Skip to content

[Infer] Change groupwise weight quant from cpu to gpu for deepseek_v2 model #10174

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
wants to merge 10 commits into
base: develop
Choose a base branch
from
Open

[Infer] Change groupwise weight quant from cpu to gpu for deepseek_v2 model #10174

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
10 commits
Select commit Hold shift + click to select a range
a63fa39
change groupwise weight quant from cpu to gpu for deepseek_v2 model
zeroRains Mar 18, 2025
e789537
fix conflict
zeroRains Mar 27, 2025
e16f4e2
Merge branch 'develop' of https://github.com/PaddlePaddle/PaddleNLP i…
zeroRains Apr 20, 2025
8d0a39a
support group-wise weight quant for qwen2 and change cpu kernel to gp…
zeroRains Apr 20, 2025
1aae565
support moe group_wise weight quant
zeroRains Apr 20, 2025
09ab2f3
fix
zeroRains Apr 20, 2025
821e4d6
Merge branch 'develop' of https://github.com/PaddlePaddle/PaddleNLP i…
zeroRains May 2, 2025
ed47c7e
add the iterscale for fused_moe
zeroRains May 9, 2025
bdf7fe0
fix the conflict
zeroRains May 9, 2025
97ed730
fix the search call
zeroRains May 14, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
272 changes: 192 additions & 80 deletions csrc/gpu/moe/fused_moe/cutlass_kernels/moe_gemm/fused_moe_cutlass_kernel.h
View file
Open in desktop

Large diffs are not rendered by default.

8 changes: 6 additions & 2 deletions csrc/gpu/moe/fused_moe/cutlass_kernels/moe_gemm/fused_moe_gemm_kernels.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -37,6 +37,7 @@ class MoeGemmRunner {
int64_t gemm_k,
int num_experts,
std::string activation_type,
const int32_t weightonly_group_size,
cudaStream_t stream);

void moe_gemm(const T* A,
Expand All @@ -48,10 +49,11 @@ class MoeGemmRunner {
int64_t gemm_n,
int64_t gemm_k,
int num_experts,
int group_size,
cudaStream_t stream);

private:
template <typename EpilogueTag>
template <typename EpilogueTag, bool FineGrained>
void dispatch_to_arch(const T* A,
const WeightType* B,
const T* weight_scales,
Expand All @@ -62,11 +64,12 @@ class MoeGemmRunner {
int64_t gemm_n,
int64_t gemm_k,
int num_experts,
int group_size,
CutlassGemmConfig gemm_config,
cudaStream_t stream,
int* occupancy = nullptr);

template <typename EpilogueTag>
template <typename EpilogueTag, bool FineGrained>
void run_gemm(const T* A,
const WeightType* B,
const T* weight_scales,
Expand All @@ -77,6 +80,7 @@ class MoeGemmRunner {
int64_t gemm_n,
int64_t gemm_k,
int num_experts,
int group_size,
cudaStream_t stream);

private:
Expand Down
247 changes: 173 additions & 74 deletions csrc/gpu/moe/fused_moe/cutlass_kernels/moe_gemm/fused_moe_gemm_kernels_template.h
View file
Open in desktop

Large diffs are not rendered by default.

