tlphams/gollm-12.8b-instruct-v2.0

gollm-instruct-all-in-one-v1

This model is a fine-tuned version of EleutherAI/polyglot-ko-12.8b on a custom mixed dataset

Model description

• No-context template

아래는 작업을 설명하는 질문어와 추가 컨텍스트를 제공하는 맥락이 함께 제공됩니다. 요청을 적절히 완료하는 답변을 작성하세요.

### 질문:
{instruction}

### 답변:

• With context template

아래는 작업을 설명하는 질문어와 추가 컨텍스트를 제공하는 맥락이 함께 제공됩니다. 요청을 적절히 완료하는 답변을 작성하세요.

### 맥락:
{input}

### 질문:
{instruction}

### 답변:

Intended uses & limitations

More information needed

Training and evaluation data

• self-introduction (20 samples)
• Combined KoAlpaca and KULLM - no-context samples only (145.8k samples)
  • KoAlpaca v1.0
  • KoAlpaca v1.1
  • KULLM (Dolly and Vicuna only)
• Naver news summarization (22.2k samples)
• KLUE MRC (17.5k samples)
• KLUE STS (5.6k samples)

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 8
• saved_checkpoint_at_epoch: 4 (condition: loss < 0.3)

Training results

Training Loss	Epoch	Step
1.5688	1.0	11947
1.0424	2.0	23895
0.5542	3.0	35843
0.2548	4.0	47791
0.1479	5.0	59738

Framework versions

• Transformers 4.32.0.dev0
• Pytorch 2.0.0+cu117
• Datasets 2.11.0
• Tokenizers 0.13.3