---
library_name: transformers
base_model: minpeter/pretrained-tiny-ko
tags:
- axolotl
- generated_from_trainer
datasets:
- lemon-mint/Korean-FineTome-100k
- lemon-mint/smol-koreantalk
- heegyu/open-korean-instructions-v20231020
- FreedomIntelligence/evol-instruct-korean
- FreedomIntelligence/alpaca-gpt4-korean
- FreedomIntelligence/sharegpt-korean
- coastral/korean-writing-style-instruct
- devngho/korean-instruction-mix
model-index:
- name: tiny-ko-sft
  results: []
---

<!-- This model card has been generated automatically according to the information the Trainer had access to. You
should probably proofread and complete it, then remove this comment. -->

[<img src="https://raw.githubusercontent.com/axolotl-ai-cloud/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>](https://github.com/axolotl-ai-cloud/axolotl)
<details><summary>See axolotl config</summary>

axolotl version: `0.10.0.dev0`
```yaml
base_model: minpeter/pretrained-tiny-ko

hub_model_id: minpeter/tiny-ko-sft
output_dir: ./outputs/tiny-ko-sft
wandb_project: "axolotl"
wandb_entity: "kasfiekfs-e"

chat_template: chatml
datasets:
  - path: lemon-mint/Korean-FineTome-100k
    type: chat_template
    split: train[:10%]
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: lemon-mint/smol-koreantalk
    type: chat_template
    split: train[:10%]
    field_messages: messages
    message_property_mappings:
      role: role
      content: content

  - path: heegyu/open-korean-instructions-v20231020
    type: chat_template
    split: train[:10%]
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value
    roles:
      user: ["human", "user"]
      assistant: ["gpt", "assistant", "bot"]
      system: ["system", "input"]

  # NOTE: https://github.com/FreedomIntelligence/MultilingualSIFT
  - path: FreedomIntelligence/evol-instruct-korean
    type: chat_template
    split: train[:10%]
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value

  - path: FreedomIntelligence/alpaca-gpt4-korean
    type: chat_template
    split: train[:10%]
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value

  - path: FreedomIntelligence/sharegpt-korean
    type: chat_template
    split: train[:10%]
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value

  - path: coastral/korean-writing-style-instruct
    type: chat_template
    split: train[:10%]
    field_messages: conversations
    message_property_mappings:
      role: from
      content: value

  - path: devngho/korean-instruction-mix
    type: chat_template
    split: train[:10%]
    field_messages: messages
    message_property_mappings:
      role: from
      content: value

dataset_prepared_path: last_run_prepared
val_set_size: 0.05

save_steps: 200
warmup_steps: 20
eval_steps: 200

sequence_len: 2048

# <<<< experimental settings <<<<
sample_packing: false
train_on_inputs: true
# >>>> experimental settings >>>

pad_to_sequence_len: true

gradient_accumulation_steps: 4
micro_batch_size: 16

optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 1e-3

bf16: auto
tf32: false

added_tokens_overrides:
  128001: "<|im_end|>"
  128002: "<|im_start|>"

special_tokens:
  bos_token: <|begin_of_text|>
  eos_token: <|im_end|>
  pad_token: <|im_end|>

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
resume_from_checkpoint:
logging_steps: 1
flash_attention: true

num_epochs: 3
weight_decay: 0.0

```

</details><br>

# tiny-ko-sft

This model is a fine-tuned version of [minpeter/pretrained-tiny-ko](https://huggingface.co/minpeter/pretrained-tiny-ko) on the lemon-mint/Korean-FineTome-100k, the lemon-mint/smol-koreantalk, the heegyu/open-korean-instructions-v20231020, the FreedomIntelligence/evol-instruct-korean, the FreedomIntelligence/alpaca-gpt4-korean, the FreedomIntelligence/sharegpt-korean, the coastral/korean-writing-style-instruct and the devngho/korean-instruction-mix datasets.
It achieves the following results on the evaluation set:
- Loss: 1.4059

## Model description

More information needed

## Intended uses & limitations

More information needed

## Training and evaluation data

More information needed

## Training procedure

### Training hyperparameters

The following hyperparameters were used during training:
- learning_rate: 0.001
- train_batch_size: 16
- eval_batch_size: 16
- seed: 42
- distributed_type: multi-GPU
- num_devices: 2
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- total_eval_batch_size: 32
- optimizer: Use OptimizerNames.PAGED_ADAMW_8BIT with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 20
- training_steps: 2972

### Training results

| Training Loss | Epoch  | Step | Validation Loss |
|:-------------:|:------:|:----:|:---------------:|
| 3.9539        | 0.0010 | 1    | 3.9757          |
| 1.6999        | 0.2019 | 200  | 1.6884          |
| 1.6123        | 0.4037 | 400  | 1.6288          |
| 1.5387        | 0.6056 | 600  | 1.5876          |
| 1.5681        | 0.8075 | 800  | 1.5429          |
| 1.3066        | 1.0091 | 1000 | 1.5208          |
| 1.395         | 1.2110 | 1200 | 1.5007          |
| 1.3474        | 1.4128 | 1400 | 1.4699          |
| 1.3025        | 1.6147 | 1600 | 1.4383          |
| 1.2566        | 1.8166 | 1800 | 1.4117          |
| 1.1672        | 2.0182 | 2000 | 1.4227          |
| 1.1267        | 2.2200 | 2200 | 1.4141          |
| 1.0195        | 2.4219 | 2400 | 1.4098          |
| 1.084         | 2.6238 | 2600 | 1.4063          |
| 1.1254        | 2.8256 | 2800 | 1.4059          |


### Framework versions

- Transformers 4.52.3
- Pytorch 2.6.0+cu124
- Datasets 3.6.0
- Tokenizers 0.21.1