AI/ML●●Solid
Mamba SSM in Rust – training and inference with custom CUDA kernels
Custom CUDA kernels for SSM recurrence with zero framework dependencies.
WizardryNiche Gem
silvermpx
103mo ago
Back to browse
SparseLab–real sparse training(CSR+custom kernel) in PyTorch, CPU-first
by DARSHANFOFADIYA·Apr 24, 2026·1 point·1 comment
AI Analysis
●●SolidNiche GemBig Brain
Custom CPU kernels for sparse training when everyone else chases GPU.
Strengths
- •CSR format with custom kernels avoids GPU dependency for sparse workloads.
- •CPU-first approach targets edge cases where GPU memory is prohibitive.
- •Native PyTorch integration means no framework switching for existing users.
Weaknesses
- •No benchmark comparisons against PyTorch's built-in sparse support or DeepSpeed.
- •CPU training will be significantly slower than GPU for most practical model sizes.
Category
Target Audience
ML researchers and PyTorch developers working with sparse models or CPU-only environments
Similar To
PyTorch Sparse · DeepSpeed · Fairscale
Similar Projects
AI/ML●●Solid
Neural Abyss – PyTorch multi-agent combat simulator
Per-agent PPO runtime with tensor-first simulation state is genuinely clever architecture.
Big BrainNiche Gem
luthor190397
103mo ago
AI/ML●●Solid
A 178K Neural Net that beats Pokémon Roguelike
178K neural net beats Pokémon roguelike with clever 1386-dim state encoding.
Rabbit HoleNiche GemCozy
farcaster
3013d ago
AI/ML●Mid
Neural Siege – A Multi-Agent RL Combat Simulation
Vectorized multi-agent RL combat sim with deterministic checkpointing and telemetry logging.
Big BrainWizardry
luthor190397
203mo ago
Infrastructure●●Solid
MCCL Distributed PyTorch training across MacBooks via Thunderbolt
Two MacBooks syncing gradients over Thunderbolt — slower than single-GPU but it works.
WizardryNiche GemShip It
sassoshots44
103mo ago
AI/ML●●●Banger
Auto GPU Kernel – Autonomous GPU-kernel discovery and optimizer
Autonomous kernel optimizer that won MLSys contest with 34.93x speedup.
WizardryBig BrainBold Bet
dogacel
1026d ago