experiments/STT-tensorflow
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Disable collective_nccl_test on single GPU and enable on multiple GPUs.

Browse Source 
PiperOrigin-RevId: 289142542
Change-Id: I6b9c41f74062accc32173cc7afa4228e500bf31c
This commit is contained in:
Ayush Dubey 2020-01-10 12:29:50 -08:00 committed by TensorFlower Gardener
parent f6404f4f24
commit bdb99e06c5
2 changed files with 21 additions and 14 deletions
tensorflow/core/kernels
BUILDcollective_nccl_test.cc

8
tensorflow/core/kernels/BUILD
View File

@ -239,7 +239,13 @@ tf_cuda_cc_test(
name = "collective_nccl_test",
size = "small",
srcs = ["collective_nccl_test.cc"],
tags = tf_cuda_tests_tags() + ["no_cuda_on_cpu_tap"],
tags = tf_cuda_tests_tags() + [
"guitar",
"manual",
"multi_gpu",
"no_oss",
"notap",
],
deps = [
"//tensorflow/core:all_kernels",
"//tensorflow/core:core_cpu",

27
tensorflow/core/kernels/collective_nccl_test.cc
View File

@ -81,20 +81,18 @@ class NcclTestBase : public ::testing::Test {
class DeviceInstance;
NcclTestBase(CollectiveType collective_type, const string& collective_name)
: collective_type_(collective_type), collective_name_(collective_name) {}
: collective_type_(collective_type),
collective_name_(collective_name),
col_exec_(nullptr) {}
~NcclTestBase() override {
if (col_exec_) col_exec_->Unref();
}
void InitGPUDevices() {
void SetUp() {
std::vector<std::unique_ptr<Device>> all_devices;
SessionOptions session_options;
session_options.config.mutable_gpu_options()
->set_per_process_gpu_memory_fraction(0.1);
session_options.env = Env::Default();
Status s = DeviceFactory::GetFactory(DEVICE_GPU)
->AddDevices(session_options, "", &all_devices);
TF_CHECK_OK(s);
TF_CHECK_OK(DeviceFactory::GetFactory(DEVICE_GPU)
->AddDevices(SessionOptions(), "", &all_devices));
for (std::unique_ptr<Device>& d : all_devices) {
if (d->device_type() == "GPU") {
gpus_.emplace_back(std::move(d));
@ -105,13 +103,11 @@ class NcclTestBase : public ::testing::Test {
void Init(const int num_ranks, const int instance_key) {
setenv("NCCL_DEBUG", "INFO", 1 /* replace */);
setenv("NCCL_LAUNCH_MODE", "PARALLEL", 1 /* replace */);
InitGPUDevices();
std::vector<std::unique_ptr<Device>> local_devices;
std::vector<string> device_names;
CHECK_LE(num_ranks, gpus_.size());
for (int rank = 0; rank < num_ranks; ++rank) {
if (rank < gpus_.size()) {
local_devices.emplace_back(std::move(gpus_[rank]));
}
local_devices.emplace_back(std::move(gpus_[rank]));
}
int num_gpus = local_devices.size();
for (const auto& device : local_devices) {
@ -180,6 +176,11 @@ class NcclTestBase : public ::testing::Test {
}
void RunTest(int num_ranks, int input_length, int instance_key) {
if (num_ranks > gpus_.size()) {
LOG(WARNING) << "Skipping test because required " << num_ranks
<< " GPUs but found " << gpus_.size();
return;
}
Init(num_ranks, instance_key);
std::vector<float> expected;
InitExpected(&expected, input_length, num_ranks);