|
--- |
|
base_model: gpt2 |
|
library_name: distily |
|
license: mit |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: distily_bench_gpt2_optim_extended2 |
|
results: [] |
|
--- |
|
|
|
# distily_bench_gpt2_optim_extended2 |
|
|
|
This student model is distilled from the teacher model [gpt2](https://huggingface.co/gpt2) using the dataset (unspecified). |
|
|
|
The [Distily](https://github.com/lapp0/distily) library was used for this distillation. |
|
|
|
It achieves the following results on the evaluation set: |
|
- eval_enwikippl: 603.2673 |
|
- eval_frwikippl: 3866.3679 |
|
- eval_zhwikippl: 9060.9883 |
|
- eval_loss: 6355.0508 |
|
- eval_runtime: 64.6366 |
|
- eval_samples_per_second: 46.413 |
|
- eval_steps_per_second: 11.603 |
|
|
|
<!-- 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. |
|
|
|
## 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: |
|
- distillation_objective: 'legacy' |
|
- loss_fn: kl |
|
- train_embeddings: True |
|
- learning_rate: 4e-05 |
|
- train_batch_size: 8 |
|
- eval_batch_size: 4 |
|
- seed: 42 |
|
- gradient_accumulation_steps: 2 |
|
- total_train_batch_size: 16 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: constant |
|
- num_epochs: 1.0 |
|
|
|
### Resource Usage |
|
Peak GPU Memory: 8.3344 GB |
|
|
|
### Eval-Phase Metrics |
|
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | zhwikippl | |
|
| --- | --- | --- | --- | --- | --- | --- | --- | --- | |
|
| **teacher eval** | | 30.2385 | 57.2728 | | | | | 18.1772 | |
|
| 0 | 0 | 55332.9297 | 57511.9648 | 333834.9375 | 63.8516 | 46.984 | 11.746 | 57797.4375 | |
|
| 500 | 0.0269 | 2446.0188 | 10865.3799 | 11817.8984 | 64.1124 | 46.793 | 11.698 | 39870.7812 | |
|
| 1000 | 0.0539 | 1804.3785 | 6767.0361 | 9836.2031 | 64.095 | 46.806 | 11.701 | 19923.8262 | |
|
| 1500 | 0.0808 | 1456.1499 | 5625.2583 | 9170.7520 | 65.271 | 45.962 | 11.491 | 18979.7988 | |
|
| 2000 | 0.1077 | 1255.2349 | 5859.0298 | 8742.0908 | 64.4753 | 46.529 | 11.632 | 17829.9570 | |
|
| 2500 | 0.1347 | 1123.1558 | 5142.7266 | 8474.3467 | 64.6172 | 46.427 | 11.607 | 18204.0723 | |
|
| 3000 | 0.1616 | 1041.0769 | 5179.4790 | 8192.3838 | 64.0965 | 46.804 | 11.701 | 16922.875 | |
|
| 3500 | 0.1886 | 948.0062 | 4929.7056 | 7911.0400 | 64.5488 | 46.476 | 11.619 | 22088.8789 | |
|
| 4000 | 0.2155 | 899.1066 | 4752.2407 | 7641.6426 | 65.5993 | 45.732 | 11.433 | 16942.1074 | |
|
| 4500 | 0.2424 | 843.3125 | 4732.0117 | 7480.7788 | 64.5158 | 46.5 | 11.625 | 13217.6758 | |
|
| 5000 | 0.2694 | 796.5746 | 4456.2817 | 7343.6479 | 65.0161 | 46.142 | 11.536 | 12772.6074 | |
|
| 5500 | 0.2963 | 772.2271 | 4386.7627 | 7222.3145 | 65.0008 | 46.153 | 11.538 | 11082.3330 | |
|
