gemm

Modules

modelopt.torch.kernels.quantization.gemm.fp4_kernel	NVFP4 Fake Quantization Triton Implementation.
modelopt.torch.kernels.quantization.gemm.fp4_kernel_hopper	NVFP4 Fake Quantization Triton kernels requiring compute capability >= 8.9 (Hopper+).
modelopt.torch.kernels.quantization.gemm.fp8_kernel	FP8 Triton Kernel Implementations.
modelopt.torch.kernels.quantization.gemm.gptq_fused_kernel	Fused Triton kernels for GPTQ blockwise weight-update.
modelopt.torch.kernels.quantization.gemm.nvfp4_quant	Composable Triton JIT functions for NVFP4 (E2M1) fake quantization.

Triton quantization kernels.