Skip to content

feat(skip-softmax): Add skip-softmax support for KV-blocked attention #1009

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

Draft
vbaddi wants to merge 1 commit into
base: main
Choose a base branch
from
+523 −8
Draft

feat(skip-softmax): Add skip-softmax support for KV-blocked attention #1009

Show file tree
Hide file tree
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
17 changes: 17 additions & 0 deletions QEfficient/blocking/attention_blocking.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -46,6 +46,11 @@ class AttentionBlockingConfig:
head_block_size: Optional[int] = None
skip_kv: Optional[bool] = True
num_batch_blocks: Optional[int] = None
skip_softmax: bool = False
skip_softmax_scale: Optional[float] = None
skip_softmax_prefill_scale: float = 1.0
skip_softmax_decode_scale: float = 1.0
skip_softmax_min_keep_blocks: int = 1


def supports_blocked_kv(past_key_value: Optional[Cache]) -> bool:
Expand Down Expand Up @@ -164,10 +169,16 @@ def generic_blocked_attention_interface(
num_kv_blocks=blocking_config.num_kv_blocks,
num_q_blocks=blocking_config.num_q_blocks,
head_block_size=blocking_config.head_block_size,
skip_kv=blocking_config.skip_kv,
num_batch_blocks=blocking_config.num_batch_blocks,
score_mod=score_mod,
position_bias=position_bias,
sinks=sinks,
skip_softmax=blocking_config.skip_softmax,
skip_softmax_scale=blocking_config.skip_softmax_scale,
skip_softmax_prefill_scale=blocking_config.skip_softmax_prefill_scale,
skip_softmax_decode_scale=blocking_config.skip_softmax_decode_scale,
skip_softmax_min_keep_blocks=blocking_config.skip_softmax_min_keep_blocks,
)

return attn_output, attn_weights
Expand Down Expand Up @@ -227,10 +238,16 @@ def generic_blocked_mla_attention_interface(
num_kv_blocks=blocking_config.num_kv_blocks,
num_q_blocks=blocking_config.num_q_blocks,
head_block_size=blocking_config.head_block_size,
skip_kv=blocking_config.skip_kv,
num_batch_blocks=blocking_config.num_batch_blocks,
score_mod=score_mod,
position_bias=position_bias,
sinks=sinks,
skip_softmax=blocking_config.skip_softmax,
skip_softmax_scale=blocking_config.skip_softmax_scale,
skip_softmax_prefill_scale=blocking_config.skip_softmax_prefill_scale,
skip_softmax_decode_scale=blocking_config.skip_softmax_decode_scale,
skip_softmax_min_keep_blocks=blocking_config.skip_softmax_min_keep_blocks,
)

return attn_output, attn_weights
Loading
Loading