|
|
--- |
|
|
base_model: gpt2 |
|
|
library_name: distily |
|
|
license: mit |
|
|
tags: |
|
|
- generated_from_trainer |
|
|
model-index: |
|
|
- name: distily_bench_gpt2_linear_objectives |
|
|
results: [] |
|
|
--- |
|
|
|
|
|
# distily_bench_gpt2_optim |
|
|
|
|
|
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: 527.0228 |
|
|
- eval_frwikippl: 3796.0032 |
|
|
- eval_zhwikippl: 4795.4683 |
|
|
- eval_loss: 2376.6721 |
|
|
- eval_runtime: 21.817 |
|
|
- eval_samples_per_second: 45.836 |
|
|
- eval_steps_per_second: 11.459 |
|
|
|
|
|
<!-- 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: LinearObjective(logits_weight=1, logits_loss_fn=<function kl_divergence_loss at 0x7f57c4b07910>, activations_weight=1, activations_loss_fn=<function kl_divergence_loss at 0x7f57c4b07910>, attentions_weight=0, attentions_loss_fn=<function mse_loss at 0x7f57c4b07880>) |
|
|
- train_embeddings: True |
|
|
- learning_rate: 4e-05 |
|
|
- train_batch_size: 4 |
|
|
- eval_batch_size: 4 |
|
|
- seed: 42 |
|
|
- gradient_accumulation_steps: 4 |
|
|
- 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: 4.5067 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 | 55339.3672 | 57682.5742 | 31197.1836 | 21.7082 | 46.065 | 11.516 | 57080.2930 | |
|
|
| 500 | 0.0808 | 1509.5735 | 7497.0439 | 3194.9919 | 21.4587 | 46.601 | 11.65 | 50589.3438 | |
|
|
| 1000 | 0.1616 | 1083.2607 | 5620.3037 | 2923.7439 | 21.5879 | 46.322 | 11.581 | 29616.2285 | |
|
|
| 1500 | 0.2424 | 906.6083 | 4937.0078 | 2796.2080 | 21.6636 | 46.16 | 11.54 | 21403.5996 | |
|
|
| 2000 | 0.3232 | 813.4678 | 4877.3267 | 2706.0481 | 21.5303 | 46.446 | 11.612 | 20010.4863 | |
|
|
| 2500 | 0.4040 | 750.0352 | 4512.8765 | 2636.6079 | 21.6059 | 46.284 | 11.571 | 16546.3457 | |
|
|
| 3000 | 0.4848 | 704.7218 | 4373.6377 | 2583.7920 | 21.6069 | 46.281 | 11.57 | 14758.0859 | |
|
|
| 3500 | 0.5657 | 667.2821 | 4153.7866 | 2537.5520 | 21.59 | 46.318 | 11.579 | 14131.2881 | |
|
|
| 4000 | 0.6465 | 635.3494 | 4060.9749 | 2505.6001 | 21.554 | 46.395 | 11.599 | 13081.5996 | |
|
|
| 4500 | 0.7273 | 605.6495 | 4037.2766 | 2468.9121 | 21.795 | 45.882 | 11.471 | 11453.9658 | |
|
|
| 5000 | 0.8081 | 573.4954 | 3881.2524 | 2437.7439 | 21.6801 | 46.125 | 11.531 | 8931.2441 | |
|
|
| 5500 | 0.8889 | 557.2740 | 3918.3730 | 2413.4880 | 21.5054 | 46.5 | 11.625 | 6643.0454 | |
|
|
| 6000 | 0.9697 | 549.7523 | 4035.1443 | 2392.2400 | 21.6194 | 46.255 | 11.564 | 5330.4404 | |
|
|
| 6187 | 0.9999 | 527.0228 | 3796.0032 | 2376.6721 | 21.817 | 45.836 | 11.459 | 4795.4683 | |
|
|
|
|
|
### Framework versions |
|
|
- Distily 0.2.0 |
|
|
- Transformers 4.44.0 |
|
|
- Pytorch 2.3.0 |
|
|
- Datasets 2.20.0 |
|
|
|
