Skip to content

GitLab

Commit 7f3a960b authored by Shucai Xiao's avatar Shucai Xiao
Browse files

some code refinement related to quantization.

parent f8613dd1
Hide whitespace changes
Inline Side-by-side
Showing with 129 additions and 148 deletions
src/include/migraphx/op/convert.hpp
View file @ 7f3a960b
......@@ -42,10 +42,10 @@ struct convert : unary<convert>
float res = scale * x + shift;
if(target_type == shape::int8_type)
{
int factor = (res > 0) ? 1 : -1;
int factor = (res >= 0.0f) ? 1 : -1;
res = res + factor * 0.5f;
res = res > 127.0 ? 127.0 : res;
res = res < -128.0 ? -128.0 : res;
res = res > 127.0f ? 127.0f : res;
res = res < -128.0f ? -128.0f : res;
}
return res;
......
src/quantization.cpp
View file @ 7f3a960b
......@@ -7,6 +7,7 @@
#include <migraphx/op/mul.hpp>
#include <migraphx/op/add.hpp>
#include <migraphx/op/quant_dot.hpp>
#include <migraphx/op/capture.hpp>
#include <migraphx/op/convolution.hpp>
#include <migraphx/op/quant_convolution.hpp>
#include <migraphx/op/multibroadcast.hpp>
......@@ -124,6 +125,23 @@ void quantize(program& prog) { quantize(prog, {"all"}); }
static std::vector<std::pair<float, float>> int8_quant_params;
// function to compute the scale for each convert operator to convert to int8
void calc_quant_params(std::size_t ins_index, std::vector<migraphx::argument> args)
{
std::pair<float, float> param_pair{1.0f, 0.0f};
// scale and shift is need for only int8 type, and we do not
// consider shift, so set shift to 0
std::vector<float> vec_val;
args.front().visit([&](auto output) { vec_val.assign(output.begin(), output.end()); });
auto max_val = *std::max_element(vec_val.begin(), vec_val.end());
auto min_val = *std::min_element(vec_val.begin(), vec_val.end());
auto max_abs = std::max(std::fabs(max_val), std::fabs(min_val));
param_pair.first = 127.0f / max_abs;
int8_quant_params[ins_index] = param_pair;
};
// int8 quantization is different from fp16 since int8 can only handle value
// -128 ~ 127. To convert the float or double to int8, we need a scale and
// a shift, then the convert can be done as v_int8 = fp * scale + shift.
......@@ -309,9 +327,7 @@ void quantize_int8(program& prog,
ins,
op::quant_convolution{padding, stride, dilation, padding_mode, group},
converted_inputs);
auto fp_conv = prog.insert_instruction(
ins, op::convert{shape::float_type, adjust_factor, 0.0f}, quant_conv);
prog.replace_instruction(ins, op::convert{orig_type, 1.0f, 0.0f}, fp_conv);
prog.replace_instruction(ins, op::convert{orig_type, adjust_factor, 0.0f}, quant_conv);
}
else
{
......@@ -333,7 +349,66 @@ void quantize_int8(program& prog, const std::vector<std::string>& ins_names)
void quantize_int8(program& prog)
{
std::vector<std::string> ins_names = {"dot", "convolution"};
quantize_int8(prog, ins_names);
quantize_int8(prog, ins_names, int8_quant_params);
}
// For the input of each input argument, we need to insert a
// capture operator to compute the scale and shift
void capture_arguments(program& prog,
const std::vector<std::string>& ins_names,
std::function<void(std::size_t, std::vector<argument>)> func)
{
size_t num_quant_params = 0;
// the int8 quantization only support dot and convolution
std::vector<std::string> op_names = {"dot", "convolution", "quant_dot", "quant_convolution"};
if(!std::all_of(ins_names.begin(), ins_names.end(), [&](auto name) {
return std::find(op_names.begin(), op_names.end(), name) != op_names.end();
}))
{
MIGRAPHX_THROW("CAPTURE_ARGUMENTS: input operator is not supported");
}
std::unordered_map<instruction_ref, instruction_ref> ins_map;
for(auto ins : iterator_for(prog))
{
if(not contains(ins_names, ins->name()))
{
continue;
}
auto inputs = ins->inputs();
std::vector<instruction_ref> new_args;
for(auto input : inputs)
{
instruction_ref new_ins{};
if(ins_map.count(input) > 0)
{
new_ins = ins_map[input];
}
else
{
new_ins = prog.insert_instruction(
std::next(input), op::capture{num_quant_params++, func}, input);
ins_map[input] = new_ins;
}
new_args.push_back(new_ins);
}
instruction::replace(ins, ins->get_operator(), ins->get_shape(), new_args);
}
// set one pair of parameter for each argument
int8_quant_params.resize(num_quant_params, std::make_pair(-1.0f, -1.0f));
}
void capture_arguments(program& prog, const std::vector<std::string>& ins_names)
{
capture_arguments(prog, ins_names, calc_quant_params);
}
void capture_arguments(program& prog)
{
std::vector<std::string> ins_names = {"dot", "convolution"};
capture_arguments(prog, ins_names);
}
} // namespace MIGRAPHX_INLINE_NS
......
