Bge-m3-567m Model | MAX Builds

Run this model

Install our magic package manager:
```
curl -ssL https://magic.modular.com/ | bash
```
Then run the source command that's printed in your terminal.
Install Max Pipelines in order to run this model.
```
magic global install max-pipelines
```

Start a local endpoint for bge-m3/567m:

max-serve serve --huggingface-repo-id BAAI/bge-m3

The endpoint is ready when you see the URI printed in your terminal:

Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)

Now open another terminal to send a request using curl:

curl -N http://0.0.0.0:8000/v1/chat/completions -H "Content-Type: application/json" -d '{
    "model": "bge-m3/567m",
    "stream": true,
    "messages": [
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Who won the World Series in 2020?"}
    ]
}' | grep -o '"content":"[^"]*"' | sed 's/"content":"//g' | sed 's/"//g' | tr -d '
' | sed 's/\n/
/g'

🎉 Hooray! You’re running Generative AI. Our goal is to make this as easy as possible.

About

BGE-M3 is a cutting-edge model based on the XLM-RoBERTa architecture, designed for unparalleled versatility across three key dimensions: Multi-Functionality, Multi-Linguality, and Multi-Granularity.

Multi-Functionality: BGE-M3 excels in three core retrieval tasks seamlessly: dense retrieval, multi-vector retrieval, and sparse retrieval, making it highly adaptable for diverse retrieval scenarios.
Multi-Linguality: The model supports over 100 languages, ensuring robust performance across a wide range of linguistic contexts.
Multi-Granularity: It effectively processes inputs from short sentences to extended documents containing up to 8192 tokens, demonstrating flexibility across different input granularities.

BGE-M3 achieves these capabilities through self-knowledge distillation, efficiently leveraging the underlying strength of XLM-RoBERTa while enhancing flexibility and performance. The model’s exceptional benchmarks demonstrate its impact as a powerful solution for multi-dimensional text representation, catering to various retrieval tasks and languages with high precision.

Benchmarks from the open-source community

@misc{bge-m3,
      title={BGE M3-Embedding: Multi-Lingual, Multi-Functionality, Multi-Granularity Text Embeddings Through Self-Knowledge Distillation},
      author={Jianlv Chen and Shitao Xiao and Peitian Zhang and Kun Luo and Defu Lian and Zheng Liu},
      year={2024},
      eprint={2402.03216},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}

bert.attention.causal	false
bert.attention.head_count	16
bert.attention.layer_norm_epsilon	1e-05
bert.block_count	24
bert.context_length	8192
bert.embedding_length	1024
bert.feed_forward_length	4096
bert.pooling_type	2
general.architecture	bert
general.file_type	1
general.license	mit
general.quantization_version	2
general.size_label	567M
general.tags	[sentence-transformers, feature-extraction, sentence-similarity, sentence-similarity]
general.type	model
tokenizer.ggml.add_bos_token	true
tokenizer.ggml.add_eos_token	true
tokenizer.ggml.add_space_prefix	true
tokenizer.ggml.bos_token_id	0
tokenizer.ggml.cls_token_id	0
tokenizer.ggml.eos_token_id	2
tokenizer.ggml.mask_token_id	250001
tokenizer.ggml.model	t5
tokenizer.ggml.padding_token_id	1
tokenizer.ggml.pre	default
tokenizer.ggml.precompiled_charsmap	[0, 180, 2, 0, 0, ...]
tokenizer.ggml.remove_extra_whitespaces	true
tokenizer.ggml.scores	[0, 0, 0, 0, -3.4635427, ...]
tokenizer.ggml.seperator_token_id	2
tokenizer.ggml.token_type	[3, 3, 3, 2, 1, ...]
tokenizer.ggml.token_type_count	1
tokenizer.ggml.tokens	[<s>, <pad>, </s>, <unk>, ,, ...]
tokenizer.ggml.unknown_token_id	3

MIT License

Copyright (c) [year] [fullname]

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

bge-m3-567m

Run this model

About