BGE-M3 in HuggingFace Transformer

This is not an official implementation of BGE-M3. Official implementation can be found in Flag Embedding project.

Introduction

Full introduction please see the github repo.

https://github.com/liuyanyi/transformers-bge-m3

Use BGE-M3 in HuggingFace Transformer

from transformers import AutoModel, AutoTokenizer

# Trust remote code is required to load the model
tokenizer = AutoTokenizer.from_pretrained(model_path, trust_remote_code=True)
model = AutoModel.from_pretrained(model_path, trust_remote_code=True)

input_str = "Hello, world!"
input_ids = tokenizer(input_str, return_tensors="pt", padding=True, truncation=True)

output = model(**input_ids, return_dict=True)

dense_output = output.dense_output # To align with Flag Embedding project, a normalization is required
colbert_output = output.colbert_output # To align with Flag Embedding project, a normalization is required
sparse_output = output.sparse_output

References

Downloads last month
75
Safetensors
Model size
568M params
Tensor type
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.