src/targets/gpu/device/convert.cpp
View file @ 7f3a960b
......@@ -21,7 +21,7 @@ void convert(hipStream_t stream,
{
gs_launch(stream, result.get_shape().elements())([=](auto i) {
float res = input_ptr[i] * scale + shift;
int factor = (res > 0) ? 1 : -1;
int factor = (res >= 0.0f) ? 1 : -1;
output_ptr[i] = static_cast<int8_t>(
std::min<float>(std::max<float>(-128.0f, res + factor * 0.5), 127.0f));
});
......
src/targets/gpu/quant_convolution.cpp
View file @ 7f3a960b
#include <migraphx/gpu/quant_convolution.hpp>
#include <migraphx/gpu/device/convert.hpp>
#include <migraphx/gpu/context.hpp>
#include <migraphx/generate.hpp>
#ifndef MIGRAPHX_GUARD_OPERATORS_CONVERT_HPP
#define MIGRAPHX_GUARD_OPERATORS_CONVERT_HPP
#include <array>
#include <migraphx/op/unary.hpp>
#include <migraphx/operation.hpp>
#include <migraphx/check_shapes.hpp>
#include <migraphx/stringutils.hpp>
#include <migraphx/streamutils.hpp>
#include <migraphx/literal.hpp>
#include <migraphx/shape_for_each.hpp>
#include <migraphx/config.hpp>
#include <cmath>
#include <utility>
namespace migraphx {
inline namespace MIGRAPHX_INLINE_NS {
namespace gpu {
namespace op {
shape miopen_quant_convolution::compute_shape(const std::vector<shape>& inputs) const
struct convert : unary<convert>
{
check_shapes{inputs, *this}.has(5).standard();
return op.compute_shape({inputs.at(0), inputs.at(1)});
}
argument miopen_quant_convolution::compute(context& ctx,
const shape& output_shape,
const std::vector<argument>& args) const
{
auto x_desc = make_tensor(args[0].get_shape());
auto x_desc_vec4 = make_tensor(args[0].get_shape(), true);
auto w_desc = make_tensor(args[1].get_shape());
auto w_desc_vec4 = make_tensor(args[1].get_shape(), true);
shape tmp_output_shape{shape::float_type, output_shape.lens()};
auto y_desc = make_tensor(tmp_output_shape);
float alpha = 1;
float beta = 0;
shape::type_t target_type = shape::half_type;
float scale = 1.0f;
float shift = 0.0f;
// pack input to vec4 format
auto status = miopenTransformTensor(ctx.get_stream().get_miopen(),
&alpha,
x_desc.get(),
args[0].implicit(),
&beta,
x_desc_vec4.get(),
arg_vec4_x.implicit());
if(status != miopenStatusSuccess)
template <class Self, class F>
static auto reflect(Self& self, F f)
{
MIGRAPHX_THROW("QUANT_CONVOLUTION: transform input tensor failed");
return pack(
f(self.target_type, "target_type"), f(self.scale, "scale"), f(self.shift, "shift"));
}
// pack input to vec4 format
status = miopenTransformTensor(ctx.get_stream().get_miopen(),
&alpha,
w_desc.get(),
args[1].implicit(),
&beta,
w_desc_vec4.get(),
arg_vec4_w.implicit());
if(status != miopenStatusSuccess)
shape compute_shape(std::vector<shape> inputs) const
{
MIGRAPHX_THROW("QUANT_CONVOLUTION: transform weight tensor failed");
check_shapes{inputs, *this}.has(1);
return {target_type, inputs.at(0).lens(), inputs.at(0).strides()};
}
status = miopenConvolutionForward(ctx.get_stream().get_miopen(),
&alpha,
x_desc_vec4.get(),
arg_vec4_x.implicit(),
w_desc_vec4.get(),
