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

[CPU] Change kvcache type if user sets INFERENCE_PRECISION_HINT #690

Merged
merged 4 commits into from
Aug 5, 2024
Merged

[CPU] Change kvcache type if user sets INFERENCE_PRECISION_HINT #690

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
ca37cec
change kv cache type if forcing precision type
dtrawins Jul 26, 2024
53ca0d1
Merge remote-tracking branch 'upstream/master' into luocheng/kvcache_…
luo-cheng2021 Aug 5, 2024
5dc319f
apply review comments
luo-cheng2021 Aug 5, 2024
a8f173c
apply review comment
luo-cheng2021 Aug 5, 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
2 changes: 1 addition & 1 deletion src/cpp/src/continuous_batching_pipeline.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -84,7 +84,7 @@ class ContinuousBatchingPipeline::Impl {
// The model can be compiled for GPU as well
std::shared_ptr<ov::Model> model = core.read_model(models_path + "/openvino_model.xml");

DeviceConfig device_config(core, scheduler_config, device);
DeviceConfig device_config(core, scheduler_config, device, plugin_config);

apply_paged_attention_transformations(model, device_config);

Expand Down
16 changes: 15 additions & 1 deletion src/cpp/src/device_config.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,7 +20,7 @@ class DeviceConfig {
std::string m_device;

public:
DeviceConfig(ov::Core& core, const SchedulerConfig& scheduling_config, const std::string& device) {
DeviceConfig(ov::Core& core, const SchedulerConfig& scheduling_config, const std::string& device, const ov::AnyMap& plugin_config = {}) {
m_device = device;

// keep information about blocsk
Expand All @@ -29,6 +29,20 @@ class DeviceConfig {
if (m_device == "CPU") {
auto inference_precision = core.get_property(device, ov::hint::inference_precision);
m_kv_cache_type = inference_precision == ov::element::bf16 ? ov::element::bf16 : ov::element::f16;
// if user sets precision hint, kv cache type should be changed
if (plugin_config.find(ov::hint::inference_precision.name()) != plugin_config.end()) {
const auto precision = plugin_config.at(ov::hint::inference_precision.name()).as<ov::element::Type>();
if (precision == ov::element::f32) {
m_kv_cache_type = ov::element::f32;
Wovchena marked this conversation as resolved.
Show resolved Hide resolved
} else if (precision == ov::element::f16) {
m_kv_cache_type = ov::element::f16;
} else if (precision == ov::element::bf16) {
m_kv_cache_type = ov::element::bf16;
} else {
// use default f32
m_kv_cache_type = ov::element::f32;
}
}
} else if (m_device == "GPU") {
OPENVINO_ASSERT("GPU is not currently supported. Please, remove this assert and fill configuration");
} else {
Expand Down
Loading