69 lines
2.2 KiB
Markdown
69 lines
2.2 KiB
Markdown
---
|
||
language: en
|
||
license: apache-2.0
|
||
tags:
|
||
- language-model
|
||
- slt
|
||
- singular-learning-theory
|
||
- patterning
|
||
- timaeus
|
||
datasets:
|
||
- HuggingFaceFW/fineweb
|
||
library_name: transformers
|
||
pipeline_tag: text-generation
|
||
---
|
||
|
||
# Dodecahedron-67M
|
||
|
||
A 67M parameter language model from the Dodecahedron family, trained for singular learning theory (SLT) research.
|
||
|
||
## Model Details
|
||
|
||
| Parameter | Value |
|
||
|-----------|-------|
|
||
| Parameters | 67M |
|
||
| Architecture | LlamaForCausalLM |
|
||
| Hidden size | 512 |
|
||
| Layers | 20 |
|
||
| Attention heads | 8 (4 KV heads, GQA 2:1) |
|
||
| Head dim | 64 |
|
||
| MLP intermediate | 1408 |
|
||
| Vocab size | 16,384 |
|
||
| Context length | 2,048 |
|
||
| Tied embeddings | Yes |
|
||
|
||
## Training
|
||
|
||
- **Data**: 40B tokens of [FineWeb](https://huggingface.co/datasets/HuggingFaceFW/fineweb)
|
||
- **Tokenizer**: Custom 16k vocab BPE trained on FineWeb (same as Dodecahedron-32M)
|
||
- **Optimizer**: AdamW (betas 0.9/0.95, weight decay 0.01, grad clip 1.0)
|
||
- **Learning rate**: 3.5e-3 with cosine decay, 1000 warmup steps
|
||
- **Batch size**: 524k tokens/step (64 micro × 4 grad accum × 2048 seq)
|
||
- **Precision**: bfloat16
|
||
- **Hardware**: 8× H100
|
||
|
||
## Checkpoints
|
||
|
||
53 checkpoints are available as branches, log-spaced from step 1 to step 9536 (~40B tokens):
|
||
|
||
```python
|
||
from transformers import AutoModelForCausalLM
|
||
|
||
# Final checkpoint
|
||
model = AutoModelForCausalLM.from_pretrained("timaeus/dodecahedron-67m-v2")
|
||
|
||
# Intermediate checkpoint
|
||
model = AutoModelForCausalLM.from_pretrained("timaeus/dodecahedron-67m-v2", revision="step5108")
|
||
```
|
||
|
||
Available steps: 1, 2, 3, 4, 5, 6, 7, 8, 10, 12, 14, 16, 19, 22, 25, 30, 35, 40, 47, 55, 65, 75, 88, 103, 121, 141, 165, 193, 225, 263, 308, 359, 420, 491, 574, 671, 785, 917, 1072, 1253, 1465, 1713, 2002, 2340, 2735, 3198, 3738, 4369, 5108, 5970, 6979, 8158, 9536
|
||
|
||
## Purpose
|
||
|
||
The Dodecahedron family is designed as small, well-characterized reference models for developmental interpretability and SLT research. Having densely-checkpointed models enables studying learning dynamics, phase transitions, and the geometry of the loss landscape throughout training.
|
||
|
||
## Related
|
||
|
||
- [Dodecahedron-32M](https://huggingface.co/timaeus/dodecahedron-32m-v2) — same architecture scaled down
|
||
- [Timaeus](https://timaeus.co) — research organization
|