gemma7_on_korean_conv-base-stm
This model is a fine-tuned version of beomi/gemma-ko-7b on the None dataset. It achieves the following results on the evaluation set:
- Loss: 1.2906
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: 5e-05
- train_batch_size: 1
- eval_batch_size: 1
- seed: 42
- gradient_accumulation_steps: 20
- total_train_batch_size: 20
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 100
- training_steps: 2000
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
1.0429 | 0.2563 | 100 | 1.0310 |
0.9035 | 0.5126 | 200 | 0.9042 |
0.8337 | 0.7688 | 300 | 0.8497 |
0.6688 | 1.0251 | 400 | 0.8408 |
0.5795 | 1.2814 | 500 | 0.7815 |
0.5576 | 1.5377 | 600 | 0.7739 |
0.5633 | 1.7940 | 700 | 0.7481 |
0.3616 | 2.0502 | 800 | 0.8965 |
0.2926 | 2.3065 | 900 | 0.8460 |
0.3151 | 2.5628 | 1000 | 0.8291 |
0.3094 | 2.8191 | 1100 | 0.8232 |
0.1368 | 3.0753 | 1200 | 1.0319 |
0.1451 | 3.3316 | 1300 | 1.0212 |
0.1486 | 3.5879 | 1400 | 1.0126 |
0.1445 | 3.8442 | 1500 | 1.0024 |
0.05 | 4.1005 | 1600 | 1.2423 |
0.0532 | 4.3567 | 1700 | 1.2510 |
0.0472 | 4.6130 | 1800 | 1.2756 |
0.0451 | 4.8693 | 1900 | 1.2673 |
0.0201 | 5.1256 | 2000 | 1.2906 |
Framework versions
- PEFT 0.11.1
- Transformers 4.41.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.1
- Tokenizers 0.19.1
- Downloads last month
- 2
Model tree for ghost613/gemma7_on_korean_conv-base-stm
Base model
beomi/gemma-ko-7b