deepseek-ai/DeepGEMM
DeepGEMM
DeepGEMM: clean and efficient FP8 GEMM kernels with fine-grained scaling
Builder
DeepSeek
deepseek-ai • ai-lab
Stars
7,342
Using upstream star count
Forks
1,025
Using upstream fork count
Open Issues
0
Activity Score
0/100
0 commits in 30d
Created
—
README Summary
DeepGEMM is a unified, high-performance tensor core kernel library that brings together the key computation primitives of modern large language models — GEMMs (FP8, FP4, BF16), fused MoE with overlapped communication (Mega MoE), MQA scoring for the lightning indexer, HyperConnection (HC), and more — into a single, cohesive CUDA codebase. All kernels are compiled at runtime via a lightweight Just-In-Time (JIT) module, requiring no CUDA compilation during installation.
Community Evaluation
Loading…AI Dev Skills
No AI dev skills recorded.
Tags
AI SafetyActiveBenchmarkingC++DeepSeekForkedGPU / CUDAPyTorchPython
Taxonomy
Recent Activity
Updated 1 months ago
7 Days
0
30 Days
0
90 Days
2
Quality
Quality signals are not available for this repo yet.
Categories
Evals & BenchmarkingPrimaryInference & ServingSecurity & SafetyModel TrainingSafety & AlignmentOther AI / ML
PM Skills
Safety & AlignmentData & Evaluation
Languages
Cuda100.0%
Timeline
- Project created
- —
- Forked
- Apr 21, 2026
- Your last push
- 1 months ago
- Upstream last push
- 25 days ago
- Tracked since
- Apr 17, 2026
Similar Repos
pgvector cosine similarity · $0
Loading…