thek0tyara/llama-cpp-turboquant
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions 4

Working implementation with tests

Browse source
This commit is contained in:
Alejandro Saucedo 2021-03-07 08:00:19 +00:00
parent cf7d46cd23
commit f02b9d6915
21 changed files with 297 additions and 216 deletions
Download patch file Download diff file Expand all files Collapse all files

12
test/TestAsyncOperations.cpp
View file

@ -73,7 +73,7 @@ TEST(TestAsyncOperations, TestManagerParallelExecution)
sq->eval<kp::OpTensorSyncLocal>(inputsSyncB);
for (uint32_t i = 0; i < numParallel; i++) {
EXPECT_EQ(inputsSyncB[i]->data(), resultSync);
EXPECT_EQ(inputsSyncB[i]->vector<float>(), resultSync);
}
kp::Manager mgrAsync(0, { 0, 2 });
@ -111,7 +111,7 @@ TEST(TestAsyncOperations, TestManagerParallelExecution)
sq->eval<kp::OpTensorSyncLocal>({ inputsAsyncB });
for (uint32_t i = 0; i < numParallel; i++) {
EXPECT_EQ(inputsAsyncB[i]->data(), resultAsync);
EXPECT_EQ((inputsAsyncB[i]->vector<float>()), resultAsync);
}
// The speedup should be at least 40%
@ -152,8 +152,8 @@ TEST(TestAsyncOperations, TestManagerAsyncExecution)
kp::Manager mgr;
std::shared_ptr<kp::Tensor> tensorA = mgr.tensor(data);
std::shared_ptr<kp::Tensor> tensorB = mgr.tensor(data);
std::shared_ptr<kp::TensorT<float>> tensorA = mgr.tensor(data);
std::shared_ptr<kp::TensorT<float>> tensorB = mgr.tensor(data);
std::shared_ptr<kp::Sequence> sq1 = mgr.sequence();
std::shared_ptr<kp::Sequence> sq2 = mgr.sequence();
@ -172,6 +172,6 @@ TEST(TestAsyncOperations, TestManagerAsyncExecution)
sq1->evalAsync<kp::OpTensorSyncLocal>({ tensorA, tensorB });
sq1->evalAwait();
EXPECT_EQ(tensorA->data(), resultAsync);
EXPECT_EQ(tensorB->data(), resultAsync);
EXPECT_EQ(tensorA->vector(), resultAsync);
EXPECT_EQ(tensorB->vector(), resultAsync);
}