File size: 417 Bytes
5e21988 b19e138 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 |
---
license: cc-by-nc-sa-4.0
---
#### Experiment Objectives
1. Is Training with Korean + Multi-lingual dataset helpful to perform Korean benchmarks?
2. Does Full Parameter Depth-Up Scaled Training (expansion method: Llama-Pro) help to perform the best Korean benchmark performance?
#### Methods
1. Training CJK + En + Glot dataset with the same ratio of data size.
2. Layer Expansion and full parameter training.
|