tinyllama_mole_sftv2_ultrachat_ep3
This model is a fine-tuned version of ondevicellm/tinyllama_mole_v1 on the HuggingFaceH4/ultrachat_200k dataset. It achieves the following results on the evaluation set:
- Loss: 1.7340
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: 2e-05
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 4
- gradient_accumulation_steps: 2
- total_train_batch_size: 128
- total_eval_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 120
- num_epochs: 3
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
2.7643 | 0.09 | 100 | 2.7492 |
2.7293 | 0.18 | 200 | 2.7330 |
2.6973 | 0.26 | 300 | 2.6920 |
2.612 | 0.35 | 400 | 2.6290 |
2.5257 | 0.44 | 500 | 2.5470 |
2.4656 | 0.53 | 600 | 2.4527 |
2.3607 | 0.61 | 700 | 2.3681 |
2.2885 | 0.7 | 800 | 2.2988 |
2.2384 | 0.79 | 900 | 2.2397 |
2.1585 | 0.88 | 1000 | 2.1877 |
2.1526 | 0.96 | 1100 | 2.1409 |
2.0845 | 1.05 | 1200 | 2.0986 |
2.049 | 1.14 | 1300 | 2.0603 |
2.0243 | 1.23 | 1400 | 2.0257 |
1.9899 | 1.31 | 1500 | 1.9950 |
1.9706 | 1.4 | 1600 | 1.9675 |
1.9414 | 1.49 | 1700 | 1.9429 |
1.8952 | 1.58 | 1800 | 1.9208 |
1.9038 | 1.66 | 1900 | 1.9013 |
1.8942 | 1.75 | 2000 | 1.8839 |
1.8652 | 1.84 | 2100 | 1.8679 |
1.823 | 1.93 | 2200 | 1.8531 |
1.8394 | 2.01 | 2300 | 1.8394 |
1.8347 | 2.1 | 2400 | 1.8268 |
1.8137 | 2.19 | 2500 | 1.8148 |
1.799 | 2.28 | 2600 | 1.8037 |
1.7774 | 2.37 | 2700 | 1.7931 |
1.771 | 2.45 | 2800 | 1.7832 |
1.7761 | 2.54 | 2900 | 1.7739 |
1.7458 | 2.63 | 3000 | 1.7652 |
1.7683 | 2.72 | 3100 | 1.7570 |
1.7389 | 2.8 | 3200 | 1.7490 |
1.7321 | 2.89 | 3300 | 1.7414 |
1.7418 | 2.98 | 3400 | 1.7340 |
Framework versions
- Transformers 4.37.0
- Pytorch 2.1.2+cu118
- Datasets 2.14.6
- Tokenizers 0.15.0
- Downloads last month
- 21
Inference API (serverless) does not yet support model repos that contain custom code.
Model tree for ondevicellm/tinyllama_mole_sftv2_ultrachat_ep3
Base model
ondevicellm/tinyllama_mole_v1