Skip to content
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

Jump to bottom

autocast support new devices #10539

Open
wants to merge 4 commits into
base: master
Choose a base branch
from
Open

autocast support new devices #10539

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ab10d2f
autocast support new devices
fpzh2011 Jun 26, 2024
62bcb09
auto format by CI
oneflow-ci-bot Jun 26, 2024
b5d2773
revert from_or_to_torch_tensor.py
fpzh2011 Jun 27, 2024
de23bcd
Merge branch 'master' into autocast
ShawnXuan Jul 11, 2024
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
31 changes: 20 additions & 11 deletions oneflow/api/python/framework/autocast.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -16,7 +16,9 @@ limitations under the License.
#include <pybind11/pybind11.h>
#include "oneflow/api/python/of_api_registry.h"

#include "oneflow/core/common/device_type.pb.h"
#include "oneflow/core/common/throw.h"
#include "oneflow/core/ep/include/device_manager_registry.h"
#include "oneflow/core/framework/autocast.h"

namespace py = pybind11;
Expand All @@ -36,7 +38,7 @@ class AutoCastMode {
public:
OF_DISALLOW_COPY_AND_MOVE(AutoCastMode);

AutoCastMode(const std::string& device_type, Symbol<DType> dtype, bool enabled,
AutoCastMode(const std::string& device_name, Symbol<DType> dtype, bool enabled,
bool cache_enabled)
: prev_enabled_(autocast::is_enabled()),
prev_cache_enabled_(autocast::is_autocast_cache_enabled()),
Expand All @@ -48,16 +50,23 @@ class AutoCastMode {
increase_nested_count();
autocast::set_enabled(enabled);
autocast::set_autocast_cache_enabled(cache_enabled);
if (device_type == "cpu") {
autocast::set_autocast_device_type(kCPU);
autocast::set_autocast_dtype(dtype);
autocast::set_autocast_cpu_dtype(dtype);
} else if (device_type == "cuda") {
autocast::set_autocast_device_type(kCUDA);
autocast::set_autocast_dtype(dtype);
autocast::set_autocast_gpu_dtype(dtype);
} else {
THROW(RuntimeError) << "User specified autocast device_type must be 'cuda' or 'cpu'";
auto device_type = ep::DeviceManagerRegistry::GetDeviceTypeByDeviceTypeName(device_name);
ShawnXuan marked this conversation as resolved.
Show resolved Hide resolved
switch (device_type) {
case kCPU:
autocast::set_autocast_device_type(device_type);
autocast::set_autocast_dtype(dtype);
autocast::set_autocast_cpu_dtype(dtype);
break;
case kCUDA:
case kMLU:
case kNPU:
autocast::set_autocast_device_type(device_type);
autocast::set_autocast_dtype(dtype);
autocast::set_autocast_gpu_dtype(dtype);
break;
default:
THROW(RuntimeError)
<< "User specified autocast device_type must be 'cuda' or 'cpu' or 'mlu' or 'npu'";
}
}

Expand Down
2 changes: 1 addition & 1 deletion oneflow/core/framework/autocast.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -181,7 +181,7 @@ std::shared_ptr<AutoCastMeta> MakeAutoCastMeta(
// autocast only supports the following device type(s) and low precision type(s):
// - device type: CUDA
// - low precision type: half, bfloat16
static std::vector<DeviceType> autocast_device_types{kCUDA};
static std::vector<DeviceType> autocast_device_types{kCUDA, kMLU, kNPU};
static std::vector<Symbol<DType>> autocast_dtypes{DType::Float16(), DType::BFloat16()};

if (autocast_meta->autocast_color() != kBlack) {
Expand Down
5 changes: 4 additions & 1 deletion oneflow/core/job_rewriter/auto_mixed_precision_lists.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,9 @@ namespace oneflow {
const AMPList& AutoMixedPrecisionLists::WhiteList() {
static AMPList white_list = {"matmul",
"batch_matmul",
"conv1d",
"conv2d",
"conv3d",
"conv_data_grad",
"conv_filter_grad",
"conv_bias_grad",
Expand Down Expand Up @@ -137,7 +139,8 @@ const AMPList& AutoMixedPrecisionLists::GrayList() {
"group_norm_grad",
"silu",
"silu_grad",
"fused_weighted_sum"};
"fused_weighted_sum",
"cast"};
return gray_list;
}

Expand Down
6 changes: 3 additions & 3 deletions python/oneflow/amp/autocast_mode.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -153,10 +153,10 @@ def __init__(
cache_enabled: Optional[bool] = None,
):
self.device = device_type
if self.device == "cuda":
self.fast_dtype = flow.get_autocast_gpu_dtype()
elif self.device == "cpu":
if self.device == "cpu":
self.fast_dtype = flow.get_autocast_cpu_dtype()
elif self.device in ["cuda", "mlu", "npu"]:
self.fast_dtype = flow.get_autocast_gpu_dtype()
else:
raise RuntimeError(
"User specified autocast device_type must be 'cuda' or 'cpu'"
Expand Down
Loading