README.md

---
base_model: gpt2
datasets:
- wikimedia/wikipedia
library_name: Distily
license: mit
tags:
- bitnet
- 1.58b
- generated_from_trainer
model-index:
- name: distily_miles_projector_experiment
  results: []
---


# Summary

Distilled with [Distily](https://github.com/lapp0/distily) library
using teacher model [gpt2](https://huggingface.co/gpt2)
on dataset [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia).

<!-- 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
-->

# Model Architecture:
- **Architecture**: `GPT2LMHeadModel`
- **Total Parameters**: 124,439,808
- **Data Type (dtype)**: torch.bfloat16
- **Model Size**: 0.24 GB


# Evaluation Metrics Comparison

| step | epoch | enwikippl | frwikippl | loss | runtime | samples_per_second | steps_per_second | tinystoriesppl | zhwikippl |
| :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: | :---: |
| **teacher eval** |  | 36.25 | 77.0 |  |  |  |  | 11.75 | 21.375 |
| 0 | 0 | 1486058684416.0 | 34084860461056.0 | 20.1302 | 40.0525 | 62.418 | 7.815 | 2281701376.0 | 15874199126016.0 |
| 2500 | 0.0404 | 756.0 | 3440.0 | 2.4552 | 40.0832 | 62.37 | 7.809 | 404.0 | 1560.0 |
| 5000 | 0.0808 | 352.0 | 1288.0 | 1.7734 | 42.1208 | 59.353 | 7.431 | 246.0 | 290.0 |
| 7500 | 0.1212 | 227.0 | 688.0 | 1.4859 | 44.2818 | 56.457 | 7.068 | 177.0 | 214.0 |
| 10000 | 0.1616 | 176.0 | 624.0 | 1.2995 | 40.5384 | 61.67 | 7.721 | 129.0 | 225.0 |
| 12500 | 0.2020 | 122.0 | 446.0 | 1.0558 | 43.2882 | 57.752 | 7.231 | 93.5 | 231.0 |
| 15000 | 0.2424 | 102.5 | 412.0 | 0.9530 | 40.2067 | 62.179 | 7.785 | 80.0 | 175.0 |
| 17500 | 0.2828 | 92.0 | 342.0 | 0.8613 | 42.4322 | 58.918 | 7.376 | 77.5 | 165.0 |
| 20000 | 0.3232 | 78.0 | 266.0 | 0.8054 | 42.4876 | 58.841 | 7.367 | 64.5 | 110.0 |
| 22500 | 0.3636 | 66.5 | 228.0 | 0.6962 | 40.1977 | 62.193 | 7.787 | 58.0 | 185.0 |
| 25000 | 0.4040 | 64.0 | 200.0 | 0.6565 | 42.3516 | 59.03 | 7.391 | 52.75 | 115.5 |
| 27500 | 0.4444 | 61.25 | 190.0 | 0.6213 | 42.9602 | 58.193 | 7.286 | 50.75 | 101.0 |
| 30000 | 0.4848 | 62.75 | 211.0 | 0.6318 | 44.9016 | 55.677 | 6.971 | 50.25 | 184.0 |
| 32500 | 0.5253 | 57.5 | 194.0 | 0.6184 | 43.9215 | 56.92 | 7.126 | 50.25 | 89.5 |
| 35000 | 0.5657 | 57.0 | 177.0 | 0.5768 | 42.6805 | 58.575 | 7.334 | 44.0 | 107.0 |
| 37500 | 0.6061 | 54.5 | 168.0 | 0.5596 | 44.1546 | 56.619 | 7.089 | 43.5 | 81.0 |
| 40000 | 0.6465 | 54.0 | 159.0 | 0.5345 | 42.0172 | 59.499 | 7.449 | 42.75 | 77.5 |
| 42500 | 0.6869 | 53.5 | 169.0 | 0.5260 | 41.7231 | 59.919 | 7.502 | 39.5 | 61.25 |
| 45000 | 0.7273 | 48.5 | 152.0 | 0.4414 | 40.3349 | 61.981 | 7.76 | 35.25 | 50.25 |
| 47500 | 0.7677 | 47.25 | 142.0 | 0.4216 | 41.3204 | 60.503 | 7.575 | 34.5 | 44.25 |
| 50000 | 0.8081 | 46.5 | 137.0 | 0.4085 | 43.1383 | 57.953 | 7.256 | 32.25 | 41.25 |
| 52500 | 0.8485 | 46.0 | 141.0 | 0.4018 | 42.0641 | 59.433 | 7.441 | 33.0 | 38.75 |
| 55000 | 0.8889 | 45.0 | 138.0 | 0.3859 | 40.373 | 61.923 | 7.753 | 31.875 | 35.75 |
| 57500 | 0.9293 | 44.75 | 133.0 | 0.3810 | 40.3972 | 61.885 | 7.748 | 31.625 | 36.0 |
| 60000 | 0.9697 | 44.75 | 132.0 | 0.3782 | 42.2203 | 59.213 | 7.413 | 31.625 | 35.5 |
| 61875 | 1.0 | 44.75 | 133.0 | 0.3778 | 44.5224 | 56.151 | 7.03 | 31.5 | 35.5 |

# Resource Usage Comparison

- VRAM Use: 7.7831 GB

# Distillation (Teacher -> Student) Architecture Difference:

- **Architecture**: `GPT2LMHeadModel` -> `GPT2LMHeadModel`
- **Total Parameters**: 124,439,808 -> 124,439,808
- **Data Type (dtype)**: torch.bfloat16 -> torch.bfloat16
- **Model Size**: 0.24 GB -> 0.24 GB

<details>
<summary>Module Diff Details</summary>

```diff

```

</details>
<br/>

# Train Dataset
Trained on 145,697,117 tokens from the [wikimedia/wikipedia](https://huggingface.co/datasets/wikimedia/wikipedia) dataset.

- Num Samples: `247,500`
- Subset: `20231101.en`
- Split: `train`


# Training Objective

```
DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2))
```

# Hyperparameters
The following hyperparameters were used during training:

<details>
<summary>Expand</summary>

- 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: `linear`
- lr_scheduler_warmup_ratio: `0.5`
- num_epochs: `1.0`
- distillation_objective: `DistillationObjective(logits_loss_component=LossComponent(label=logits, weight=1, loss_fn=kl), attn_loss_component=LossComponent(label=attn, weight=5, loss_fn=raw_mse, layer_mapper=layer-2))`
- train_embeddings: `True`
- lr_scheduler: `<torch.optim.lr_scheduler.LambdaLR object at 0x7fd776b0cd90>`
- student_model_name_or_path: `None`
- student_config_name_or_path: `None`
- student_model_config: `None`
- reinitialize_weights: `None`
- copy_teacher_modules: `[('lm_head', False)]`
- student_model_as_bitnet: `True`
- student_model_compile: `False`
- dropout: `None`
- teacher_model_name_or_path: `gpt2`
- teacher_load_in_8bit: `False`
- teacher_load_in_4bit: `False`
- teacher_model_compile: `False`
- dataset_uri: `wikimedia/wikipedia`
- dataset_subset: `20231101.en`
- dataset_split: `train`
- dataset_column_name: `text`
- dataset_sample_size: `250000`
- dataset_test_size: `0.01`
- gradient_accumulation_steps: `1`
- weight_decay: `0.0`
- max_grad_norm: `1.0`
- warmup_ratio: `0.5`
- warmup_steps: `0`
- gradient_checkpointing: `True`

</details>
<br/>


# Framework Versions
- Distily 0.2.0
- Transformers 4.44.2
- Pytorch 2.3.0
- Datasets 2.21.0