Model request: Snowflake's Arctic embed v2.0 and gte-Qwen2-1.5B

[lovewormcn]

Hi WebLLM team,

I would like to request adding two new models to the list of pre-built embedding models in WebLLM:

1. Snowflake's Arctic Embed v2.0

• https://huggingface.co/Snowflake/snowflake-arctic-embed-l-v2.0
• https://huggingface.co/Snowflake/snowflake-arctic-embed-m-v2.0

2. gte-Qwen2-1.5B-instruct

• https://huggingface.co/Alibaba-NLP/gte-Qwen2-1.5B-instruct

These additions would significantly benefit users of the WebLLM-Extension (used by SillyTavern ), as it directly depends on WebLLM's model registry for features like Vector Storage and Summarize.

Reasons for the request:

For Snowflake's Arctic Embed v2.0:

1. Performance & Scalability:
   Arctic Embed v2.0 is optimized for enterprise-grade throughput with a lightweight design (under 1B parameters), enabling faster and more efficient embeddings compared to current options. This would enhance SillyTavern's performance in tasks like semantic search and text summarization.

2. Multilingual Support:
   The model provides high-quality Chinese text embeddings, aligning with WebLLM's goal of supporting diverse language use cases (as highlighted in MTEB benchmarks).

For gte-Qwen2-1.5B-instruct:

1. Strong Chinese Language Support:
   The gte-Qwen2-1.5B-instruct model is specifically fine-tuned for Chinese language tasks and demonstrates excellent performance in generating high-quality embeddings for Chinese text. This makes it an ideal choice for applications targeting Chinese-speaking users.

2. Efficient Architecture:
   With 1.5B parameters, this model strikes a balance between performance and resource efficiency, making it suitable for deployment in environments with limited computational resources .

3. Compatibility with Qwen2 Series:
   As part of the Qwen2 series, this model inherits the robust capabilities of the Qwen2 architecture, including strong instruction-following abilities and multilingual support .

Thank you for considering this request! Let me know if further details are needed.