Open Issues Need Help
View All on GitHub Benchmark the kernels on hardware other than an A100-80GB about 1 hour ago
enhancement help wanted
Proof-of-concept Triton kernels for Forge — SwiGLU, RoPE, Cross-Entropy, LoRA MLP Fused, LoRA QKV Fused, GeGLU
Python
Embedding: dispatch to F.embedding below a dimension threshold about 1 hour ago
enhancement help wanted
Proof-of-concept Triton kernels for Forge — SwiGLU, RoPE, Cross-Entropy, LoRA MLP Fused, LoRA QKV Fused, GeGLU
Python