---
library_name: transformers
license: apache-2.0
tags: [ablation, sparse-attention, dsa, custom_code]
---
# Standard attention ablation model (1.50M params, 3072 context)

Part of a controlled ablation: **BananaMind/dsa-model-ablation** (DSA) vs **BananaMind/normal-model-ablation** (standard attention).
Same seed, data order, tokenizer, architecture and hyper-parameters. Only the attention differs.

- Attention: Standard dense causal multi-head attention.
- Architecture: 5 layers, hidden 128, 4 heads, SwiGLU 336, RoPE, RMSNorm, tied embeddings, vocab 4096 (custom BPE)
- Context length: 3072
- Params: 1,498,496
- Data: HuggingFaceFW/fineweb-edu (sample-10BT), 2000M tokens, 1 epoch, seed 1337
- Optimiser: AdamW lr 0.002 (cosine), batch 16x3072 tokens, bf16 autocast

## Usage
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "BananaMind/normal-model-ablation"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(repo, trust_remote_code=True)
```
(No KV cache is implemented; `generate` recomputes the full sequence each step.)

## Benchmarks (0-shot, lm-evaluation-harness; acc / acc_norm where available)
| Model | Params | Val loss | Val ppl | piqa | hellaswag | arc_easy | arc_challenge |
|---|---|---|---|---|---|---|---|
| DSA | 1,530,496 | 3.4150 | 30.42 | 54.90 / 53.05 | 26.84 / 26.72 | 30.18 / 31.02 | 16.98 / 20.73 |
| Standard | 1,498,496 | 3.4248 | 30.72 | 53.81 / 53.75 | 26.86 / 26.75 | 30.68 / 31.44 | 17.83 / 20.99 |

Per-model raw results: `benchmarks/results.json`. Comparison with the other model: `benchmarks/comparison.md`
(other repo: BananaMind/dsa-model-ablation). Training curve: `train_log.json`.

Note: at this size models are close to chance on these benchmarks, and the benchmark prompts are shorter than
the top-k (512), where DSA selects every token and behaves identically to dense attention at inference,
so differences mostly reflect how training differed.