Views
No views yet
Standard: Q = W_Q·x, K = W_K·x, V = W_V·x [3d² params]
Unified: u = W·x → [seeking|offering|content] [d² params]1import torch
2from huggingface_hub import hf_hub_download
3
4# Download model
5model_path = hf_hub_download(repo_id="Reinforce-ai/yocto", filename="model.pt")
6tokenizer_path = hf_hub_download(repo_id="Reinforce-ai/yocto", filename="tokenizer.json")
7
8# Load and generate (see GitHub for full code)| Metric | Value |
|---|---|
| Parameters | 484,272 |
| Size (fp16) | 946 KB |
| Attention share | 5.7% |
| Perplexity | 9.58 |
| Speed (CPU) | 700+ tok/s |
Once upon a time, there was a little girl named Lily. She loved to play with her toys all day long. One day, she found a shiny thing on the shelf. The little girl said, "Look, mommy, look!" Her mommy explained that it's very cool, so Lily and her mommy went to the store to buy some tasty food.
| Component | Value |
|---|---|
| Embedding dim | 72 |
| Layers | 4 |
| Attention heads | 3 |
| FFN dim | 288 |
| Vocab size | 4,000 |
| Context length | 512 |
1@misc{deshwal2026yocto,
2 title={Attention Fields: Unified Projections for Efficient Language Models},
3 author={Deshwal, Viraj},
4 year={2026},
5 url={https://www.reinforceai.com/yocto},
6 howpublished={\url{https://github.com/reinforceai/yocto}}
7}