hungeni commited on
Commit
d25a484
1 Parent(s): 1b85f85

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +7 -3
README.md CHANGED
@@ -5,6 +5,10 @@ language:
5
  - vi
6
  ---
7
 
8
- Base Model: LLaMa2 7B Chat HF
9
- + Continual Pre-Train with 2B tokens Vietnamese
10
- + Trainning profile: LoRa
 
 
 
 
 
5
  - vi
6
  ---
7
 
8
+ ## Base Model: LLaMa2 7B Chat HF
9
+ + Extend vocab to 44,800 for better Vietnamese understanding
10
+ + Continual Pre-Train with >2B tokens Vietnamese
11
+ + Trainning profile: LoRa (rank=32, alpha=128, 16fp), 1 epoch, block size = 512. Takes 300GPU Hours x RXT4090 24GB
12
+
13
+ ## Can be better use for
14
+ + Futher training / Fine-tuning for Vietnamese tasks