ai/mxbai-embed-large Docker 镜像 - 轩辕镜像

mxbai-embed-large-v1

!logo

mxbai-embed-large-v1 is a state-of-the-art English language embedding model developed by Mixedbread AI. It converts text into dense vector representations, capturing the semantic essence of the input. Trained on a vast dataset exceeding 700 million pairs using contrastive training methods and fine-tuned on over 30 million high-quality triplets with the AnglE loss function, this model adapts to a wide range of topics and domains, making it suitable for various real-world applications and Retrieval-Augmented Generation (RAG) use cases.

Intended uses

mxbai-embed-large-v1 is designed for generating sentence embeddings suitable for various NLP applications.

• SemanticsSearch and information retrieval: Specifically designed for RAG, this model enhances search systems by providing relevant document embeddings, improving the accuracy and relevance of search results.
• Semantic textual similarity: Measures the similarity between sentences, aiding in tasks such as clustering, duplicate detection, and paraphrase identification.
• Text classification: Serves as input features for classifiers in tasks like sentiment analysis, topic categorization, and intent detection.

Characteristics

Available model variants

¹: VRAM estimated based on model characteristics.

latest → 335M-F16

Use this AI model with Docker Model Runner

First, pull the model:

bash
docker model pull ai/mxbai-embed-large

Then run the model:

bash
docker model run ai/mxbai-embed-large

For more information on Docker Model Runner, explore the documentation.

***ations

• Prompt usage: For retrieval tasks, prepend the query with the prompt. For example, "Represent this sentence for searching relevant passages:". This practice helps the model understand the context and improves performance. For other tasks, the text can be used as-is without any additional prompt.
• Language limitation: The model is trained exclusively on English text and is specifically designed for the English language.
• Sequence length: The suggested maximum sequence length is 512 tokens. Longer sequences may be truncated, leading to a loss of information.

Benchmark performance

Links

• Open Source Strikes Bread - New Fluffy Embedding Model
• Mixelbread Docs:mxbai-embed-large-v1