README.md

metadata

tags:
  - sparse sparsity quantized onnx embeddings int8
license: mit
language:
  - en

gte-base-quant

This is the quantized (INT8) ONNX variant of the gte-base embeddings model created with DeepSparse Optimum for ONNX export/inference and Neural Magic's Sparsify for one-shot quantization.

Current list of sparse and quantized gte ONNX models:

Links	Sparsification Method
zeroshot/gte-large-sparse	Quantization (INT8) & 50% Pruning
zeroshot/gte-large-quant	Quantization (INT8)
zeroshot/gte-base-sparse	Quantization (INT8) & 50% Pruning
zeroshot/gte-base-quant	Quantization (INT8)
zeroshot/gte-small-sparse	Quantization (INT8) & 50% Pruning
zeroshot/gte-small-quant	Quantization (INT8)

pip install -U deepsparse-nightly[sentence_transformers]

from deepsparse.sentence_transformers import SentenceTransformer
model = SentenceTransformer('zeroshot/gte-base-quant', export=False)

# Our sentences we like to encode
sentences = ['This framework generates embeddings for each input sentence',
    'Sentences are passed as a list of string.',
    'The quick brown fox jumps over the lazy dog.']

# Sentences are encoded by calling model.encode()
embeddings = model.encode(sentences)

# Print the embeddings
for sentence, embedding in zip(sentences, embeddings):
    print("Sentence:", sentence)
    print("Embedding:", embedding.shape)
    print("")

For further details regarding DeepSparse & Sentence Transformers integration, refer to the DeepSparse README.

For general questions on these models and sparsification methods, reach out to the engineering team on our community Slack.