|
--- |
|
base_model: gpt2 |
|
library_name: Distily |
|
license: mit |
|
tags: |
|
- generated_from_trainer |
|
model-index: |
|
- name: distily_bench_obj_cross_v2.15_gpt2 |
|
results: [] |
|
--- |
|
|
|
# distily_bench_obj_cross_v2.15_gpt2 |
|
|
|
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: 84.0 |
|
- eval_frwikippl: 342.0 |
|
- eval_zhwikippl: 217.0 |
|
- eval_tinystoriesppl: 69.5 |
|
- eval_loss: 0.6877 |
|
- eval_runtime: 16.9969 |
|
- eval_samples_per_second: 58.834 |
|
- eval_steps_per_second: 7.354 |
|
|
|
<!-- 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: DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl, layer_mapper=None, projector=None), hs_loss_component=LossComponent(label=hs, weight=1.0, loss_fn=mse, layer_mapper=last, projector=None), attn_loss_component=LossComponent(label=attn, weight=0, loss_fn=None, layer_mapper=None, projector=None)) |
|
- train_embeddings: True |
|
- learning_rate: 0.0001 |
|
- train_batch_size: 4 |
|
- eval_batch_size: 8 |
|
- seed: 42 |
|
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08 |
|
- lr_scheduler_type: constant |
|
- lr_scheduler_warmup_ratio: 0.2 |
|
- num_epochs: 1.0 |
|
|
|
### Resource Usage |
|
Peak GPU Memory: 7.7252 GB |
|
|
|
### Eval-Phase Metrics |
|
| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl | |
|
| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | |
|
| **teacher eval** | | 43.75 | 61.75 | | | | | 11.8125 | 19.125 | |
|
| 0 | 0 | 2473901162496.0 | 170424302305280.0 | 20.7680 | 17.0409 | 58.682 | 7.335 | 4060086272.0 | 71468255805440.0 | |
|
| 1000 | 0.0404 | 334.0 | 1464.0 | 1.5419 | 17.0178 | 58.762 | 7.345 | 243.0 | 596.0 | |
|
| 2000 | 0.0808 | 232.0 | 756.0 | 1.3235 | 16.9755 | 58.909 | 7.364 | 189.0 | 250.0 | |
|
| 3000 | 0.1212 | 180.0 | 628.0 | 1.1620 | 16.9923 | 58.85 | 7.356 | 149.0 | 171.0 | |
|
| 4000 | 0.1616 | 150.0 | 576.0 | 1.0434 | 16.9803 | 58.892 | 7.361 | 121.5 | 172.0 | |
|
| 5000 | 0.2020 | 130.0 | 504.0 | 0.9520 | 17.0128 | 58.779 | 7.347 | 100.5 | 144.0 | |
|
| 6000 | 0.2424 | 113.5 | 420.0 | 0.8702 | 17.0074 | 58.798 | 7.35 | 91.0 | 137.0 | |
|
| 7000 | 0.2828 | 106.0 | 408.0 | 0.8100 | 16.9821 | 58.885 | 7.361 | 80.5 | 160.0 | |
|
| 8000 | 0.3232 | 96.5 | 396.0 | 0.7421 | 16.9749 | 58.911 | 7.364 | 70.5 | 127.0 | |
|
| 9000 | 0.3636 | 84.0 | 342.0 | 0.6877 | 16.9969 | 58.834 | 7.354 | 69.5 | 217.0 | |
|
| 10000 | 0.4040 | 78.0 | 300.0 | 0.6467 | 16.9846 | 58.877 | 7.36 | 65.0 | 139.0 | |
|
| 11000 | 0.4444 | 77.0 | 278.0 | 0.5957 | 16.9903 | 58.857 | 7.357 | 60.0 | 127.5 | |
|
| 12000 | 0.4848 | 75.0 | 272.0 | 0.5789 | 16.9858 | 58.873 | 7.359 | 56.5 | 140.0 | |
|
| 13000 | 0.5253 | 71.5 | 266.0 | 0.5525 | 16.9418 | 59.026 | 7.378 | 56.5 | 116.0 | |
|
| 14000 | 0.5657 | 71.0 | 252.0 | 0.5416 | 17.088 | 58.521 | 7.315 | 53.75 | 132.0 | |
|
| 15000 | 0.6061 | 68.0 | 221.0 | 0.5283 | 16.9524 | 58.989 | 7.374 | 50.25 | 112.5 | |
|
| 16000 | 0.6465 | 70.0 | 244.0 | 0.5200 | 17.0495 | 58.653 | 7.332 | 52.5 | 109.5 | |
|
| 17000 | 0.6869 | 67.0 | 225.0 | 0.5097 | 17.0223 | 58.747 | 7.343 | 51.5 | 109.0 | |
|
| 18000 | 0.7273 | 71.0 | 239.0 | 0.5016 | 17.0519 | 58.644 | 7.331 | 49.5 | 150.0 | |
|
| 19000 | 0.7677 | 68.0 | 212.0 | 0.4887 | 17.0831 | 58.537 | 7.317 | 51.25 | 98.0 | |
|
| 20000 | 0.8081 | 65.0 | 211.0 | 0.4865 | 17.0098 | 58.789 | 7.349 | 49.0 | 101.5 | |
|
| 21000 | 0.8485 | 64.5 | 217.0 | 0.4791 | 17.0253 | 58.736 | 7.342 | 47.5 | 142.0 | |
|
| 22000 | 0.8889 | 66.5 | 230.0 | 0.4798 | 16.9954 | 58.839 | 7.355 | 48.5 | 147.0 | |
|
| 23000 | 0.9293 | 62.5 | 212.0 | 0.4675 | 16.9835 | 58.881 | 7.36 | 45.5 | 134.0 | |
|
| 24000 | 0.9697 | 63.5 | 220.0 | 0.4712 | 16.9973 | 58.833 | 7.354 | 47.0 | 138.0 | |
|
| 24750 | 1.0 | 63.75 | 247.0 | 0.4679 | 17.0597 | 58.618 | 7.327 | 45.75 | 205.0 | |
|
|
|
### Framework versions |
|
- Distily 0.2.0 |
|
- Transformers 4.44.0 |
|
- Pytorch 2.3.0 |
|
- Datasets 2.21.0 |
|
|
