jastorj/couchmind-rlt-v5.8.0.1_rl-cw-11K-16bit

TEXT GENERATIONConcurrent Unit Cost:1Model Size:7.6BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 20, 2026License:apache-2.0Architecture:Transformer Open Weights Featherless Exclusive Cold

The jastorj/couchmind-rlt-v5.8.0.1_rl-cw-11K-16bit is a 7.6 billion parameter language model developed by jastorj, fine-tuned for Text-to-SQL generation. This model specializes in converting natural language questions into syntactically valid Couchbase SQL++ queries, leveraging a reinforcement learning approach with GRPO on the NL2SQL++ v5.8.0.1_rl dataset. It is optimized for precise SQL generation based on provided database schemas, making it suitable for applications requiring accurate database interaction from natural language inputs.

Loading preview...

Model Overview

This model, jastorj/couchmind-rlt-v5.8.0.1_rl-cw-11K-16bit, is a 7.6 billion parameter language model developed by jastorj. It is a reinforcement-learning fine-tuned version of jastorj/couchmind-v5.8_rl_cold_start-cw-26K-16bit, specifically designed for Text-to-SQL generation.

Key Capabilities

  • Specialized Text-to-SQL: Converts natural language questions into Couchbase SQL++ queries.
  • Reinforcement Learning Fine-tuning: Utilizes GRPO (Group Relative Policy Optimisation) with Unsloth for enhanced performance on SQL generation tasks.
  • Couchbase SQL++ Expertise: Trained on the NL2SQL++ v5.8.0.1_rl dataset, focusing on generating accurate queries for Couchbase databases.
  • Schema-Aware Query Generation: Adheres to provided database schemas, including bucket, scope, and collection names, to ensure syntactically valid and precise SQL++ output.
  • Quantized Weights: Features 16-bit merged weights for potentially improved efficiency.

Training Details

The model was fine-tuned on a dataset of 5000 examples from the NL2SQL++ v5.8.0.1_rl dataset. It is designed to generate SQL++ queries that select only explicitly requested columns and use exact naming conventions from the schema.

Good For

  • Developers building applications that require converting natural language into Couchbase SQL++ queries.
  • Automating database interactions where precise SQL generation based on a given schema is critical.
  • Use cases involving Couchbase databases where a robust Text-to-SQL solution is needed.

Finetunes
1 models