[Refactor] Rename gptq_marlin to marlin to match MoE (#32952)
Signed-off-by: mgoin <mgoin64@gmail.com>
This commit is contained in:
@@ -3,8 +3,8 @@
|
||||
#define MARLIN_NAMESPACE_NAME marlin_moe_wna16
|
||||
#endif
|
||||
|
||||
#include "quantization/gptq_marlin/marlin.cuh"
|
||||
#include "quantization/gptq_marlin/marlin_dtypes.cuh"
|
||||
#include "quantization/marlin/marlin.cuh"
|
||||
#include "quantization/marlin/marlin_dtypes.cuh"
|
||||
#include "core/scalar_type.hpp"
|
||||
|
||||
#define MARLIN_KERNEL_PARAMS \
|
||||
|
||||
@@ -23,10 +23,10 @@
|
||||
#define MARLIN_NAMESPACE_NAME marlin_moe_wna16
|
||||
#endif
|
||||
|
||||
#include "quantization/gptq_marlin/marlin.cuh"
|
||||
#include "quantization/gptq_marlin/marlin_dtypes.cuh"
|
||||
#include "quantization/gptq_marlin/dequant.h"
|
||||
#include "quantization/gptq_marlin/marlin_mma.h"
|
||||
#include "quantization/marlin/marlin.cuh"
|
||||
#include "quantization/marlin/marlin_dtypes.cuh"
|
||||
#include "quantization/marlin/dequant.h"
|
||||
#include "quantization/marlin/marlin_mma.h"
|
||||
#include "core/scalar_type.hpp"
|
||||
|
||||
#define STATIC_ASSERT_SCALAR_TYPE_VALID(scalar_t) \
|
||||
|
||||
@@ -7,7 +7,7 @@
|
||||
#include <cuda_fp16.h>
|
||||
#include <cuda_bf16.h>
|
||||
#include <iostream>
|
||||
#include "../gptq_marlin/marlin_dtypes.cuh"
|
||||
#include "../marlin/marlin_dtypes.cuh"
|
||||
using marlin::MarlinScalarType2;
|
||||
|
||||
namespace allspark {
|
||||
|
||||
@@ -46,7 +46,7 @@ __global__ void permute_cols_kernel(int4 const* __restrict__ a_int4_ptr,
|
||||
|
||||
} // namespace marlin
|
||||
|
||||
torch::Tensor gptq_marlin_gemm(
|
||||
torch::Tensor marlin_gemm(
|
||||
torch::Tensor& a, std::optional<torch::Tensor> c_or_none,
|
||||
torch::Tensor& b_q_weight,
|
||||
std::optional<torch::Tensor> const& b_bias_or_none, torch::Tensor& b_scales,
|
||||
@@ -528,7 +528,7 @@ void marlin_mm(const void* A, const void* B, void* C, void* C_tmp, void* b_bias,
|
||||
|
||||
} // namespace marlin
|
||||
|
||||
torch::Tensor gptq_marlin_gemm(
|
||||
torch::Tensor marlin_gemm(
|
||||
torch::Tensor& a, std::optional<torch::Tensor> c_or_none,
|
||||
torch::Tensor& b_q_weight,
|
||||
std::optional<torch::Tensor> const& b_bias_or_none, torch::Tensor& b_scales,
|
||||
@@ -856,5 +856,5 @@ torch::Tensor gptq_marlin_gemm(
|
||||
#endif
|
||||
|
||||
TORCH_LIBRARY_IMPL_EXPAND(TORCH_EXTENSION_NAME, CUDA, m) {
|
||||
m.impl("gptq_marlin_gemm", &gptq_marlin_gemm);
|
||||
m.impl("marlin_gemm", &marlin_gemm);
|
||||
}
|
||||
@@ -303,9 +303,9 @@ TORCH_LIBRARY_EXPAND(TORCH_EXTENSION_NAME, ops) {
|
||||
ops.def("permute_cols(Tensor A, Tensor perm) -> Tensor");
|
||||
ops.impl("permute_cols", torch::kCUDA, &permute_cols);
|
||||
|
||||
// gptq_marlin Optimized Quantized GEMM for GPTQ.
|
||||
// Marlin Optimized Quantized GEMM (supports GPTQ, AWQ, FP8, NVFP4, MXFP4).
|
||||
ops.def(
|
||||
"gptq_marlin_gemm(Tensor a, Tensor? c_or_none, Tensor b_q_weight, "
|
||||
"marlin_gemm(Tensor a, Tensor? c_or_none, Tensor b_q_weight, "
|
||||
"Tensor? b_bias_or_none,Tensor b_scales, "
|
||||
"Tensor? a_scales, Tensor? global_scale, Tensor? b_zeros_or_none, "
|
||||
"Tensor? "
|
||||
|
||||
Reference in New Issue
Block a user