Llama-2-13b-best_ratio90 (90% Parameters)

This model is a pruned and finetuned version of meta-llama/Llama-2-13b-hf, retaining approximately 90% of parameters while maintaining strong performance through genetic algorithm pruning and RMSNorm fine-tuning.

Model Details

  • Base Model: meta-llama/Llama-2-13b-hf
  • Parameter Retention: ~90%
  • Pruning Method: Genetic Algorithm
  • Fine-tuning Method: RMSNorm calibration

Performance

Metric Value
PPL (Before Fine-tuning) 5.80
PPL (After Fine-tuning) 5.01
Improvement 13.64%

Performance Comparison

Model PPL (After FT)
50% params 10.03
70% params 6.59
80% params 5.64
90% params 5.01

Files Included

  • : Full model state dict
  • : This documentation

License

Llama 2 Community License (inherited from base model)

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for ra225/Llama-2-13b-best_ratio90_rmsnorm_finetuned

Finetuned
(60)
this model