|
--- |
|
datasets: |
|
- nilq/babylm-10M |
|
language: |
|
- en |
|
--- |
|
|
|
This autoregressive model belongs to a series of rather small language models trained on the [BabyLM](https://babylm.github) data: |
|
- the [baby_llama](https://huggingface.co/bbunzeck/baby_llama) model has few parameters and was trained on a small data set (10M tokens) |
|
- the [**t**eenie_llama](https://huggingface.co/bbunzeck/teenie_llama) model has the same number of parameters but was trained on more **t**okens of text (100M) |
|
- the [**w**eenie_llama](https://huggingface.co/bbunzeck/weenie_llama) model was trained on the small data set, but has more parameters/**w**eights |
|
- the [**tw**eenie_llama](https://huggingface.co/bbunzeck/tweenie_llama) model features both -- more **t**okens (the larger data set) and more **w**eights (*viz.* parameters) |
|
|
|
|
|
| | baby_llama | teenie_llama | weenie_llama | tweenie_llama | |
|
|-----------------|-----------|-------------|-------------|--------------| |
|
| Parameters | 2.97M | 2.97M | 11.44M | 11.44M | |
|
| hidden layers | 8 | 8 | 16 | 16 | |
|
| Attention heads | 8 | 8 | 16 | 16 | |
|
| Embedding size | 128 | 128 | 256 | 256 | |
|
| Context size | 128 | 128 | 256 | 256 | |
|
| Vocab size | 16k | 16k | 16k | 16k | |