Create README.md
#1
by
metterian
- opened
README.md
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
license: apache-2.0
|
| 3 |
+
language:
|
| 4 |
+
- ko
|
| 5 |
+
---
|
| 6 |
+
# KULLM-Polyglot-12.8B-v2
|
| 7 |
+
|
| 8 |
+
This model is a fine-tuned version of [EleutherAI/polyglot-ko-12.8b](https://huggingface.co/EleutherAI/polyglot-ko-12.8b) on a KULLM v2
|
| 9 |
+
|
| 10 |
+
Detail Codes are available at [KULLM Github Repository](https://github.com/nlpai-lab/KULLM)
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
## Training procedure
|
| 14 |
+
|
| 15 |
+
### Training hyperparameters
|
| 16 |
+
|
| 17 |
+
The following hyperparameters were used during training:
|
| 18 |
+
|
| 19 |
+
- learning_rate: 3e-4
|
| 20 |
+
- train_batch_size: 6
|
| 21 |
+
- seed: 42
|
| 22 |
+
- distributed_type: multi-GPU (A100 80G)
|
| 23 |
+
- num_devices: 4
|
| 24 |
+
- gradient_accumulation_steps: 21.3
|
| 25 |
+
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 26 |
+
- lr_scheduler_type: linear
|
| 27 |
+
- num_epochs: 8.0
|
| 28 |
+
|
| 29 |
+
### Framework versions
|
| 30 |
+
|
| 31 |
+
- Transformers 4.28.1
|
| 32 |
+
- Pytorch 2.0.0+cu117
|
| 33 |
+
- Datasets 2.11.0
|
| 34 |
+
- Tokenizers 0.13.3
|