File size: 1,115 Bytes
afeba10
 
f7a230f
 
4b1bc0e
 
 
f7a230f
7dff9e7
f7a230f
4b1bc0e
 
 
 
 
afeba10
f7a230f
c53f71b
 
 
 
f7a230f
7dff9e7
f7a230f
4b1bc0e
f7a230f
726a252
 
 
f7a230f
 
 
 
 
 
 
 
4b1bc0e
f7a230f
a1edf0e
 
f7a230f
 
 
a1edf0e
f7a230f
 
 
 
 
 
4b1bc0e
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
---
license: apache-2.0
tags:
- generated_from_trainer
- polyglot-ko
- gpt-neox
- KoAlpaca
model-index:
- name: KoAlpaca-Polyglot-12.8B
  results: []
language:
- ko
datasets:
- KoAlpaca-v1.1b
pipeline_tag: text-generation
---

Update @ 2023.06.01

- Add Safetensor sharded model weight (max shard = 1GB)


# KoAlpaca-Polyglot-12.8B (v1.1b)

This model is a fine-tuned version of [EleutherAI/polyglot-ko-12.8b](https://huggingface.co/EleutherAI/polyglot-ko-12.8b) on a KoAlpaca Dataset v1.1b

Detail Codes are available at [KoAlpaca Github Repository](https://github.com/Beomi/KoAlpaca)


## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 5e-05
- train_batch_size: 1
- seed: 42
- distributed_type: multi-GPU (A100 80G)
- num_devices: 4
- gradient_accumulation_steps: 64
- total_train_batch_size: 256
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- num_epochs: 2.0

### Framework versions

- Transformers 4.28.1
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.3