---

license: odc-by

datasets:

- HuggingFaceFW/fineweb-edu

language:

- en

library_name: transformers

pipeline_tag: text-generation

tags:

- nanodex

- tiny-lm

- pretrained-from-scratch

---

nanodex-500k-200m

A 492,192-parameter decoder-only language model pre-trained

from scratch on fineweb-edu,

using the NanoDex Trainer Space.

Architecture

A standard LlamaForCausalLM decoder-only transformer — SiLU MLP, RMSNorm,

rotary position embeddings, grouped-query attention, tied embeddings, no biases —

scaled down in width and depth to fit the parameter budget.

| | |

|---|---|

| Parameters | 492,192 |

| Hidden size | 96 |

| Layers | 3 |

| Attention heads | 6 (KV: 2) |

| FFN size | 256 |

| Context length | 512 |

| Vocab | 2,048 (custom BPE trained on fineweb-edu) |

Training

| | |

|---|---|

| Tokens seen | 199,884,800 |

| Steps | 1,525 |

| Tokens / step | 131,072 |

| Optimizer | AdamW(0.9, 0.95) wd=0.1 clip=1.0 |

| LR schedule | warmup 2% + cosine to 10% (peak 4e-03) |

| Final loss | 3.8150 (ppl 45.4) |

| Wall time | 4.5 min |

| Trained by | @GGUFGuy |

Usage

from transformers import AutoModelForCausalLM, AutoTokenizer

tok = AutoTokenizer.from_pretrained("GGUFGuy/nanodex-500k-200m")
model = AutoModelForCausalLM.from_pretrained("GGUFGuy/nanodex-500k-200m")

ids = tok("The mitochondria is", return_tensors="pt").input_ids
print(tok.decode(model.generate(ids, max_new_tokens=60, do_sample=True,
                                temperature=0.8, top_k=50)[0]))

Caveats

This is a nano-scale research artifact. At this parameter count and token

budget the model learns word shapes, common collocations and a little syntax —

it is not a useful assistant and its output is not factual. It exists to make

"pre-train a transformer from scratch" something you can actually watch happen.