metadata
license: apache-2.0
tags:
- pruned
- math
- optimized
base_model: Qwen/Qwen2.5-3B-Instruct
Qwen2.5-3B-Instruct-math-medium-light
This model is a medium-light pruned version of Qwen/Qwen2.5-3B-Instruct, specialized for MATH tasks.
Pruning Details
- Base Model: Qwen/Qwen2.5-3B-Instruct
- Specialization: Math
- Prune Mode: Medium-light
- Method: Activation-based weight pruning
Performance Comparison
| Category | Original | Pruned |
|---|---|---|
| Python | 100.0% | 40.0% |
| HTML | 6.7% | 6.7% |
| Trivia | 100.0% | 100.0% |
| Math | 60.0% | 60.0% |
| Reasoning | N/A | N/A |
| Medical | 100.0% | 100.0% |
| Linux | 100.0% | 100.0% |
| Writing | 73.3% | 66.7% |
Usage
from transformers import AutoModelForCausalLM, AutoTokenizer
model = AutoModelForCausalLM.from_pretrained("CompactAI/Qwen2.5-3B-Instruct-math-medium-light")
tokenizer = AutoTokenizer.from_pretrained("CompactAI/Qwen2.5-3B-Instruct-math-medium-light")
License
This model inherits the license from the base model.