A LoRA fine-tuned model based on zai-org/GLM-4.6 specialized for the Hyperswitch Rust codebase. This model excels at understanding payment processing patterns, Hyperswitch architecture, and Rust development practices.

🎯 Model Description

This LoRA adapter was trained on 16,731 samples extracted from the Hyperswitch codebase to enhance code understanding, explanation, and generation within the payment processing domain.

Base Model: zai-org/GLM-4.6
Training Type: Causal Language Modeling (CLM) with LoRA
Domain: Payment Processing, Rust Development
Specialization: Hyperswitch codebase patterns and architecture

📊 Training Details

LoRA Configuration

yaml

r: 16                   # LoRA rank
alpha: 32              # LoRA alpha (2*r)
dropout: 0.05           # LoRA dropout
  target_modules:
    - "q_proj"   
    - "k_proj"    
    - "v_proj"   
    - "o_proj"    
  
  exclude_modules:
    - "block_sparse_moe"  
    - "w1"                
    - "w2"               
    - "w3"               
    - "gate"             

Training Hyperparameters

Epochs: 5
Learning Rate: 2e-4 (cosine schedule)
Hardware: 8 x NVIDIA H200

🛠️ Technical Specifications

Precision: bfloat16
Inference Speed: Optimized with Flash Attention 2

🙏 Acknowledgments

Zai Team for the excellent GLM 4.6 base model
Hyperswitch Team for the open-source payment processing platform
Hugging Face for the transformers and PEFT libraries

📞 Citation

bibtex

@misc{GLM-4.6-CPT-LoRA-HyperSwitch-v2,
  title={AdityaNarayan/GLM-4.6-CPT-LoRA-HyperSwitch-v2},
  author={Aditya Narayan},
  year={2024},
  publisher={Hugging Face},
  url={https://huggingface.co/AdityaNarayan/GLM-4.6-CPT-LoRA-HyperSwitch-v2}
}