Decision Workspace
oxicuda-train vs oxicuda-infer vs burn-cuda
Side-by-side comparison of Rust crates
47
oxicuda-train
experimentalv0.5.5
GPU-accelerated training engine: fused optimizer update kernels, gradient checkpointing, LR schedulers, and ZeRO optimizer state sharding
48
oxicuda-infer
experimentalv0.5.5
High-performance GPU inference engine for OxiCUDA: PagedAttention KV cache, continuous batching, speculative decoding, structured sampling, and quantized inference.
74
burn-cuda
growingv0.22.0
CUDA backend for the Burn framework
Core Metrics
| oxicuda-train | oxicuda-infer | burn-cuda | |
|---|---|---|---|
| Health Score | 47 | 48 | 74 |
| Total Downloads | 336 | 338 | 1.3M |
| 30d Downloads | 0 | 0 | 0 |
| Dependents | 0 | 0 | 368 |
| Releases | 19 | 19 | 28 |
| Last Updated | 57d ago | 57d ago | 3d ago |
| Age | 6m | 6m | 2y 1m |
Health Breakdown
oxicuda-train
Maintenance
17
Quality
9
Community
6
Popularity
3
Documentation
12
oxicuda-infer
Maintenance
17
Quality
10
Community
6
Popularity
3
Documentation
12
burn-cuda
Maintenance
20
Quality
17
Community
17
Popularity
7
Documentation
13
Technical Details
| oxicuda-train | oxicuda-infer | burn-cuda | |
|---|---|---|---|
| Version | 0.5.5 | 0.5.5 | 0.22.0 |
| Stable (≥1.0) | ✗ No | ✗ No | ✗ No |
| License | Apache-2.0 | Apache-2.0 | MIT OR Apache-2.0 |
| Dependencies | 9 | 3 | 4 |
| Crate Size | 134KB | 146KB | 37KB |
| Features | 3 | 1 | 7 |
| Yanked % | 10.5% | 10.5% | 0.0% |
| Edition | 2024 | 2024 | 2024 |
| MSRV | 1.85 | 1.85 | — |
| Owners | 1 | 1 | 3 (team) |
Links
Quick Verdict
- •burn-cuda leads with a health score of 74/100, but none of the options score above 80.
- •burn-cuda is depended on by 368 crates — strongest ecosystem trust.