AI/ML●●Solid
PyTorch on Java
LibTorch bindings bring CUDA and MPS backends to Java with LLaMA-3 inference included.
Niche GemBig Brain
pdsminer
2010d ago
Back to browse
GitHub Repository
Machine Learning
12 starsC#
GPT-2 inference in pure C#, 0 bytes allocated per token
by dev-on-bike·May 17, 2026·1 point·1 comment
AI Analysis
●●●BangerWizardryBig Brain
GPT-2 inference in pure C# allocating zero bytes per token beats ONNX Runtime.
Strengths
- •Zero-allocation hot paths eliminate GC pressure for predictable real-time latency.
- •Matches PyTorch output within 1e-4 without native binaries or Python runtime.
- •KV-cache implementation scales O(N) while keeping managed memory usage flat.
Weaknesses
- •Supports only 14 operators currently, limiting complex model architecture imports.
- •Lacks the broad ecosystem and tooling surrounding established ONNX runtimes.
Category
Target Audience
.NET developers needing high-performance local inference
Similar To
ONNX Runtime · TensorFlow.NET · ML.NET
Similar Projects
AI/ML●●Solid
NanoEuler – GPT-2 scale model in pure C/CUDA from scratch
Hand-written FlashAttention and full gradient checks in pure CUDA with no PyTorch.
WizardryBig Brain
vforno
602d ago
AI/ML●●●●Gem
Verified Deep Learning with Lean 4
Formally verifies ResNet and ViT architectures using Lean 4 proofs.
WizardryBig BrainNiche Gem
asparagui
602mo ago
AI/ML●●●Banger
Static-allocation MLP inference in ANSI C using a 2-slot ring buffer
Two-slot ring buffer cuts MLP RAM usage to the practical lower bound on microcontrollers.
Niche GemWizardryBig Brain
xou
4024d ago
Open Source●●Solid
I mirrored war.gov's UAP archive in pure Rail with verifiable bytes
Cryptographic signing of government UFO docs is a clever trust layer.
Big BrainNiche Gem
zem0g
301mo ago
AI/ML●●●Banger
We built an LLM inference engine in pure Python – no PyTorch, no Triton
30x faster cold start than vLLM with zero PyTorch dependencies.
WizardryBig BrainZero to One
zyoraclub
2019d ago