jastorj/snowflake_arctic_text2sql_r1_7b-nl2sqlpp-16bit-v5.1-cw-15K
The jastorj/snowflake_arctic_text2sql_r1_7b-nl2sqlpp-16bit-v5.1-cw-15K model is a 7.6 billion parameter language model fine-tuned from Snowflake/Arctic-Text2SQL-R1-7B. It specializes in Text-to-SQL generation, specifically for SQL++ queries, utilizing the NL2SQL++ v8 dataset with code-with-thought reasoning. This model is optimized for converting natural language queries into accurate SQL++ queries, making it suitable for database interaction and data retrieval tasks.
Loading preview...
Model Overview
This model, jastorj/snowflake_arctic_text2sql_r1_7b-nl2sqlpp-16bit-v5.1-cw-15K, is a specialized 7.6 billion parameter language model derived from the Snowflake/Arctic-Text2SQL-R1-7B base. It is meticulously fine-tuned for Text-to-SQL generation, focusing on the SQL++ query language.
Key Capabilities
- Natural Language to SQL++ Conversion: Translates complex natural language queries into executable SQL++ statements.
- Code-with-Thought Reasoning: Enhanced with a training methodology that incorporates reasoning steps, improving the accuracy and relevance of generated SQL++ queries.
- Optimized for Database Interaction: Specifically designed for use with databases that support SQL++ syntax, such as Couchbase.
- Efficient Fine-tuning: Utilizes LoRA (Low-Rank Adaptation) with Unsloth for efficient fine-tuning, resulting in a 16-bit quantized model.
Training Details
The model was fine-tuned on the NL2SQL++ v8 dataset, which includes 12,265 training examples and 1,000 validation examples. The training configuration involved a learning rate of 0.0002, a batch size of 8, and 3 epochs, with gradient checkpointing enabled to manage memory efficiently. The maximum sequence length supported is 15,000 tokens, allowing for complex schema and query inputs.