Skip to content

P
Paddle

机器未来 / Paddle
与 Fork 源项目一致

Fork自 PaddlePaddle / Paddle

通知 1
Star 1
Fork 0
P
Paddle
未验证 提交 844d9855 编写于 作者: J Jacek Czaja 提交者： GitHub
浏览文件
操作

Replacing set_format with set_mem_desc in FC onednn kernel (#46372) 

* added fc int8 tests

* CI fix

* added skipping UTs for GPUs

* fixes for CI

* added support for residual connections inside fc

* fix for quant int8 bias

* - lint
Co-authored-by: Njakpiase <jakpia21@gmail.com>
上级 806b252c
隐藏空白更改
内联 并排
Showing with 383 addition and 544 deletion
paddle/fluid/operators/mkldnn/fc_mkldnn_op.cc
浏览文件 @ 844d9855
...@@ -16,10 +16,7 @@ limitations under the License. */ ...@@ -16,10 +16,7 @@ limitations under the License. */
#include "paddle/fluid/operators/fc_op.h" #include "paddle/fluid/operators/fc_op.h"
#include "paddle/fluid/platform/mkldnn_helper.h" #include "paddle/fluid/platform/mkldnn_helper.h"
#include "paddle/fluid/platform/mkldnn_reuse.h"
namespace phi {
class DenseTensor;
} // namespace phi
namespace paddle { namespace paddle {
namespace operators { namespace operators {
...@@ -34,388 +31,127 @@ using framework::DDim; ...@@ -34,388 +31,127 @@ using framework::DDim;
using framework::ExecutionContext; using framework::ExecutionContext;
using framework::LoDTensor; using framework::LoDTensor;
using framework::Tensor; using framework::Tensor;
using phi::vectorize;
using platform::GetMKLDNNFormat; using platform::GetMKLDNNFormat;
using platform::MKLDNNDeviceContext; using platform::MKLDNNDeviceContext;
using platform::MKLDNNGetDataType;
using platform::to_void_cast; using platform::to_void_cast;
template <typename T>
constexpr bool IsInt8() {
return std::is_same<T, int8_t>::value || std::is_same<T, uint8_t>::value;
}
template <typename T_in, typename T_w, typename T_out> template <typename T_in, typename T_w, typename T_out>
class FCPrimitiveFactory { class FCMKLDNNHandler
: public platform::MKLDNNHandlerNoCachingT<T_in,
dnnl::inner_product_forward> {
public: public:
explicit FCPrimitiveFactory(const dnnl::engine& engine) : engine_(engine) {} FCMKLDNNHandler(const paddle::framework::ExecutionContext& ctx,
const platform::MKLDNNDeviceContext& dev_ctx,
void ExecuteFcPrimitive(const LoDTensor* input, const Tensor* x,
const Tensor* weights, const Tensor* weights,
const Tensor* bias, const Tensor* bias,
LoDTensor* output, Tensor* out,
const MKLDNNDeviceContext& dev_ctx, const int in_num_col_dims,
const ExecutionContext& ctx) { dnnl::engine mkldnn_engine,
RecomputeOutputDims(ctx, input, weights, output); platform::Place cpu_place)
// If primitive has already been created and cached, don't create new one, : platform::MKLDNNHandlerNoCachingT<T_in, dnnl::inner_product_forward>(
// but update input and output data pointers and return it. mkldnn_engine, cpu_place),
if (fc_) { dev_ctx_(dev_ctx) {
UpdateDataPointers(ctx, output, input); this->memory_key_ = ctx.InputName("W");
this->Execute();
return; auto x_vec_dims = phi::vectorize(x->dims());
} // Otherwise, create a new one. auto weights_vec_dims = phi::vectorize(weights->dims());
auto in_col_dims = ctx.Attr<int>("in_num_col_dims"); int MB = 1;
PADDLE_ENFORCE_LE( for (int i = 0; i < in_num_col_dims; ++i) {
in_col_dims, MB *= x_vec_dims[i];
2,
platform::errors::Unimplemented(
"DNNL FC doesn't support in_num_col_dims parameter to "
"be higher than "
"2."));
if (in_col_dims == 2) {
PADDLE_ENFORCE_EQ(
input->dims().size(),
3,
platform::errors::Unimplemented(
"DNNL FC only supports in_num_col_dims equal to 2 when "
"3 dim input is provided."));
PADDLE_ENFORCE_EQ(
input->format(),
MKLDNNMemoryFormat::ncw,
platform::errors::Unimplemented(
"DNNL FC only supports in_num_col_dims equal to 2 when "
"input format is equal to ncw."));
} }
weights_ = CreateWeightsMemory(weights); int IC = 1;
for (size_t i = in_num_col_dims; i < x_vec_dims.size(); ++i) {
// Since MKL-DNN has a lot of limitations on what the input/weights/output IC *= x_vec_dims[i];
// dimensions should be, to simplify the code, the creation of primitive
// descriptor has been divided into separate cases, based on the number
// of input dimensions.
size_t input_dim_num = input->dims().size();
paddle::optional<dnnl::inner_product_forward::primitive_desc> fc_prim_desc;
memory::desc usr_weights_desc = {};
switch (input_dim_num) {
case 2:
fc_prim_desc =
Create2DFcPrimDescriptor(input, weights, bias, output, ctx);
usr_weights_desc = Create2DUserWeightsDesc();
break;
case 3:
fc_prim_desc =
Create3DFcPrimDescriptor(input, weights, bias, output, ctx);
usr_weights_desc = Create3DUserWeightsDesc(weights);
break;
case 4:
fc_prim_desc =
Create4DFcPrimDescriptor(input, weights, bias, output, ctx);
usr_weights_desc = Create4DUserWeightsDesc(input, weights);
break;
default:
PADDLE_THROW(platform::errors::Unimplemented(
"DNNL FC doesn't support input dims different than 2, 3, 4."));
break;
} }
input_ = CreateMemory<T_in>(fc_prim_desc->src_desc(), input);
// Update weights format inside of its memory
weights_ = Reorder(
usr_weights_desc, usr_weights_desc, weights_->get_data_handle());
// Quantize weights and reorder to format chosen by FC primitive descriptor.
