Downloads · 30 days
2
8% of all-time downloads
cayley-10b-k16-2l-attn_in
markhenry/cayley-10b-k16-2l-attn_in
cayley-10b-k16-2l-attn_in is a machine learning model from markhenry. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Trained by @amack (aemack-org). Previously uploaded as aemack-org/cayley-10b.
All-time downloads
25
Public
Repo size
978 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.pt978 MB · 100%
At a glance
- Access
- Public
- Created
- Mar 30, 2026
- Updated
- Mar 30, 2026
- SHA
- 87258d1c
Model card
From the Hugging Face model README
cayley-10b-2l-attn_in
Trained by @amack (aemack-org). Previously uploaded as aemack-org/cayley-10b.
Architecture
- 205M-param GPT (12 layers, 8 heads, d=1024)
- CayleySAE inserted at
attn_inin every block (2-level hierarchy) - Trained on FineWeb-Edu-10B for 16k iters (~10.5B tokens)
CayleySAE config
| Level | n | k | delta | Features |
|---|---|---|---|---|
| L0 | 10 | 16 | 0 (root) | 1,024 |
| L1 | 15 | 32 | 256 | 32,768 |
- Total features per layer: 33,792
- Active per token: 48
per_parent_budget=True,score_standardize=True- Location:
attn_in(pre-attention in each block)
Training config
optimizer: muon (lr=0.006), adamw (lr=0.006)
lr_schedule: linear_warmdown, warmdown_frac=0.2
batch_size: 80, seq_len: 1024, gradient_accumulation_steps: 1
max_iters: 16000
- Created
- Mar 30, 2026
- Updated
- Mar 30, 2026