Create README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
KoAlpaca-Polyglot-12.8B (v1.1b)
|
| 2 |
+
This model is a fine-tuned version of EleutherAI/polyglot-ko-12.8b on a KoAlpaca Dataset v1.1b
|
| 3 |
+
|
| 4 |
+
Detail Codes are available at KoAlpaca Github Repository
|
| 5 |
+
|
| 6 |
+
Training procedure
|
| 7 |
+
Training hyperparameters
|
| 8 |
+
The following hyperparameters were used during training:
|
| 9 |
+
|
| 10 |
+
learning_rate: 5e-05
|
| 11 |
+
train_batch_size: 1
|
| 12 |
+
seed: 42
|
| 13 |
+
distributed_type: multi-GPU (A100 80G)
|
| 14 |
+
num_devices: 4
|
| 15 |
+
gradient_accumulation_steps: 64
|
| 16 |
+
total_train_batch_size: 256
|
| 17 |
+
total_eval_batch_size: 32
|
| 18 |
+
optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
|
| 19 |
+
lr_scheduler_type: linear
|
| 20 |
+
num_epochs: 2.0
|
| 21 |
+
Framework versions
|
| 22 |
+
Transformers 4.28.1
|
| 23 |
+
Pytorch 2.0.0+cu117
|
| 24 |
+
Datasets 2.11.0
|
| 25 |
+
Tokenizers 0.13.3
|