QuantizeWeights(ctx, fc_prim_desc->weights_desc());
bias_ = CreateMemoryToBeCached<float>(fc_prim_desc->bias_desc(), bias); int OC = weights_vec_dims[1];
// If int8 is desired, quantize bias into 32-bit signed int
QuantizeBias(*fc_prim_desc, ctx);
// Store weights and bias in the mkldnn cache dnnl::memory::desc bias_md;
CacheWeightsAndBias(dev_ctx, ctx);
// Based on format determined by inner_product, create output in desired auto src_md = dnnl::memory::desc(
// memory format {MB, IC}, MKLDNNGetDataType<T_in>(), dnnl::memory::format_tag::any);
output_ = CreateDstMemory(*fc_prim_desc, ctx, output); auto weights_md = dnnl::memory::desc(
{OC, IC}, MKLDNNGetDataType<T_w>(), dnnl::memory::format_tag::any);
auto dst_md = dnnl::memory::desc(
{MB, OC}, MKLDNNGetDataType<T_out>(), dnnl::memory::format_tag::any);
if (bias) {
bias_md = dnnl::memory::desc({bias->numel()},
MKLDNNGetDataType<float>(),
dnnl::memory::format_tag::a);
}
// Return MKL-DNN primitive ready to be fed into pipeline and executed dnnl::primitive_attr attrs;
fc_ = inner_product_forward(*fc_prim_desc); HandlePostOps(ctx, &attrs);
this->Execute();
}
void Execute() { this->AcquireForwardPrimitiveDescriptor(attrs,
auto& astream = platform::MKLDNNDeviceContext::tls().get_stream(); prop_kind::forward_inference,
if (bias_) { src_md,
fc_->execute(astream, weights_md,
{{DNNL_ARG_SRC, *input_}, bias_md,
{DNNL_ARG_WEIGHTS, *weights_}, dst_md);
{DNNL_ARG_BIAS, *bias_},
{DNNL_ARG_DST, *output_}});
} else {
fc_->execute(astream,
{{DNNL_ARG_SRC, *input_},
{DNNL_ARG_WEIGHTS, *weights_},
{DNNL_ARG_DST, *output_}});
}
astream.wait();
} }
private: private:
// DNNL always returns 2-dimensional data block as a result of computing void HandlePostOps(const paddle::framework::ExecutionContext& ctx,
// inner product. Hence the format 'nc' is always set for its output dnnl::primitive_attr* attrs) {
// primitive. Therefore, function SetOutputFormat is needed to choose static std::unordered_map<std::string, dnnl::algorithm> algo_map = {
// an appropriate format based on the number of input dimensions and {"relu", dnnl::algorithm::eltwise_relu},
// format of an input tensor. {"gelu", dnnl::algorithm::eltwise_gelu},
void SetOutputFormat(MKLDNNMemoryFormat in_format, Tensor* out) { {"gelu_tanh", dnnl::algorithm::eltwise_gelu_tanh},
int dim_num = out->dims().size(); {"gelu_erf", dnnl::algorithm::eltwise_gelu_erf},
// In case of 2 dims, we set the only possible format, nc {"tanh", dnnl::algorithm::eltwise_tanh},
if (dim_num == 2) { {"sigmoid", dnnl::algorithm::eltwise_logistic},
out->set_format(MKLDNNMemoryFormat::nc); {"hard_swish", dnnl::algorithm::eltwise_hardswish},
out->set_mem_desc({phi::vectorize(out->dims()), {"mish", dnnl::algorithm::eltwise_mish}};
platform::MKLDNNGetDataType<T_out>(),
out->format()});
// In case of 3 dims, we generate a format that is based on number
// of output dims and the layout of input format (nchw or nhwc).
} else if (dim_num == 3) {
if (in_format == MKLDNNMemoryFormat::nwc ||
in_format == MKLDNNMemoryFormat::nhwc) {
out->set_format(
platform::MKLDNNFormatForSize(dim_num, MKLDNNMemoryFormat::nhwc));
} else {
out->set_format(
platform::MKLDNNFormatForSize(dim_num, MKLDNNMemoryFormat::nchw));
}
// In any other case we overwrite the output format with the input one.
} else {
out->set_format(in_format);
}
}
void UpdateDataPointers(const ExecutionContext& ctx, std::vector<float> output_shift_scale;
Tensor* out, float scale = 1.0f;
const Tensor* in) { if (IsInt8<T_w>()) {
input_->set_data_handle(to_void_cast(in->data<T_in>())); std::tie(output_shift_scale, scale) = ComputeOutputShiftScale(ctx);
output_->set_data_handle(out->mutable_data<T_out>(ctx.GetPlace())); int mask = CreateMask(1, output_shift_scale.size() > 1);
// If the primitive exists, but the output tensor has changed its attrs->set_output_scales(mask, output_shift_scale);
// variable, update its format to what has been determined in first
// call to CreateFcPrimitive method.
