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.

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

[CB]: allow int8 KV cache precision for CPU #1552

Open
wants to merge 3 commits into
base: master
Choose a base branch
from
Open
Changes from all commits
Commits
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
5 changes: 0 additions & 5 deletions src/cpp/src/continuous_batching_impl.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -53,11 +53,6 @@ void apply_kv_cache_precision(const std::shared_ptr<ov::Model>& model, const std
// x86 and ARM have different default kv cache type, take this information from the plugin
m_kv_cache_type = core.get_property(device, ov::hint::kv_cache_precision);
}

// TEMP WA: currently FP16 / BF16 KV cache is faster than U8 for PagedAttention
if (m_kv_cache_type == ov::element::u8) {
m_kv_cache_type = inference_precision == ov::element::bf16 ? ov::element::bf16 : ov::element::f16;
}
} else if (device.find("GPU") != std::string::npos) {
if (accuracy_mode) {
inference_precision = ov::element::f32;
Expand Down
Loading