<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
gpt-expt-sp
This model is a fine-tuned version of gpt2 on an unknown dataset. It achieves the following results on the evaluation set:
- Loss: 0.2561
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: 0.0005
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 8
- total_train_batch_size: 256
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 1000
- num_epochs: 300
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss |
---|---|---|---|
3.8623 | 3.12 | 100 | 1.7653 |
1.6403 | 6.24 | 200 | 1.5635 |
1.5806 | 9.37 | 300 | 1.5326 |
1.5433 | 12.49 | 400 | 1.4568 |
1.362 | 15.61 | 500 | 0.9368 |
0.8739 | 18.73 | 600 | 0.5006 |
0.5905 | 21.85 | 700 | 0.3875 |
0.4755 | 24.98 | 800 | 0.3440 |
0.4252 | 28.12 | 900 | 0.3238 |
0.3904 | 31.24 | 1000 | 0.3093 |
0.366 | 34.37 | 1100 | 0.3004 |
0.3492 | 37.49 | 1200 | 0.2922 |
0.3345 | 40.61 | 1300 | 0.2860 |
0.3277 | 43.73 | 1400 | 0.2819 |
0.324 | 46.85 | 1500 | 0.2800 |
0.318 | 49.98 | 1600 | 0.2766 |
0.314 | 53.12 | 1700 | 0.2736 |
0.308 | 56.24 | 1800 | 0.2740 |
0.306 | 59.37 | 1900 | 0.2716 |
0.3037 | 62.49 | 2000 | 0.2708 |
0.2993 | 65.61 | 2100 | 0.2685 |
0.2991 | 68.73 | 2200 | 0.2680 |
0.297 | 71.85 | 2300 | 0.2670 |
0.2964 | 74.98 | 2400 | 0.2662 |
0.2964 | 78.12 | 2500 | 0.2653 |
0.2942 | 81.24 | 2600 | 0.2664 |
0.2937 | 84.37 | 2700 | 0.2655 |
0.2886 | 87.49 | 2800 | 0.2631 |
0.2877 | 90.61 | 2900 | 0.2634 |
0.2859 | 93.73 | 3000 | 0.2628 |
0.2852 | 96.85 | 3100 | 0.2629 |
0.2841 | 99.98 | 3200 | 0.2629 |
0.2848 | 103.12 | 3300 | 0.2625 |
0.2811 | 106.24 | 3400 | 0.2611 |
0.281 | 109.37 | 3500 | 0.2608 |
0.2794 | 112.49 | 3600 | 0.2599 |
0.2787 | 115.61 | 3700 | 0.2604 |
0.2781 | 118.73 | 3800 | 0.2601 |
0.2777 | 121.85 | 3900 | 0.2604 |
0.2776 | 124.98 | 4000 | 0.2600 |
0.2786 | 128.12 | 4100 | 0.2597 |
0.2757 | 131.24 | 4200 | 0.2597 |
0.2754 | 134.37 | 4300 | 0.2590 |
0.2758 | 137.49 | 4400 | 0.2596 |
0.2742 | 140.61 | 4500 | 0.2598 |
0.2731 | 143.73 | 4600 | 0.2581 |
0.2738 | 146.85 | 4700 | 0.2587 |
0.273 | 149.98 | 4800 | 0.2583 |
0.2736 | 153.12 | 4900 | 0.2579 |
0.271 | 156.24 | 5000 | 0.2580 |
0.2709 | 159.37 | 5100 | 0.2578 |
0.2708 | 162.49 | 5200 | 0.2582 |
0.2697 | 165.61 | 5300 | 0.2578 |
0.2695 | 168.73 | 5400 | 0.2578 |
0.269 | 171.85 | 5500 | 0.2582 |
0.2691 | 174.98 | 5600 | 0.2574 |
0.2705 | 178.12 | 5700 | 0.2574 |
0.2678 | 181.24 | 5800 | 0.2572 |
0.2692 | 184.37 | 5900 | 0.2582 |
0.2687 | 187.49 | 6000 | 0.2572 |
0.2673 | 190.61 | 6100 | 0.2571 |
0.2666 | 193.73 | 6200 | 0.2568 |
0.2662 | 196.85 | 6300 | 0.2573 |
0.2662 | 199.98 | 6400 | 0.2568 |
0.2688 | 203.12 | 6500 | 0.2567 |
0.2658 | 206.24 | 6600 | 0.2570 |
0.2666 | 209.37 | 6700 | 0.2567 |
0.2652 | 212.49 | 6800 | 0.2565 |
0.2651 | 215.61 | 6900 | 0.2568 |
0.2649 | 218.73 | 7000 | 0.2566 |
0.2648 | 221.85 | 7100 | 0.2564 |
0.2645 | 224.98 | 7200 | 0.2564 |
0.2662 | 228.12 | 7300 | 0.2564 |
0.2641 | 231.24 | 7400 | 0.2564 |
0.2641 | 234.37 | 7500 | 0.2563 |
0.2639 | 237.49 | 7600 | 0.2563 |
0.2638 | 240.61 | 7700 | 0.2563 |
0.2637 | 243.73 | 7800 | 0.2562 |
0.2635 | 246.85 | 7900 | 0.2562 |
0.2633 | 249.98 | 8000 | 0.2563 |
0.2653 | 253.12 | 8100 | 0.2562 |
0.2631 | 256.24 | 8200 | 0.2562 |
0.2631 | 259.37 | 8300 | 0.2561 |
0.263 | 262.49 | 8400 | 0.2561 |
0.263 | 265.61 | 8500 | 0.2561 |
0.2629 | 268.73 | 8600 | 0.2561 |
0.2628 | 271.85 | 8700 | 0.2561 |
0.2628 | 274.98 | 8800 | 0.2561 |
0.2646 | 278.12 | 8900 | 0.2561 |
0.2626 | 281.24 | 9000 | 0.2561 |
0.2626 | 284.37 | 9100 | 0.2561 |
0.2625 | 287.49 | 9200 | 0.2561 |
0.2626 | 290.61 | 9300 | 0.2561 |
0.2626 | 293.73 | 9400 | 0.2561 |
0.2626 | 296.85 | 9500 | 0.2561 |
0.2625 | 299.98 | 9600 | 0.2561 |
Framework versions
- Transformers 4.25.1
- Pytorch 1.13.1+cu117
- Datasets 2.8.0
- Tokenizers 0.13.2