if (out->format() == MKLDNNMemoryFormat::undef) {
SetOutputFormat(in->format(), out);
} }
}
dnnl::inner_product_forward::primitive_desc Create2DFcPrimDescriptor(
const LoDTensor* input,
const Tensor* weights,
const Tensor* bias,
LoDTensor* output,
const ExecutionContext& ctx) {
auto src_desc = CreateMemDescriptor<T_in>(input, MKLDNNMemoryFormat::any);
auto weight_dims = Get2DWeightDimsForDNNL(weights);
auto weights_desc =
CreateMemDescriptor<T_w>(weight_dims, MKLDNNMemoryFormat::any);
auto bias_desc = CreateMemDescriptor<float>(bias, MKLDNNMemoryFormat::x);
auto dst_desc = CreateMemDescriptor<T_out>(output, MKLDNNMemoryFormat::any);
const auto attrs = CreateFCAttrs(ctx);
return CreateFcPrimDesc(src_desc, weights_desc, bias_desc, dst_desc, attrs);
}
std::vector<int64_t> Get2DWeightDimsForDNNL(const Tensor* weights) { dnnl::post_ops post_ops;
auto dims = phi::vectorize(weights->dims());
std::swap(dims[0], dims[1]); // swap input dim with output dim
return dims;
}
memory::desc Create2DUserWeightsDesc() { return weights_->get_desc(); }
dnnl::inner_product_forward::primitive_desc Create3DFcPrimDescriptor(
const LoDTensor* input,
const Tensor* weights,
const Tensor* bias,
LoDTensor* output,
const ExecutionContext& ctx) {
auto input_dims = phi::vectorize(input->dims());
std::vector<int64_t> new_input_dims = {
input_dims[0] * input_dims[1], input_dims[2], 1};
auto src_desc =
CreateMemDescriptor<T_in>(new_input_dims, MKLDNNMemoryFormat::any);
auto weight_dims = Get3DWeightDimsForDNNL(weights);
auto weights_desc =
CreateMemDescriptor<T_w>(weight_dims, MKLDNNMemoryFormat::any);
auto bias_desc = CreateMemDescriptor<float>(bias, MKLDNNMemoryFormat::x);
auto dst_dims = {input_dims[0] * input_dims[1], weight_dims[0]};
auto dst_desc =
CreateMemDescriptor<T_out>(dst_dims, MKLDNNMemoryFormat::any);
const auto attrs = CreateFCAttrs(ctx);
return CreateFcPrimDesc(src_desc, weights_desc, bias_desc, dst_desc, attrs);
}
std::vector<int64_t> Get3DWeightDimsForDNNL(const Tensor* weights) {
auto paddle_w_dims = phi::vectorize(weights->dims());
return {paddle_w_dims[1], paddle_w_dims[0], 1};
}
memory::desc Create3DUserWeightsDesc(const Tensor* weights) {
auto dims = Get3DWeightDimsForDNNL(weights);
return CreateMemDescriptor<float>(dims, MKLDNNMemoryFormat::oiw);
}
dnnl::inner_product_forward::primitive_desc Create4DFcPrimDescriptor(
const LoDTensor* input,
const Tensor* weights,
const Tensor* bias,
LoDTensor* output,
const ExecutionContext& ctx) {
auto src_desc = CreateMemDescriptor<T_in>(input, MKLDNNMemoryFormat::any);
// Since MKL-DNN doesn't support 4D column-major data formats in
// inner_product primitive, transpose the weights to be in
// row-major format
auto dims = Get4DWeightDimsForDNNL(input, weights);
auto weights_desc = CreateMemDescriptor<T_w>(dims, MKLDNNMemoryFormat::any);
auto bias_desc = CreateMemDescriptor<float>(bias, MKLDNNMemoryFormat::x);
auto dst_desc = CreateMemDescriptor<T_out>(output, MKLDNNMemoryFormat::any);
const auto attrs = CreateFCAttrs(ctx);
return CreateFcPrimDesc(src_desc, weights_desc, bias_desc, dst_desc, attrs);
}
std::vector<int64_t> Get4DWeightDimsForDNNL(const LoDTensor* input, constexpr float sum_scale = 1.0f;
const Tensor* weights) { if (ctx.HasAttr("fuse_residual_connection") &&
auto old_w_dims = phi::vectorize(weights->dims()); ctx.Attr<bool>("fuse_residual_connection")) {
auto old_in_dims = phi::vectorize(input->dims()); post_ops.append_sum(sum_scale);
auto dims = {old_w_dims[1], old_in_dims[1], old_in_dims[2], old_in_dims[3]};
return dims;
}
memory::desc Create4DUserWeightsDesc(const LoDTensor* input,
const Tensor* weights) {
auto dims = Get4DWeightDimsForDNNL(input, weights);
return CreateMemDescriptor<float>(dims, MKLDNNMemoryFormat::oihw);
}
// Convert data from one data format to another
std::shared_ptr<dnnl::memory> Reorder(const memory::desc& src_desc,
const memory::desc& dst_desc,
void* src_data) {
auto src_mem = memory(src_desc, engine_, src_data);
auto dst_mem = std::make_shared<memory>(dst_desc, engine_);
auto reorder = dnnl::reorder(src_mem, *dst_mem);
auto& astream = platform::MKLDNNDeviceContext::tls().get_stream();
{
platform::RecordEvent record_reorder(
"int_reorder",
platform::TracerEventType::UserDefined,
2,
platform::EventRole::kUniqueOp);
reorder.execute(astream, src_mem, *dst_mem);
astream.wait();
} }
return dst_mem; std::string activation_type = ctx.Attr<std::string>("activation_type");
}
// Convert data from one data format to another and rescale it. if (activation_type.empty() == false) {
// If the desired data type is (un)signed int8, quantization occurs here. constexpr float alpha = 0.0f;
std::shared_ptr<dnnl::memory> ReorderWithScale( constexpr float beta = 0.0f;
const std::shared_ptr<memory> src_mem,
const memory::desc& dst_md,
const std::vector<float>& scale_data) {
auto dst_mem = std::make_shared<dnnl::memory>(dst_md, engine_);
dnnl::primitive_attr attributes;
// According to MKL-DNN's documentation mask determines along which
// dimensions should the scale be applied.
// 0 - Single scale applied to whole tensor
// 1 - Apply Scale along a slice of each dimension which index is 1.
// In case of weights quantization, that dimension is output,
// becuase we perform per-output-channel quantization
int mask = CreateMask(0, scale_data.size() > 1);
attributes.set_output_scales(mask, scale_data);
auto reorder = dnnl::reorder(*src_mem, *dst_mem, attributes);
auto& astream = platform::MKLDNNDeviceContext::tls().get_stream(); post_ops.append_eltwise(scale, algo_map[activation_type], alpha, beta);
{
platform::RecordEvent record_reorder(
"int_reorder",
platform::TracerEventType::UserDefined,
2,
platform::EventRole::kUniqueOp);
reorder.execute(astream,
{{DNNL_ARG_FROM, *src_mem}, {DNNL_ARG_TO, *dst_mem}});
astream.wait();
} }
attrs->set_post_ops(post_ops);
return dst_mem;
}
template <typename T>
static dnnl::memory::desc CreateMemDescriptor(
const std::vector<int64_t>& dims, MKLDNNMemoryFormat format) {
return platform::MKLDNNMemDesc(
dims, platform::MKLDNNGetDataType<T>(), format);
}
template <typename T>
static dnnl::memory::desc CreateMemDescriptor(const Tensor* tensor,
MKLDNNMemoryFormat format) {
auto dims = phi::vectorize(tensor->dims());
return CreateMemDescriptor<T>(dims, format);
}
template <typename T>
dnnl::memory CreateMemory(const dnnl::memory::desc& desc,
const Tensor* tensor) {
return CreateMemory(desc, platform::to_void_cast<T>(tensor->data<T>()));
}
dnnl::memory CreateMemory(const dnnl::memory::desc& desc, void* data) {
return memory(desc, engine_, data);
}
template <typename T>
std::shared_ptr<dnnl::memory> CreateMemoryToBeCached(
const dnnl::memory::desc& desc, const Tensor* tensor) {
return CreateMemoryToBeCached(desc,
platform::to_void_cast<T>(tensor->data<T>()));
}
std::shared_ptr<dnnl::memory> CreateMemoryToBeCached(
const dnnl::memory::desc& desc, void* data) {
return std::make_shared<memory>(desc, engine_, data);
}
// Create weights memory and transform to default MKL-DNN format
std::shared_ptr<dnnl::memory> CreateWeightsMemory(const Tensor* weights) {
auto dims = phi::vectorize(weights->dims());
std::swap(dims[0], dims[1]); // Correct output dimensions
auto src_desc = CreateMemDescriptor<float>(dims, MKLDNNMemoryFormat::io);
auto dst_desc = CreateMemDescriptor<float>(dims, MKLDNNMemoryFormat::oi);
// Transpose weights through MKL-DNN's reorder from io to oi format.
return Reorder(src_desc,
dst_desc,
platform::to_void_cast<float>(weights->data<float>()));
}
void CacheWeightsAndBias(const MKLDNNDeviceContext& dev_ctx,
const ExecutionContext& ctx) {
std::string key = platform::CreateKey(dev_ctx);
key = platform::ExtendKeyWithThreadInfoIfNeeded(dev_ctx, key);
const std::string weights_key = key + ctx.InputName("W");
const std::string bias_key = key + ctx.InputName("Bias");
dev_ctx.SetBlob(weights_key, weights_);
dev_ctx.SetBlob(bias_key, bias_);
} }
// Compute the bias scales so that its values correspond to the // Compute the bias scales so that its values correspond to the
// scale of data being an output of weights and input multiplication // scale of data being an output of weights and input multiplication
std::vector<float> ComputeBiasScales(const ExecutionContext& ctx) { std::vector<float> ComputeBiasScales(
auto scale_in_data = ctx.Attr<float>("Scale_in"); const float scale_in, const std::vector<float>& scale_weights) {
auto scale_weights_data = ctx.Attr<std::vector<float>>("Scale_weights"); std::vector<float> bias_scales(scale_weights.size());
const size_t weight_scales_num = scale_weights_data.size();
std::vector<float> bias_scales(weight_scales_num);
#pragma omp parallel for for (size_t i = 0; i < bias_scales.size(); ++i) {
for (size_t i = 0; i < weight_scales_num; i++) { if (scale_weights[i] == 0.0)
if (scale_weights_data[i] == 0.0)
bias_scales[i] = 1.0f; bias_scales[i] = 1.0f;
else else
bias_scales[i] = scale_in_data * scale_weights_data[i]; bias_scales[i] = scale_in * scale_weights[i];
} }
return bias_scales; return bias_scales;
...@@ -444,7 +180,6 @@ class FCPrimitiveFactory { ...@@ -444,7 +180,6 @@ class FCPrimitiveFactory {
const size_t weight_scales_num = scale_weights_data.size(); const size_t weight_scales_num = scale_weights_data.size();
std::vector<float> output_shift_scale(weight_scales_num); std::vector<float> output_shift_scale(weight_scales_num);
#pragma omp parallel for
for (size_t i = 0; i < weight_scales_num; i++) { for (size_t i = 0; i < weight_scales_num; i++) {
if (scale_weights_data[i] == 0.0) if (scale_weights_data[i] == 0.0)
output_shift_scale[i] = inner_scale; output_shift_scale[i] = inner_scale;
...@@ -464,131 +199,218 @@ class FCPrimitiveFactory { ...@@ -464,131 +199,218 @@ class FCPrimitiveFactory {
