Snowflake/Arctic-Text2SQL-R1-7B Fine-tuned for NL2SQL++ v8

Model Details

• Base Model: Snowflake/Arctic-Text2SQL-R1-7B
• Task: Text-to-SQL generation
• Dataset: NL2SQL++ v8 with code-with-thought reasoning
• Fine-tuning Method: LoRA (Low-Rank Adaptation) with Unsloth
• Quantization: 16-bit merged weights
• Maximum Sequence Length: 32768 tokens
• Training Dataset Size: 46344 examples
• Validation Dataset Size: 1986 examples

Training Configuration

LoRA Parameters

• LoRA Rank (r): 64
• LoRA Alpha: 128
• Target Modules: q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj, down_proj

Training Hyperparameters

• Learning Rate: 0.0002
• Training Epochs: 2
• Max Steps: N/A (using epochs)
• Train Batch Size: 64
• Eval Batch Size: 50
• Gradient Accumulation Steps: 2
• Effective Batch Size: 128
• Warmup Steps: 0
• Warmup Ratio: 0.1
• Optimizer: AdamW (torch)
• Learning Rate Scheduler: Cosine
• Weight Decay: 0.01
• Max Gradient Norm: 1.0
• Seed: 3407