pretraining7
This model is a fine-tuned version of gpt2 on the None dataset. It achieves the following results on the evaluation set:
- Loss: 2.9516
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: 0.0006
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 10
- total_train_batch_size: 320
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: inverse_sqrt
- lr_scheduler_warmup_steps: 250
- training_steps: 1500
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
9.1748 | 0.1830 | 50 | 7.4163 |
6.7378 | 0.3660 | 100 | 6.2275 |
5.9659 | 0.5490 | 150 | 5.6559 |
5.4898 | 0.7321 | 200 | 5.2576 |
5.1248 | 0.9151 | 250 | 4.9392 |
4.7951 | 1.0981 | 300 | 4.6297 |
4.4843 | 1.2811 | 350 | 4.3591 |
4.2178 | 1.4641 | 400 | 4.0591 |
3.9729 | 1.6471 | 450 | 3.7843 |
3.766 | 1.8302 | 500 | 3.6225 |
3.6046 | 2.0132 | 550 | 3.5064 |
3.41 | 2.1962 | 600 | 3.4262 |
3.3702 | 2.3792 | 650 | 3.3577 |
3.309 | 2.5622 | 700 | 3.3027 |
3.2562 | 2.7452 | 750 | 3.2583 |
3.2027 | 2.9283 | 800 | 3.2192 |
3.1139 | 3.1113 | 850 | 3.1779 |
3.0442 | 3.2943 | 900 | 3.1549 |
3.0144 | 3.4773 | 950 | 3.1266 |
3.0016 | 3.6603 | 1000 | 3.0997 |
3.0001 | 3.8433 | 1050 | 3.0770 |
2.9655 | 4.0264 | 1100 | 3.0554 |
2.8328 | 4.2094 | 1150 | 3.0422 |
2.8343 | 4.3924 | 1200 | 3.0261 |
2.8266 | 4.5754 | 1250 | 3.0105 |
2.8236 | 4.7584 | 1300 | 2.9962 |
2.8194 | 4.9414 | 1350 | 2.9807 |
2.7161 | 5.1245 | 1400 | 2.9717 |
2.6842 | 5.3075 | 1450 | 2.9632 |
2.6898 | 5.4905 | 1500 | 2.9516 |
Framework versions
- Transformers 4.40.1
- Pytorch 2.3.0+cu121
- Datasets 2.19.0
- Tokenizers 0.19.1
- Downloads last month
- 3
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.