return is_multi_channel_quantizied ? 1 << slice_dimension : 0; return is_multi_channel_quantizied ? 1 << slice_dimension : 0;
} }
void QuantizeWeights(const ExecutionContext& ctx, memory::desc dst) { std::shared_ptr<dnnl::memory> AcquireMemoryWithReorderAndAttrs(
weights_ = ReorderWithScale( const dnnl::memory::desc& user_md,
weights_, dst, ctx.Attr<std::vector<float>>("Scale_weights")); const dnnl::memory::desc& target_md,
} void* ptr,
const dnnl::primitive_attr& attrs) {
std::shared_ptr<dnnl::memory> target_memory_p;
void QuantizeBias(const inner_product_forward::primitive_desc& fc_prim_desc, auto user_memory_p =
const ExecutionContext& ctx) { std::make_shared<dnnl::memory>(user_md, this->engine_, ptr);
auto bias_scales = ComputeBiasScales(ctx); target_memory_p = std::make_shared<dnnl::memory>(target_md, this->engine_);
bias_ = ReorderWithScale(bias_, fc_prim_desc.bias_desc(), bias_scales); auto reorder_p = std::make_shared<dnnl::reorder>(
} *user_memory_p, *target_memory_p, attrs);
dnnl::primitive_attr CreateFCAttrs(const ExecutionContext& ctx) { auto& astream = platform::MKLDNNDeviceContext::tls().get_stream();
dnnl::primitive_attr attributes; reorder_p->execute(
dnnl::post_ops post_operations; astream,
{{DNNL_ARG_FROM, *user_memory_p}, {DNNL_ARG_TO, *target_memory_p}});
astream.wait();
std::vector<float> output_shift_scale; return target_memory_p;
float scale; }
std::tie(output_shift_scale, scale) = ComputeOutputShiftScale(ctx);
int mask = CreateMask(1, output_shift_scale.size() > 1);
attributes.set_output_scales(mask, output_shift_scale);
float sum_scale = 1.0f; std::string memory_key_;
if (ctx.HasAttr("fuse_residual_connection") && const platform::MKLDNNDeviceContext& dev_ctx_;
ctx.Attr<bool>("fuse_residual_connection")) {
post_operations.append_sum(sum_scale);
}
if (ctx.Attr<std::string>("activation_type") == "relu") { public:
constexpr float negative_slope = 0.0f; std::shared_ptr<dnnl::memory> AcquireSrcMemoryWithReorder(const Tensor* x) {
constexpr float placeholder = 1.0f; // beta const T_in* x_data = x->data<T_in>();
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_relu, negative_slope, placeholder); auto user_md = x->mem_desc();
} else if (ctx.Attr<std::string>("activation_type") == "gelu") { if (x->dims().size() != 2) {
constexpr float alpha = 0.0f; // reshape restrictions are always satisfied because in case of 3 or 4 dim
constexpr float beta = 0.0f; // input, plain layout is enforced
post_operations.append_eltwise( user_md = user_md.reshape(this->fwd_pd_->src_desc().dims());
scale, dnnl::algorithm::eltwise_gelu, alpha, beta);
} else if (ctx.Attr<std::string>("activation_type") == "gelu_tanh") {
constexpr float alpha = 0.0f;
constexpr float beta = 0.0f;
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_gelu_tanh, alpha, beta);
} else if (ctx.Attr<std::string>("activation_type") == "gelu_erf") {
constexpr float alpha = 0.0f;
constexpr float beta = 0.0f;
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_gelu_erf, alpha, beta);
} else if (ctx.Attr<std::string>("activation_type") == "tanh") {
constexpr float alpha = 0.0f;
constexpr float beta = 0.0f;
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_tanh, alpha, beta);
} else if (ctx.Attr<std::string>("activation_type") == "sigmoid") {
constexpr float alpha = 0.0f;
constexpr float beta = 0.0f;
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_logistic, alpha, beta);
} else if (ctx.Attr<std::string>("activation_type") == "mish") {
constexpr float alpha = 0.0f;
constexpr float beta = 0.0f;
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_mish, alpha, beta);
} else if (ctx.Attr<std::string>("activation_type") == "hard_swish") {
constexpr float alpha = 0.0f;
constexpr float beta = 0.0f;
post_operations.append_eltwise(
scale, dnnl::algorithm::eltwise_hardswish, alpha, beta);
} }
attributes.set_post_ops(post_operations); return this->AcquireMemoryWithReorder(
return attributes; user_md, this->fwd_pd_->src_desc(), to_void_cast<T_in>(x_data));
} }
dnnl::inner_product_forward::primitive_desc CreateFcPrimDesc( std::shared_ptr<dnnl::memory> AcquireBiasMemoryWithReorder(
const dnnl::memory::desc& input_desc, const Tensor* bias,
const dnnl::memory::desc& weights_desc, const float scale_in,
const dnnl::memory::desc& bias_desc, const std::vector<float>& scale_weights) {
const dnnl::memory::desc& dst_desc, const float* bias_data = bias->data<float>();
const dnnl::primitive_attr& attrs) {
auto fc_desc = inner_product_forward::desc(prop_kind::forward_scoring, if (IsInt8<T_w>() == false) {
input_desc, // for BF16/FP32 bias is 1D and has no scales, so reorder is not needed
weights_desc, return this->AcquireMemoryFromPrimitive(this->fwd_pd_->bias_desc(),
bias_desc, to_void_cast<float>(bias_data));
dst_desc); } else {
const std::string bias_key = this->memory_key_ + "@bias";
auto memory_p = std::static_pointer_cast<dnnl::memory>(
this->dev_ctx_.GetBlob(bias_key));
if (!memory_p) {
const auto& scale_data = ComputeBiasScales(scale_in, scale_weights);
dnnl::primitive_attr attrs;
int mask = CreateMask(0, scale_data.size() > 1);
attrs.set_output_scales(mask, scale_data);
auto user_md = dnnl::memory::desc({bias->dims()[0]},
MKLDNNGetDataType<float>(),
dnnl::memory::format_tag::a);
memory_p = this->AcquireMemoryWithReorderAndAttrs(
user_md,
this->fwd_pd_->bias_desc(),
to_void_cast<float>(bias_data),
attrs);
}
return memory_p;
}
}
std::shared_ptr<dnnl::memory> AcquireWeightsMemoryWithReorder(
const Tensor* weights, const std::vector<float>& scale_data) {
const std::string weights_key = this->memory_key_ + "@weights";
auto memory_p = std::static_pointer_cast<dnnl::memory>(
this->dev_ctx_.GetBlob(weights_key));
return inner_product_forward::primitive_desc(fc_desc, attrs, engine_); if (!memory_p) {
const float* weights_data = weights->data<float>();
auto weights_dims = this->fwd_pd_->weights_desc().dims();
auto user_md = dnnl::memory::desc(weights_dims,
MKLDNNGetDataType<float>(),
dnnl::memory::format_tag::io);
if (IsInt8<T_w>()) {
dnnl::primitive_attr attrs;
int mask = CreateMask(0, scale_data.size() > 1);
attrs.set_output_scales(mask, scale_data);
memory_p = this->AcquireMemoryWithReorderAndAttrs(
user_md,
this->fwd_pd_->weights_desc(),
to_void_cast<float>(weights_data),
attrs);
} else {
memory_p =
this->AcquireMemoryWithReorder(user_md,
this->fwd_pd_->weights_desc(),
to_void_cast<float>(weights_data));
}
this->dev_ctx_.SetBlob(weights_key, memory_p);
}
return memory_p;
} }
// Create output memory based on output tensor and inner_product std::shared_ptr<dnnl::memory> AcquireCustomDstMemory(
// primitive descriptor format chosen for output const ExecutionContext& ctx, Tensor* out) {
dnnl::memory CreateDstMemory(
const dnnl::inner_product_forward::primitive_desc& fc_prim_desc,
const ExecutionContext& ctx,
Tensor* output) {
if (ctx.HasAttr("fuse_residual_connection") && if (ctx.HasAttr("fuse_residual_connection") &&
ctx.Attr<bool>("fuse_residual_connection")) { ctx.Attr<bool>("fuse_residual_connection")) {
auto* residual_param = ctx.Output<Tensor>("ResidualData"); auto* residual_param = ctx.Output<Tensor>("ResidualData");
PADDLE_ENFORCE_EQ( PADDLE_ENFORCE_EQ(
output->dims(), out->dims(),
residual_param->dims(), residual_param->dims(),
platform::errors::InvalidArgument( platform::errors::InvalidArgument(
"Output and elementwise parameter need to have the " "Output and elementwise parameter need to have the "
"same dimension sizes, but got output's dimension = %d" "same dimension sizes, but got output's dimension = %d"
" and residual param's dimension =%d .", " and residual param's dimension =%d .",
output->dims().size(), out->dims().size(),
residual_param->dims().size())); residual_param->dims().size()));
output->ShareDataWith(*residual_param); out->ShareDataWith(*residual_param);
} }
return this->template AcquireDstMemory<T_out>(out);
}
};
auto dst_desc = fc_prim_desc.dst_desc(); template <typename T_in, typename T_w>
auto buffer_size = dst_desc.get_size(); class FCMKLDNNKernel : public framework::OpKernel<T_in> {
T_out* output_data = public:
output->mutable_data<T_out>(ctx.GetPlace(), buffer_size); void Compute(const framework::ExecutionContext& ctx) const override {
memory dst_mem(dst_desc, engine_, to_void_cast<T_out>(output_data)); bool force_fp32_output = ctx.Attr<bool>("force_fp32_output");
SetOutputFormat(ctx.Input<LoDTensor>("Input")->format(), output); bool fuse_relu = ctx.Attr<std::string>("activation_type") == "relu";
return dst_mem; if (force_fp32_output) {
this->RunKernel<float>(ctx);
} else if (IsInt8<T_in>()) {
if (fuse_relu) {
this->RunKernel<uint8_t>(ctx);
} else {
this->RunKernel<int8_t>(ctx);
}
} else {
this->RunKernel<T_in>(ctx);
}
}
template <typename T_out = T_w>
void RunKernel(const framework::ExecutionContext& ctx) const {
const auto& dev_ctx =
ctx.template device_context<platform::MKLDNNDeviceContext>();
const auto& mkldnn_engine = dev_ctx.GetEngine();
const auto* x = ctx.Input<LoDTensor>("Input");
const auto* weights = ctx.Input<Tensor>("W");
const auto* bias = ctx.Input<Tensor>("Bias");
auto out = ctx.Output<LoDTensor>("Out");
auto in_col_dims = ctx.Attr<int>("in_num_col_dims");
const float scale_in = ctx.Attr<float>("Scale_in");
const auto& scale_weights = ctx.Attr<std::vector<float>>("Scale_weights");
RecomputeOutputDims(ctx, x, weights, out);
FCMKLDNNHandler<T_in, T_w, T_out> handler(ctx,
dev_ctx,
x,
weights,
bias,
out,
in_col_dims,
mkldnn_engine,
ctx.GetPlace());
auto src_memory_p = handler.AcquireSrcMemoryWithReorder(x);
auto weights_memory_p =
handler.AcquireWeightsMemoryWithReorder(weights, scale_weights);
auto dst_memory_p = handler.AcquireCustomDstMemory(ctx, out);
auto fc_p = handler.AcquireForwardPrimitive();
auto& astream = paddle::platform::MKLDNNDeviceContext::tls().get_stream();
std::unordered_map<int, dnnl::memory> fc_args = {
{DNNL_ARG_SRC, *src_memory_p},
{DNNL_ARG_WEIGHTS, *weights_memory_p},
{DNNL_ARG_DST, *dst_memory_p}};
if (bias) {
auto bias_memory_p =
handler.AcquireBiasMemoryWithReorder(bias, scale_in, scale_weights);
fc_args.insert({DNNL_ARG_BIAS, *bias_memory_p});
}
fc_p->execute(astream, fc_args);
astream.wait();
out->set_mem_desc(
dst_memory_p->get_desc().reshape(phi::vectorize(out->dims())));
} }
void RecomputeOutputDims(const ExecutionContext& ctx, void RecomputeOutputDims(const ExecutionContext& ctx,
const LoDTensor* input, const LoDTensor* x,
const Tensor* w, const Tensor* weights,
LoDTensor* output) { LoDTensor* out) const {
int in_num_col_dims = ctx.Attr<int>("in_num_col_dims"); int in_num_col_dims = ctx.Attr<int>("in_num_col_dims");
bool padding_weights = ctx.Attr<bool>("padding_weights"); bool padding_weights = ctx.Attr<bool>("padding_weights");
PADDLE_ENFORCE_EQ(padding_weights, PADDLE_ENFORCE_EQ(padding_weights,
...@@ -596,102 +418,16 @@ class FCPrimitiveFactory { ...@@ -596,102 +418,16 @@ class FCPrimitiveFactory {
platform::errors::PermissionDenied( platform::errors::PermissionDenied(
"Weight padding in fc can not be used in MKLDNN.")); "Weight padding in fc can not be used in MKLDNN."));
std::vector<int64_t> output_dims; std::vector<int64_t> output_dims;
FCOutputSize(input->dims(), FCOutputSize(x->dims(),
w->dims(), weights->dims(),
output_dims, output_dims,
in_num_col_dims, in_num_col_dims,
padding_weights); padding_weights);
output->Resize(phi::make_ddim(output_dims)); out->Resize(phi::make_ddim(output_dims));
output->set_lod(input->lod()); out->set_lod(x->lod());
} }
private:
const dnnl::engine& engine_;
paddle::optional<memory> input_;
paddle::optional<memory> output_;
std::shared_ptr<memory> bias_;
std::shared_ptr<memory> weights_;
paddle::optional<inner_product_forward> fc_;
}; };
// Attempt to fetch cached primitive factory based on provided parameters
// of input format, weight dimensions and output name.
// If not cached, create a new one.
template <typename T_in, typename T_w, typename T_out>
static std::shared_ptr<FCPrimitiveFactory<T_in, T_w, T_out>>
GetPrimitiveFactory(const MKLDNNDeviceContext& dev_ctx,
const std::string& key) {
auto prim_creator =
std::static_pointer_cast<FCPrimitiveFactory<T_in, T_w, T_out>>(
dev_ctx.GetBlob(key));
if (prim_creator == nullptr) {
prim_creator = std::make_shared<FCPrimitiveFactory<T_in, T_w, T_out>>(
dev_ctx.GetEngine());
dev_ctx.SetBlob(key, prim_creator);
}
return prim_creator;
}
// Choose appropriate primitive factory implementation based on inferred
// output type (uint8, int8 or float).
template <typename T_in, typename T_w>
static void ExecuteFc(const ExecutionContext& ctx,
const LoDTensor* input,
const Tensor* w,
const Tensor* bias,
LoDTensor* output,
bool fuse_relu,
bool force_fp32_output) {
auto& dev_ctx = ctx.template device_context<MKLDNNDeviceContext>();
std::string prim_key = platform::CreateKey(dev_ctx,
input->format(),
input->dims()[0],
phi::vectorize<int>(w->dims()),
ctx.OutputName("Out"));
prim_key = platform::ExtendKeyWithThreadInfoIfNeeded(dev_ctx, prim_key);
constexpr bool is_int8 =
std::is_same<T_in, int8_t>::value || std::is_same<T_in, uint8_t>::value;
bool is_bfloat16 = std::is_same<T_in, paddle::platform::bfloat16>::value;
if ((!is_int8 && !is_bfloat16) || force_fp32_output) {
GetPrimitiveFactory<T_in, T_w, float>(dev_ctx, prim_key)
->ExecuteFcPrimitive(input, w, bias, output, dev_ctx, ctx);
} else if (is_bfloat16) {
GetPrimitiveFactory<T_in, T_w, platform::bfloat16>(dev_ctx, prim_key)
->ExecuteFcPrimitive(input, w, bias, output, dev_ctx, ctx);
} else if (fuse_relu) {
GetPrimitiveFactory<T_in, T_w, uint8_t>(dev_ctx, prim_key)
->ExecuteFcPrimitive(input, w, bias, output, dev_ctx, ctx);
} else {
GetPrimitiveFactory<T_in, T_w, int8_t>(dev_ctx, prim_key)
->ExecuteFcPrimitive(input, w, bias, output, dev_ctx, ctx);
}
}
template <typename T_in, typename T_w>
class FCMKLDNNOpKernel : public framework::OpKernel<T_in> {
public:
void Compute(const paddle::framework::ExecutionContext& ctx) const override {
PADDLE_ENFORCE_EQ(
platform::is_cpu_place(ctx.GetPlace()),
true,
platform::errors::PreconditionNotMet("FC MKL-DNN must use CPUPlace."));
platform::MKLDNNDeviceContext::tls().log_lib_version();
auto input = ctx.Input<LoDTensor>("Input");
auto w = ctx.Input<Tensor>("W");
auto bias = ctx.Input<Tensor>("Bias");
auto output = ctx.Output<LoDTensor>("Out");
bool fuse_relu = ctx.Attr<std::string>("activation_type") == "relu";
bool force_fp32_output = ctx.Attr<bool>("force_fp32_output");
ExecuteFc<T_in, T_w>(
ctx, input, w, bias, output, fuse_relu, force_fp32_output);
output->set_layout(DataLayout::kMKLDNN);
}
};
} // namespace operators } // namespace operators
} // namespace paddle } // namespace paddle
...@@ -704,7 +440,7 @@ REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(fc, ...@@ -704,7 +440,7 @@ REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(fc,
::paddle::platform::CPUPlace, ::paddle::platform::CPUPlace,
FP32, FP32,
ops::kFCMKLDNNFP32, ops::kFCMKLDNNFP32,
ops::FCMKLDNNOpKernel<float, float>); ops::FCMKLDNNKernel<float, float>);
REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE( REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(
fc, fc,
...@@ -712,19 +448,19 @@ REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE( ...@@ -712,19 +448,19 @@ REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(
::paddle::platform::CPUPlace, ::paddle::platform::CPUPlace,
BF16, BF16,
ops::kFCMKLDNNFP32, ops::kFCMKLDNNFP32,
ops::FCMKLDNNOpKernel<paddle::platform::bfloat16, ops::FCMKLDNNKernel<paddle::platform::bfloat16,
paddle::platform::bfloat16>); paddle::platform::bfloat16>);
REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(fc, REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(fc,
MKLDNN, MKLDNN,
::paddle::platform::CPUPlace, ::paddle::platform::CPUPlace,
U8, U8,
ops::kFCMKLDNNINT8, ops::kFCMKLDNNINT8,
ops::FCMKLDNNOpKernel<uint8_t, int8_t>); ops::FCMKLDNNKernel<uint8_t, int8_t>);
REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(fc, REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE(fc,
MKLDNN, MKLDNN,
::paddle::platform::CPUPlace, ::paddle::platform::CPUPlace,
S8, S8,
ops::kFCMKLDNNINT8, ops::kFCMKLDNNINT8,
ops::FCMKLDNNOpKernel<int8_t, int8_t>); ops::FCMKLDNNKernel<int8_t, int8_t>);
python/paddle/fluid/tests/unittests/mkldnn/test_fc_int8_mkldnn_op.py 0 → 100644
浏览文件 @ 844d9855
# Copyright (c) 2022 PaddlePaddle Authors. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from __future__ import print_function
import unittest
import numpy as np
from paddle.fluid.tests.unittests.op_test import OpTest, OpTestTool
@OpTestTool.skip_if_not_cpu()
class TestFCINT8OneDNNOp(OpTest):
def setUp(self):
self.op_type = "fc"
self._cpu_only = True
self.configure()
self.generate_data()
self.set_inputs()
self.attrs = {
'use_mkldnn': True,
'Scale_in': self.x_scale,
'Scale_weights': [self.y_scale],
'Scale_out': self.out_scale,
'force_fp32_output': self.force_fp32_output
}
if self.force_fp32_output:
out = self.out_float
else:
out = self.out
self.outputs = {'Out': out}
def configure(self):
self.use_bias = True
self.force_fp32_output = False
def set_inputs(self):
self.inputs = {'Input': self.x, 'W': self.y_float, 'Bias': self.bias}
def quantize(self, tensor):
scale = 63. / np.abs(np.amax(tensor))
quantized = np.round(scale * tensor).astype("int8")
return scale, quantized
def generate_data(self):
self.x_float = np.random.random((10, 5)).astype("float32") * 10
self.x_scale, self.x = self.quantize(self.x_float)
self.y_float = np.random.random((5, 10)).astype("float32") * 10
self.y_scale, self.y = self.quantize(self.y_float)
self.out_float = np.dot(self.x_float, self.y_float)
if self.use_bias:
self.bias = np.random.random((10)).astype("float32") * 10
self.out_float += self.bias
self.out_scale, self.out = self.quantize(self.out_float)
def test_check_output(self):
int_atol = 2
self.check_output(int_atol)
class TestFCINT8NoBiasOneDNNOp(TestFCINT8OneDNNOp):
def configure(self):
self.use_bias = False
self.force_fp32_output = False
def set_inputs(self):
self.inputs = {
'Input': self.x,
'W': self.y_float,
}
class TestFCINT8ForceFP32OutputOneDNNOp(TestFCINT8NoBiasOneDNNOp):
def configure(self):
self.use_bias = False
self.force_fp32_output = True
if __name__ == "__main__":
import paddle
paddle.enable_static()
unittest.main()
python/paddle/fluid/tests/unittests/mkldnn/test_fc_mkldnn_op.py
浏览文件 @ 844d9855
...@@ -73,4 +73,6 @@ class TestFCMKLDNNOp1(TestFCMKLDNNOp): ...@@ -73,4 +73,6 @@ class TestFCMKLDNNOp1(TestFCMKLDNNOp):
if __name__ == "__main__": if __name__ == "__main__":
import paddle
paddle.enable_static()
unittest.main() unittest.main()
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或