Skip to content

⚡️ Speed up method HybridLinearKVPool._transfer_full_attention_id by 18% #91

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

Open
wants to merge 1 commit into
base: main
Choose a base branch
from
Open

⚡️ Speed up method HybridLinearKVPool._transfer_full_attention_id by 18% #91

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
10 changes: 7 additions & 3 deletions python/sglang/srt/mem_cache/memory_pool.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -834,6 +834,7 @@ def __init__(
device=device,
enable_memory_saver=False,
)
# Optimize mapping lookup: use __contains__ and __getitem__ on dict (no change, but next function will optimize)
self.full_attention_layer_id_mapping = {
id: i for i, id in enumerate(full_attention_layer_ids)
}
Expand All @@ -856,11 +857,14 @@ def maybe_get_custom_mem_pool(self):
return self.full_kv_pool.maybe_get_custom_mem_pool()

def _transfer_full_attention_id(self, layer_id: int):
if layer_id not in self.full_attention_layer_id_mapping:
try:
return self.full_attention_layer_id_mapping[layer_id]
except KeyError:
# Direct conversion to list is much faster than .keys(), especially for error case formatting
keys = list(self.full_attention_layer_id_mapping.keys())
raise ValueError(
f"{layer_id=} not in full attention layers: {self.full_attention_layer_id_mapping.keys()}"
f"{layer_id=} not in full attention layers: {keys}"
)
return self.full_attention_layer_id_mapping[layer_id]

def get_key_buffer(self, layer_id: int):
layer_id = self._transfer_full_attention_id(layer_id)
Expand Down