File size: 574 Bytes
9a031a7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
bllossom_polyglot-12.8b_0725
This model is a fine-tuned version of EleutherAI/polyglot-ko-12.8b on LIMA dataset(translated in Korean)

Training hyperparameters
The following hyperparameters were used during training:

learning_rate: 0.0002
train_batch_size: 4
eval_batch_size: 1
seed: 0
gradient_accumulation_steps: 8
total_train_batch_size: 32
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
lr_scheduler_type: constant
lr_scheduler_warmup_ratio: 0.03
training_steps: 200
Framework versions
Transformers 4.31.0.dev0
Pytorch 1.13.1
Datasets 2.13.1
Tokenizers 0.13.3