File size: 199 Bytes
b85ba72
 
 
 
 
 
1
2
3
4
5
6
7
# fokan/medgemma-4b-it-fp16-pruned20
FP16 + 20% pruned version of `google/medgemma-4b-it`

- Precision: FP16
- Pruning: L1 (20%) on all Linear layers
- Balanced size/speed with minimal accuracy loss