README update
Browse files
README.md
ADDED
|
@@ -0,0 +1,42 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language:
|
| 3 |
+
- ko
|
| 4 |
+
base_model:
|
| 5 |
+
- EleutherAI/polyglot-ko-1.3b
|
| 6 |
+
pipeline_tag: text-generation
|
| 7 |
+
tags:
|
| 8 |
+
- ingredient
|
| 9 |
+
- chatbot
|
| 10 |
+
- review
|
| 11 |
+
- usage
|
| 12 |
+
license: apache-2.0
|
| 13 |
+
---
|
| 14 |
+
|
| 15 |
+
## 🐥 Base Model
|
| 16 |
+
|
| 17 |
+
**EleutherAI/polyglot-ko-1.3b**
|
| 18 |
+
|
| 19 |
+
This model is based on the Korean version of Polyglot-1.3B, an open-source language model released by EleutherAI.
|
| 20 |
+
It is pre-trained on a large-scale Korean corpus and designed for general-purpose Korean language understanding and generation tasks.
|
| 21 |
+
|
| 22 |
+
---
|
| 23 |
+
|
| 24 |
+
## Training Procedure
|
| 25 |
+
|
| 26 |
+
### Training Hyperparameters
|
| 27 |
+
|
| 28 |
+
The following hyperparameters were used during training:
|
| 29 |
+
|
| 30 |
+
- `output_dir`: `./qlora_model_eleutherai`
|
| 31 |
+
- `per_device_train_batch_size`: `2`
|
| 32 |
+
- `gradient_accumulation_steps`: `4`
|
| 33 |
+
- `total_batch_size`: `8 (2 x 4)`
|
| 34 |
+
- `learning_rate`: `2e-5`
|
| 35 |
+
- `num_train_epochs`: `2`
|
| 36 |
+
- `fp16`: `True`
|
| 37 |
+
- `logging_dir`: `./logs`
|
| 38 |
+
- `logging_steps`: `5`
|
| 39 |
+
- `save_steps`: `100`
|
| 40 |
+
- `save_total_limit`: `1`
|
| 41 |
+
- `load_best_model_at_end`: `True`
|
| 42 |
+
- `metric_for_best_model`: `loss`
|