| 6000 | 0.3232 | 723.9974 | 4267.7817 | 7016.9600 | 64.7743 | 46.315 | 11.579 | 9581.7812 | |
|
| 6500 | 0.3502 | 696.7773 | 4287.5391 | 6892.1387 | 64.6727 | 46.387 | 11.597 | 8422.7246 | |
|
| 7000 | 0.3771 | 679.4652 | 4046.8250 | 6773.9629 | 64.6977 | 46.369 | 11.592 | 7275.9604 | |
|
| 7500 | 0.4040 | 667.8522 | 4138.4370 | 6713.6533 | 65.028 | 46.134 | 11.533 | 8175.5986 | |
|
| 8000 | 0.4310 | 647.4772 | 3977.0999 | 6626.9331 | 64.3886 | 46.592 | 11.648 | 5914.0166 | |
|
| 8500 | 0.4579 | 627.8210 | 3850.3174 | 6548.4160 | 64.3532 | 46.618 | 11.654 | 7728.6548 | |
|
| 9000 | 0.4848 | 608.0646 | 3773.8511 | 6449.4614 | 64.3549 | 46.616 | 11.654 | 7419.7065 | |
|
| 9500 | 0.5118 | 603.2673 | 3866.3679 | 6355.0508 | 64.6366 | 46.413 | 11.603 | 9060.9883 | |
|
| 10000 | 0.5387 | 588.2559 | 3563.7371 | 6282.0479 | 65.1489 | 46.048 | 11.512 | 7187.1206 | |
|
| 10500 | 0.5657 | 569.4130 | 3654.1926 | 6309.9839 | 64.8852 | 46.235 | 11.559 | 7732.7837 | |
|
| 11000 | 0.5926 | 572.8280 | 3728.8887 | 6206.9868 | 65.1196 | 46.069 | 11.517 | 6973.9194 | |
|
| 11500 | 0.6195 | 551.1736 | 3640.4358 | 6146.9331 | 65.3439 | 45.911 | 11.478 | 5983.9292 | |
|
| 12000 | 0.6465 | 544.3150 | 3507.0312 | 6073.0454 | 65.3717 | 45.891 | 11.473 | 5726.3408 | |
|
| 12500 | 0.6734 | 538.8688 | 3312.2402 | 6032.6079 | 65.1968 | 46.015 | 11.504 | 5642.0854 | |
|
| 13000 | 0.7003 | 525.2048 | 3317.0325 | 6042.6240 | 65.216 | 46.001 | 11.5 | 11299.7695 | |
|
| 13500 | 0.7273 | 516.2283 | 3381.7358 | 5946.3682 | 67.4205 | 44.497 | 11.124 | 7501.9004 | |
|
| 14000 | 0.7542 | 508.5393 | 3201.6807 | 5921.8345 | 65.0932 | 46.088 | 11.522 | 7485.8843 | |
|
| 14500 | 0.7811 | 499.8382 | 3091.7612 | 5887.8721 | 65.2716 | 45.962 | 11.49 | 5927.4609 | |
|
| 15000 | 0.8081 | 491.9155 | 3132.6841 | 5930.3252 | 65.4781 | 45.817 | 11.454 | 7431.6040 | |
|
| 15500 | 0.8350 | 485.9736 | 3050.2964 | 5844.8960 | 65.1349 | 46.058 | 11.515 | 6106.6260 | |
|
| 16000 | 0.8620 | 483.0016 | 2964.3213 | 5828.2241 | 65.6241 | 45.715 | 11.429 | 5001.1572 | |
|
| 16500 | 0.8889 | 480.1220 | 2957.3284 | 5789.1626 | 65.5498 | 45.767 | 11.442 | 4932.5088 | |
|
| 17000 | 0.9158 | 470.7449 | 2851.3689 | 5783.3174 | 65.2632 | 45.968 | 11.492 | 4651.6655 | |
|
| 17500 | 0.9428 | 471.4951 | 2821.2729 | 5762.3945 | 65.89 | 45.53 | 11.383 | 4335.2710 | |
|
| 18000 | 0.9697 | 467.4575 | 2898.3936 | 5772.5654 | 65.0307 | 46.132 | 11.533 | 3703.9866 | |
|
| 18500 | 0.9966 | 465.3025 | 2792.5769 | 5640.9438 | 65.1725 | 46.032 | 11.508 | 4174.7715 | |
|
| 18562 | 1.0000 | 459.9143 | 2775.4995 | 5699.1255 | 65.9141 | 45.514 | 11.378 | 4052.5564 | |
|
|
|
### Framework versions |
|
- Distily 0.2.0 |
|
- Transformers 4.44.0 |
|
- Pytorch 2.3.0 |
|
- Datasets 2.20.0 |
|
|
