tinyllama_mole_sft_routeraux_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: 1.2128
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 |
---|---|---|---|
1.4007 | 0.09 | 100 | 1.3781 |
1.3255 | 0.18 | 200 | 1.3159 |
1.2919 | 0.26 | 300 | 1.2922 |
1.2696 | 0.35 | 400 | 1.2771 |
1.2426 | 0.44 | 500 | 1.2666 |
1.263 | 0.53 | 600 | 1.2584 |
1.2502 | 0.61 | 700 | 1.2514 |
1.237 | 0.7 | 800 | 1.2457 |
1.2322 | 0.79 | 900 | 1.2407 |
1.2156 | 0.88 | 1000 | 1.2360 |
1.2396 | 0.96 | 1100 | 1.2319 |
1.1564 | 1.05 | 1200 | 1.2315 |
1.1594 | 1.14 | 1300 | 1.2296 |
1.1711 | 1.23 | 1400 | 1.2274 |
1.1625 | 1.31 | 1500 | 1.2256 |
1.1652 | 1.4 | 1600 | 1.2234 |
1.1625 | 1.49 | 1700 | 1.2214 |
1.1457 | 1.58 | 1800 | 1.2196 |
1.1666 | 1.66 | 1900 | 1.2178 |
1.1701 | 1.75 | 2000 | 1.2158 |
1.1567 | 1.84 | 2100 | 1.2142 |
1.1304 | 1.93 | 2200 | 1.2128 |
1.1133 | 2.01 | 2300 | 1.2170 |
1.1204 | 2.1 | 2400 | 1.2170 |
1.1089 | 2.19 | 2500 | 1.2168 |
1.102 | 2.28 | 2600 | 1.2162 |
1.1004 | 2.37 | 2700 | 1.2157 |
1.1058 | 2.45 | 2800 | 1.2157 |
1.1119 | 2.54 | 2900 | 1.2150 |
1.0941 | 2.63 | 3000 | 1.2148 |
1.1127 | 2.72 | 3100 | 1.2147 |
1.104 | 2.8 | 3200 | 1.2144 |
1.1001 | 2.89 | 3300 | 1.2144 |
1.1188 | 2.98 | 3400 | 1.2144 |
Framework versions
- Transformers 4.37.0
- Pytorch 2.1.2+cu118
- Datasets 2.16.1
- Tokenizers 0.15.0
- Downloads last month
- 20
Inference API (serverless) does not yet support model repos that contain custom code.
Model tree for ondevicellm/tinyllama_mole_sft_routeraux_ultrachat_ep3
Base model
ondevicellm/tinyllama_mole_v1