arg_vec4_w.implicit(),
cd.get(),
algo,
&beta,
y_desc.get(),
args[3].implicit(),
args[2].implicit(),
args[2].get_shape().bytes());
if(status != miopenStatusSuccess)
auto apply() const
{
MIGRAPHX_THROW("QUANT_CONVOLUTION: run convolution forward failed");
return [&](auto x) {
float res = scale * x + shift;
if(target_type == shape::int8_type)
{
int factor = (res >= 0.0f) ? 1 : -1;
res = res + factor * 0.5f;
res = res > 127.0f ? 127.0f : res;
res = res < -128.0f ? -128.0f : res;
}
return res;
};
}
// Add a conversion from float to int32_t
device::convert(ctx.get_stream().get(), args[4], args[3], 1.0f, 0.0f, shape::int32_type);
return args[4];
}
shape miopen_quant_convolution::compile(context& ctx,
const shape& output_shape,
std::vector<shape> inputs)
{
shape workspace_shape{};
auto x_desc = make_tensor(inputs[0], true);
auto w_desc = make_tensor(inputs[1], true);
shape tmp_output_shape{shape::float_type, output_shape.lens()};
auto y_desc = make_tensor(tmp_output_shape);
std::size_t workspace_size = 0;
miopenConvolutionForwardGetWorkSpaceSize(ctx.get_stream().get_miopen(),
w_desc.get(),
x_desc.get(),
cd.get(),
y_desc.get(),
&workspace_size);
workspace_shape = shape{shape::int8_type, {workspace_size}};
arg_vec4_x = to_gpu(generate_argument(pack_int8_shape(inputs[0])));
arg_vec4_w = to_gpu(generate_argument(pack_int8_shape(inputs[1])));
auto y = allocate_gpu(tmp_output_shape);
auto workspace = allocate_gpu(workspace_shape);
int algo_count = 1;
miopenConvAlgoPerf_t perf;
auto status = miopenFindConvolutionForwardAlgorithm(ctx.get_stream().get_miopen(),
x_desc.get(),
arg_vec4_x.implicit(),
w_desc.get(),
arg_vec4_w.implicit(),
cd.get(),
y_desc.get(),
y.implicit(),
1,
&algo_count,
&perf,
workspace.implicit(),
workspace_size,
false);
if(status != miopenStatusSuccess)
{
MIGRAPHX_THROW("QUANT_CONVOLUTION: find convolution failed");
}
handle = ctx.get_stream().get_miopen();
algo = perf.fwd_algo;
return shape{shape::int8_type, {perf.memory}};
}
convert(shape::type_t t) : target_type{t} {}
convert(shape::type_t t, float sle, float sft) : target_type{t}, scale{sle}, shift{sft} {}
convert() {}
};
void miopen_quant_convolution::finalize(context& ctx,
const shape& output_shape,
std::vector<shape> inputs)
{
if(handle == ctx.get_stream().get_miopen())
return;
// Check that workspace hasn't changed
auto size = inputs.at(2).bytes();
auto ws = compile(ctx, output_shape, std::move(inputs));
if(ws.bytes() > size)
MIGRAPHX_THROW("Workspace has changed during finalization.");
}
shape miopen_quant_convolution::pack_int8_shape(shape& s)
{
if(s.type() != shape::int8_type)
{
MIGRAPHX_THROW("PACK_INT8_SHAPE: only process int8_type");
}
auto lens = s.lens();
auto strides = s.strides();
lens[1] = (lens[1] + 3) / 4 * 4;
strides[0] = strides[1] * lens[1];
return {s.type(), lens, strides};
}
} // namespace gpu
} // namespace op
} // namespace MIGRAPHX_INLINE_NS
} // namespace migraphx
#endif
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment