GitLab

- Introduced specialized DispatchInstruction templates for fp8_e4_t and fp8_e5_t types, enhancing support for new data formats in CUDA GEMM operations.
- Each specialization defines the corresponding MMA and MMA_Group types, optimizing performance for specific configurations.