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

Updated single include

Browse source
This commit is contained in:
Alejandro Saucedo 2021-02-10 08:44:33 +00:00
parent c83ba44549
commit b386113849
1 changed files with 49 additions and 48 deletions
Download patch file Download diff file Expand all files Collapse all files

97
single_include/kompute/Kompute.hpp
View file

@ -1513,23 +1513,7 @@ class Manager
std::shared_ptr<Tensor> tensor(
const std::vector<float>& data,
Tensor::TensorTypes tensorType = Tensor::TensorTypes::eDevice,
bool syncDataToGPU = true)
{
SPDLOG_DEBUG("Kompute Manager tensor triggered");
SPDLOG_DEBUG("Kompute Manager creating new tensor shared ptr");
std::shared_ptr<Tensor> tensor =
std::make_shared<Tensor>(kp::Tensor(data, tensorType));
tensor->init(this->mPhysicalDevice, this->mDevice);
if (syncDataToGPU) {
this->evalOpDefault<OpTensorSyncDevice>({ tensor });
}
this->mManagedTensors.insert(tensor);
return tensor;
}
bool syncDataToGPU = true);
/**
* Function that simplifies the common workflow of tensor initialisation. It
@ -1539,22 +1523,9 @@ class Manager
*
* @param tensors Array of tensors to rebuild
* @param syncDataToGPU Whether to sync the data to GPU memory
* @returns Initialized Tensor with memory Syncd to GPU device
*/
void rebuild(std::vector<std::shared_ptr<kp::Tensor>> tensors,
bool syncDataToGPU = true)
{
SPDLOG_DEBUG("Kompute Manager rebuild triggered");
for (std::shared_ptr<Tensor> tensor : tensors) {
// False syncData to run all tensors at once instead one by one
this->rebuild(tensor, false);
}
if (syncDataToGPU) {
this->evalOpDefault<OpTensorSyncDevice>(tensors);
}
}
bool syncDataToGPU = true);
/**
* Function that simplifies the common workflow of tensor initialisation. It
@ -1564,29 +1535,59 @@ class Manager
*
* @param tensors Single tensor to rebuild
* @param syncDataToGPU Whether to sync the data to GPU memory
* @returns Initialized Tensor with memory Syncd to GPU device
*/
void rebuild(std::shared_ptr<kp::Tensor> tensor,
bool syncDataToGPU = true)
{
SPDLOG_DEBUG("Kompute Manager rebuild Tensor triggered");
bool syncDataToGPU = true);
if (tensor->isInit()) {
tensor->freeMemoryDestroyGPUResources();
}
/**
* Destroy owned Vulkan GPU resources and free GPU memory for
* single tensor.
*
* @param tensors Single tensor to rebuild
*/
void destroy(std::shared_ptr<kp::Tensor> tensor);
tensor->init(this->mPhysicalDevice, this->mDevice);
/**
* Destroy owned Vulkan GPU resources and free GPU memory for
* vector of tensors.
*
* @param tensors Single tensor to rebuild
*/
void destroy(std::vector<std::shared_ptr<kp::Tensor>> tensors);
std::set<std::shared_ptr<Tensor>>::iterator it =
this->mManagedTensors.find(tensor);
if (it == this->mManagedTensors.end()) {
this->mManagedTensors.insert(tensor);
}
/**
* Destroy owned Vulkan GPU resources and free GPU memory for
* vector of sequences. Destroying by sequence name is more efficent
* and hence recommended instead of by object.
*
* @param sequences Vector for shared ptrs with sequences to destroy
*/
void destroy(std::vector<std::shared_ptr<kp::Sequence>> sequences);
if (syncDataToGPU) {
this->evalOpDefault<OpTensorSyncDevice>({ tensor });
}
}
/**
* Destroy owned Vulkan GPU resources and free GPU memory for
* single sequence. Destroying by sequence name is more efficent
* and hence recommended instead of by object.
*
* @param sequences Single sequence to rebuild
*/
void destroy(std::shared_ptr<kp::Sequence> sequence);
/**
* Destroy owned Vulkan GPU resources and free GPU memory for
* sequence by name.
*
* @param sequenceName Single name of named sequence to destroy
*/
void destroy(const std::string& sequenceName);
/**
* Destroy owned Vulkan GPU resources and free GPU memory for
* sequences using vector of named sequence names.
*
* @param sequenceName Vector of sequence names to destroy
*/
void destroy(const std::vector<std::string>& sequenceNames);
private:
// -------------- OPTIONALLY OWNED RESOURCES