6 changes: 6 additions & 0 deletions csrc/gpu/moe/fused_moe/fused_moe.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,6 +57,7 @@ void FusedMoeKernel(const paddle::Tensor& input,
const paddle::optional<paddle::Tensor>& ffn2_scale,
const paddle::optional<paddle::Tensor>& ffn2_bias,
const std::string& quant_method,
const int weightonly_group_size,
const int moe_topk,
const bool group_moe,
const bool norm_topk_prob,
Expand Down Expand Up @@ -86,6 +87,7 @@ void FusedMoeKernel(const paddle::Tensor& input,
ffn2_scale ? ffn2_scale.get_ptr() : nullptr,
ffn2_bias ? ffn2_bias.get_ptr() : nullptr,
nullptr,
weightonly_group_size,
moe_topk,
group_moe,
norm_topk_prob,
Expand All @@ -105,6 +107,7 @@ std::vector<paddle::Tensor> FusedExpertMoe(
const paddle::optional<paddle::Tensor>& ffn2_bias,
const paddle::optional<paddle::Tensor>& ffn2_scale,
const std::string& quant_method,
const int weightonly_group_size,
const int moe_topk,
const bool norm_topk_prob,
const bool group_moe) {
Expand All @@ -122,6 +125,7 @@ std::vector<paddle::Tensor> FusedExpertMoe(
ffn2_scale,
ffn2_bias,
quant_method,
weightonly_group_size,
moe_topk,
group_moe,
norm_topk_prob,
Expand All @@ -137,6 +141,7 @@ std::vector<paddle::Tensor> FusedExpertMoe(
ffn2_scale,
ffn2_bias,
quant_method,
weightonly_group_size,
moe_topk,
group_moe,
norm_topk_prob,
Expand Down Expand Up @@ -184,6 +189,7 @@ PD_BUILD_OP(fused_expert_moe)
paddle::Optional("ffn2_scale")})
.Outputs({"output"})
.Attrs({"quant_method:std::string",
"weightonly_group_size:int",
"moe_topk:int",
"norm_topk_prob:bool",
"group_moe:bool"})
Expand Down
7 changes: 7 additions & 0 deletions csrc/gpu/moe/fused_moe/moe/fused_moe_helper.h
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -126,6 +126,7 @@ class MoeHelper {
const paddle::Tensor *ffn2_scale,
const paddle::Tensor *ffn2_bias,
const paddle::Tensor *moe_token_type_ids,
const int weightonly_group_size,
const int moe_topk,
const bool group_moe,
const bool norm_topk_prob,
Expand Down Expand Up @@ -304,6 +305,7 @@ class MoeHelper {
hidden_size,
num_experts,
"none",
weightonly_group_size,
stream);
} else if (gemm_method_ == "weight_only_int4") {
int4_moe_gemm_runner_->moe_gemm_bias_act(
Expand All @@ -319,6 +321,7 @@ class MoeHelper {
hidden_size,
num_experts,
"none",
weightonly_group_size,
stream);
} else {
fp16_moe_gemm_runner_->moe_gemm_bias_act(
Expand All @@ -333,6 +336,7 @@ class MoeHelper {
hidden_size,
num_experts,
"none",
weightonly_group_size,
stream);
}

Expand All @@ -356,6 +360,7 @@ class MoeHelper {
hidden_size,
inter_size / 2,
num_experts,
weightonly_group_size,
stream);
} else if (gemm_method_ == "weight_only_int4") {
int4_moe_gemm_runner_->moe_gemm(
Expand All @@ -369,6 +374,7 @@ class MoeHelper {
hidden_size,
inter_size / 2,
num_experts,
weightonly_group_size,
stream);
} else {
fp16_moe_gemm_runner_->moe_gemm(
Expand All @@ -381,6 +387,7 @@ class MoeHelper {
hidden_size,
inter_size / 2,
num_experts,
weightonly_group_size,
stream);
}

Expand Down
30 changes: 26 additions & 4 deletions csrc/gpu/moe/fused_moe/moe_ffn.cu
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
const paddle::optional<paddle::Tensor>& ffn1_scale,
const paddle::optional<paddle::Tensor>& ffn2_scale,
const std::string& quant_method,
const int32_t weightonly_group_size,
paddle::Tensor ffn_out) {
typedef PDTraits<T> traits_;
typedef typename traits_::DataType DataType_;
Expand Down Expand Up @@ -62,8 +63,16 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
ffn1_bias
? const_cast<paddle::Tensor*>(ffn1_bias.get_ptr())->data<data_t>()
: nullptr;

/*
group size
different quant and no quant, no quant and quant channel wise have same group
size no quant : group_size = -1
quant channel wise : group_size = -1
quant group wise : group_size = 64 || 128
*/
int group_size = weightonly_group_size;
if (quant_method == "weight_only_int8") {
group_size = group_size == -1 ? hidden_size : group_size;
int8_moe_gemm_runner.moe_gemm_bias_act(
reinterpret_cast<const NvType*>(permuted_data),
reinterpret_cast<const uint8_t*>(ffn1_weight.data<int8_t>()),
Expand All @@ -77,8 +86,10 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
hidden_size,
num_experts,
"none",
group_size,
stream);
} else if (quant_method == "weight_only_int4") {
group_size = group_size == -1 ? hidden_size : group_size;
int4_moe_gemm_runner.moe_gemm_bias_act(
reinterpret_cast<const NvType*>(permuted_data),
reinterpret_cast<const cutlass::uint4b_t*>(ffn1_weight.data<int8_t>()),
Expand All @@ -92,6 +103,7 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
hidden_size,
num_experts,
"none",
group_size,
stream);
} else {
fp16_moe_gemm_runner.moe_gemm_bias_act(
Expand All @@ -106,13 +118,16 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
hidden_size,
num_experts,
"none",
-1,
stream);
}

auto act_out_tensor = paddle::experimental::swiglu(fc1_out_tensor, nullptr);
auto act_out = act_out_tensor.data<data_t>();

// reset group_size
group_size = weightonly_group_size;
if (quant_method == "weight_only_int8") {
group_size = group_size == -1 ? inter_size / 2 : group_size;
int8_moe_gemm_runner.moe_gemm(
reinterpret_cast<const NvType*>(act_out),
reinterpret_cast<const uint8_t*>(ffn2_weight.data<int8_t>()),
Expand All @@ -124,9 +139,11 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
hidden_size,
inter_size / 2,
num_experts,
group_size,
stream);

} else if (quant_method == "weight_only_int4") {
group_size = group_size == -1 ? inter_size / 2 : group_size;
int4_moe_gemm_runner.moe_gemm(
reinterpret_cast<const NvType*>(act_out),
reinterpret_cast<const cutlass::uint4b_t*>(ffn2_weight.data<int8_t>()),
Expand All @@ -138,6 +155,7 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
hidden_size,
inter_size / 2,
num_experts,
group_size,
stream);
} else {
fp16_moe_gemm_runner.moe_gemm(
Expand All @@ -150,6 +168,7 @@ void MoeFFNKernel(const paddle::Tensor& permute_input,
hidden_size,
inter_size / 2,
num_experts,
-1,
stream);
}
}
Expand All @@ -162,7 +181,8 @@ std::vector<paddle::Tensor> MoeExpertFFN(
const paddle::optional<paddle::Tensor>& ffn1_bias,
const paddle::optional<paddle::Tensor>& ffn1_scale,
const paddle::optional<paddle::Tensor>& ffn2_scale,
const std::string& quant_method) {
const std::string& quant_method,
const int32_t weightonly_group_size) {
const auto input_type = permute_input.dtype();
auto ffn_out = paddle::empty_like(permute_input);

Expand All @@ -176,6 +196,7 @@ std::vector<paddle::Tensor> MoeExpertFFN(
ffn1_scale,
ffn2_scale,
quant_method,
weightonly_group_size,
ffn_out);
break;
case paddle::DataType::FLOAT16:
Expand All @@ -187,6 +208,7 @@ std::vector<paddle::Tensor> MoeExpertFFN(
ffn1_scale,
ffn2_scale,
quant_method,
weightonly_group_size,
ffn_out);
break;
default:
Expand Down Expand Up @@ -226,7 +248,7 @@ PD_BUILD_OP(moe_expert_ffn)
paddle::Optional("ffn1_scale"),
paddle::Optional("ffn2_scale")})
.Outputs({"output_tensor"})
.Attrs({"quant_method:std::string"})
.Attrs({"quant_method:std::string", "weightonly_group_size: int"})
.SetKernelFn(PD_KERNEL(MoeExpertFFN))
.SetInferShapeFn(PD_INFER_SHAPE(MoeExpertFFNInferShape))
.SetInferDtypeFn(PD_INFER_DTYPE(MoeExpertFFNInferDtype));
Loading