers.

Parameters

model (str) – model name, defaults to ‘google/vit-base-patch16-384’.

Example

import io
from PIL import Image
from gptcache.embedding import ImageEmbedding

def prepare_image(image_data: str = None):
    if not image_data:
        image_data = io.BytesIO()
        Image.new('RGB', (244, 244), color=(255, 0, 0)).save(image_data, format='JPEG')
        image_data.seek(0)
    image = Image.open(image_data)
    return image

image = prepare_image()
encoder = ImageEmbeddings(model="google/vit-base-patch16-384")
embed = encoder.to_embeddings(image)
to_embeddings(data, **__)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

preprocess(data)[source]#
property dimension#

Embedding dimension.

Returns

embedding dimension

langchain#

class gptcache.embedding.langchain.LangChain(embeddings: langchain.embeddings.base.Embeddings, dimension: int = 0)[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate text embedding for given text using LangChain

Parameters

  • embeddings (Embeddings) – the LangChain Embeddings object.

  • dimension (int) – The vector dimension after embedding is calculated by calling embed once by default. If you confirm the dimension, you can assign a value to this parameter to reduce this request.

Example

from gptcache.embedding import LangChain
from langchain.embeddings.openai import OpenAIEmbeddings

test_sentence = 'Hello, world.'
embeddings = OpenAIEmbeddings(model="your-embeddings-deployment-name")
encoder = LangChain(embeddings=embeddings)
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **kwargs)[source]#
property dimension: int#

huggingface#

class gptcache.embedding.huggingface.Huggingface(model: str = 'distilbert-base-uncased')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate sentence embedding for given text using pretrained models from Huggingface transformers.

Parameters

model (str) – model name, defaults to ‘distilbert-base-uncased’.

Example

from gptcache.embedding import Huggingface

test_sentence = 'Hello, world.'
encoder = Huggingface(model='distilbert-base-uncased')
embed = encoder.to_embeddings(test_sentence)

test_sentence = '什么是Github'
huggingface = Huggingface(model='uer/albert-base-chinese-cluecorpussmall')
embed = huggingface.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

post_proc(token_embeddings, inputs)[source]#
property dimension#

Embedding dimension.

Returns

embedding dimension

data2vec#

class gptcache.embedding.data2vec.Data2VecAudio(model_name='facebook/data2vec-audio-base-960h')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate audio embedding for given audio using pretrained models from Data2Vec.

Parameters

model (str) – model name, defaults to ‘facebook/data2vec-audio-base-960h’.

Example

from gptcache.embedding import Data2VecAudio

audio_file = 'test.wav'
encoder = Data2VecAudio(model='facebook/data2vec-audio-base-960h')
embed = encoder.to_embeddings(audio_file)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – path to audio file.

Returns

a text embedding in shape of (dim,).

load_audio(audio_path, target_sr)[source]#
property dimension#

Embedding dimension.

Returns

embedding dimension

uform#

class gptcache.embedding.uform.UForm(model: Union[str, uform.TritonClient] = 'unum-cloud/uform-vl-english', embedding_type: str = 'text')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate multi-modal embeddings using pretrained models from UForm.

Parameters

  • model (str) – model name, defaults to ‘unum-cloud/uform-vl-english’.

  • embedding_type (str) – type of embedding, defaults to ‘text’. options: text, image

Example

from gptcache.embedding import UForm

test_sentence = 'Hello, world.'
encoder = UForm(model='unum-cloud/uform-vl-english')
embed = encoder.to_embeddings(test_sentence)

test_sentence = '什么是Github'
encoder = UForm(model='unum-cloud/uform-vl-multilingual')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data: Any, **_)[source]#

Generate embedding given text input or a path to a file.

Parameters

data (str) – text in string, or a path to an image file.

Returns

an embedding in shape of (dim,).

property dimension#

Embedding dimension.

Returns

embedding dimension

onnx#

class gptcache.embedding.onnx.Onnx(model='GPTCache/paraphrase-albert-onnx')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate text embedding for given text using ONNX Model.

Example

from gptcache.embedding import Onnx

test_sentence = 'Hello, world.'
encoder = Onnx(model='GPTCache/paraphrase-albert-onnx')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input.

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

post_proc(token_embeddings, attention_mask)[source]#
property dimension#

Embedding dimension.

Returns

embedding dimension

rwkv#

class gptcache.embedding.rwkv.Rwkv(model: str = 'sgugger/rwkv-430M-pile')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate sentence embedding for given text using RWKV models.

Parameters

model (str) – model name, defaults to ‘sgugger/rwkv-430M-pile’. Check https://huggingface.co/docs/transformers/model_doc/rwkv for more avaliable models.

Example

from gptcache.embedding import Rwkv

test_sentence = 'Hello, world.'
encoder = Rwkv(model='sgugger/rwkv-430M-pile')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

property dimension#

Embedding dimension.

Returns

embedding dimension

timm#

class gptcache.embedding.timm.Timm(model: str = 'resnet18', device: str = 'default')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate image embedding for given image using pretrained models from Timm.

Parameters

model (str) – model name, defaults to ‘resnet34’.

Example

import requests
from io import BytesIO
from gptcache.embedding import Timm


encoder = Timm(model='resnet50')
embed = encoder.to_embeddings('path/to/image')
to_embeddings(data, skip_preprocess: bool = False, **_)[source]#

Generate embedding given image data

Parameters

  • data (str) – image path.

  • skip_preprocess (bool) – flag to skip preprocess, defaults to False, enable this if the input data is torch.tensor.

Returns

an image embedding in shape of (dim,).

post_proc(features)[source]#
preprocess(image_path)[source]#

Load image from path and then transform image to torch.tensor with model transformations.

Parameters

image_path (str) – image path.

Returns

an image tensor (without batch size).

property dimension#

Embedding dimension.

Returns

embedding dimension

paddlenlp#

class gptcache.embedding.paddlenlp.PaddleNLP(model: str = 'ernie-3.0-medium-zh')[source]#

Bases: gptcache.embedding.base.BaseEmbedding

Generate sentence embedding for given text using pretrained models from PaddleNLP transformers.

Parameters

model (str) – model name, defaults to ‘ernie-3.0-medium-zh’.

Example

from gptcache.embedding import PaddleNLP

test_sentence = 'Hello, world.'
encoder = PaddleNLP(model='ernie-3.0-medium-zh')
embed = encoder.to_embeddings(test_sentence)
to_embeddings(data, **_)[source]#

Generate embedding given text input

Parameters

data (str) – text in string.

Returns

a text embedding in shape of (dim,).

post_proc(token_embeddings, inputs)[source]#
property dimension#

Embedding dimension.

Returns

embedding dimension