Marqo Ecommerce Embeddings L
| Property | Value |
|---|---|
| Parameter Count | 652M |
| Embedding Dimension | 1024 |
| License | Apache 2.0 |
| Author | Marqo |
What is marqo-ecommerce-embeddings-L?
Marqo-ecommerce-embeddings-L is a cutting-edge embedding model specifically designed for ecommerce applications. It represents a significant advancement in multimodal retrieval technology, demonstrating superior performance with a 17.6% improvement in MRR and 20.5% in nDCG@10 compared to existing solutions.
Implementation Details
Built on OpenCLIP architecture, this model generates high-quality embeddings for both product images and text descriptions. The model supports both image-to-text and text-to-image retrieval tasks, making it particularly valuable for ecommerce search applications.
- 652M parameters for robust feature extraction
- 1024-dimensional embedding space
- Supports both Transformers and OpenCLIP implementations
- Optimized for batch processing with support for up to 1024 batch size
Core Capabilities
- Product image and text embedding generation
- Cross-modal retrieval between images and text
- Superior performance on Google Shopping and Amazon Products datasets
- Batch processing support for efficient large-scale operations
Frequently Asked Questions
Q: What makes this model unique?
The model achieves state-of-the-art performance in ecommerce search tasks, with significant improvements over both open-source and private models like Amazon-Titan-Multimodal. It's specifically optimized for product search and retrieval scenarios.
Q: What are the recommended use cases?
The model excels in ecommerce applications including product search, recommendation systems, and catalog organization. It's particularly effective for text-to-image and category-to-image retrieval tasks in large-scale product databases.
