Skip to content

[bugfix][torchair] fix kv_nz accuracy problem and remove redundant reshape_and_cache operation #3066

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
linfeng-yuan wants to merge 1 commit into
base: main
Choose a base branch
from
+3 −11
Open

[bugfix][torchair] fix kv_nz accuracy problem and remove redundant reshape_and_cache operation #3066

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
14 changes: 3 additions & 11 deletions vllm_ascend/torchair/torchair_mla.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1221,17 +1221,9 @@ def forward(
assert len(
kv_cache
) > 1, "the number of kv cache should be greater than 1, namely (nope_cache and rope_cache)"
if self.torchair_graph_enabled:
if kv_cache[0].numel() > 0 and has_prefill:
slots = attn_metadata.slot_mapping
# NOTE: Separate the kv cache in advance to avoid OOM or other issues
torch_npu._npu_reshape_and_cache(
key=kv_c_normed.view(num_tokens, self.num_kv_heads, -1),
value=prefill_k_pe,
key_cache=kv_cache[0],
value_cache=kv_cache[1],
slot_indices=slots[num_decode_tokens:])
else:
# NOTE: Since CP/SP and shared_expert_dp features temporarily depend on torchair modeling
# and attention backend, cases without torchair_graph_enabled should be considered here.
if not self.torchair_graph_enabled:
kv_c_normed = kv_c_normed.view(
[num_actual_toks, self.num_kv_heads, -1])
torch_npu._npu_reshape_and_cache(
Expand Down
Loading