Decision Workspace
burn_dsa vs burn-msa vs gllm-kernels
Side-by-side comparison of Rust crates
43
burn_dsa
experimentalv0.21.0
Burn-native dynamic sparse attention reference blocks and kernel boundary
39
burn-msa
experimentalv0.2.4
MiniMax Sparse Attention (MSA) - block-sparse attention with per-GQA-group top-k block selection for Burn
53
gllm-kernels
experimentalv0.1.3
Low-level attention kernels for gllm with CUDA/ROCm support
Core Metrics
| burn_dsa | burn-msa | gllm-kernels | |
|---|---|---|---|
| Health Score | 43 | 39 | 53 |
| Total Downloads | 25 | 169 | 9.6K |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 0 | 3 |
| Releases | 1 | 8 | 4 |
| Last Updated | 136d ago | 64d ago | 270d ago |
| Age | 4m | 2m | 9m |
Health Breakdown
burn_dsa
Maintenance
6
Quality
14
Community
6
Popularity
2
Documentation
15
burn-msa
Maintenance
10
Quality
10
Community
3
Popularity
3
Documentation
13
gllm-kernels
Maintenance
15
Quality
15
Community
8
Popularity
5
Documentation
10
Technical Details
| burn_dsa | burn-msa | gllm-kernels | |
|---|---|---|---|
| Version | 0.21.0 | 0.2.4 | 0.1.3 |
| Stable (≥1.0) | ✗ No | ✗ No | ✗ No |
| License | MIT OR Apache-2.0 | AGPL-3.0 | Apache-2.0 |
| Dependencies | 9 | 10 | 10 |
| Crate Size | 52KB | 78KB | 111KB |
| Features | 4 | 5 | 13 |
| Yanked % | 0.0% | 25.0% | 0.0% |
| Edition | 2024 | 2021 | 2021 |
| MSRV | 1.92 | — | — |
| Owners | 1 | 1 | 1 |
Links
Quick Verdict
- •gllm-kernels leads with a health score of 53/100, but none of the options score above 80.