NSQL-Llama-2-7B Sharded Version

This model, czurita/nsql-llama-2-7B-sharded-bf16-2GB, is a sharded, bfloat16 version of the original NumbersStation/nsql-llama-2-7B model. It is specifically reconfigured for environments with limited RAM, such as Google Colab or Kaggle, while retaining its core capabilities.

Key Capabilities

• SQL Generation: Designed to convert natural language questions and provided database schemas into SQL queries.
• Llama-2 Base: Built upon Meta's Llama-2 7B architecture, providing a strong foundation for language understanding.
• Specialized Training: Pre-trained on 1 million SQL queries from The Stack and fine-tuned on over 20 public text-to-SQL datasets.
• Evaluation: Performance is evaluated on standard text-to-SQL benchmarks like Spider and GeoQuery.
• Optimized Output: Best suited for generating SELECT queries based on a defined prompt format.

Good For

• Text-to-SQL Applications: Ideal for developers building applications that require converting natural language into executable SQL.
• Low-Resource Environments: Its sharded, bfloat16 format makes it suitable for deployment and experimentation in environments with constrained memory.
• Database Interaction: Facilitating easier interaction with databases by automating SQL query generation from user prompts.