Skip to content

Restore fsdpa calibration #1086

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

Merged
merged 3 commits into from
Apr 23, 2025
Merged

Restore fsdpa calibration #1086

Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
d0110e1
Restore fsdpa calibration
madamczyk-intel Apr 15, 2025
93447bf
Merge branch 'habana_main' into dev/madamczyk/fix_fsdpa_calibration
michalkuligowski Apr 16, 2025
a64531c
Merge branch 'habana_main' into dev/madamczyk/fix_fsdpa_calibration
madamczyk-intel Apr 22, 2025
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
12 changes: 6 additions & 6 deletions vllm/attention/backends/hpu_attn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,7 +11,8 @@
import vllm_hpu_extension.kernels as kernels
import vllm_hpu_extension.ops as ops
from vllm_hpu_extension.flags import enabled_flags
from vllm_hpu_extension.utils import Matmul, Softmax, VLLMKVCache
from vllm_hpu_extension.utils import (Matmul, ModuleFusedSDPA, Softmax,
VLLMKVCache)

import vllm.envs as envs
from vllm.attention.backends.abstract import (AttentionBackend, AttentionImpl,
Expand Down Expand Up @@ -133,7 +134,9 @@ def __init__(
self.block2batch_matmul = Matmul()
self.k_cache = VLLMKVCache()
self.v_cache = VLLMKVCache()
self.fused_scaled_dot_product_attention = kernels.fsdpa()
HPUFusedSDPA = kernels.fsdpa()
self.fused_scaled_dot_product_attention = None if HPUFusedSDPA is None \
else ModuleFusedSDPA(HPUFusedSDPA)

self.prefill_impl = 'naive'
if "flex_attention" in enabled_flags():
Expand Down Expand Up @@ -272,16 +275,13 @@ def common_attention_args(self,
block_list=None,
key_cache=None,
value_cache=None):
fsdpa_op = self.fused_scaled_dot_product_attention.apply \
if self.fused_scaled_dot_product_attention is not None else None

return {
'scale': self.scale,
'matmul_qk_op': self.matmul_qk,
'matmul_av_op': self.matmul_av,
'batch2block_matmul_op': self.batch2block_matmul,
'block2batch_matmul_op': self.block2batch_matmul,
'fsdpa_op': fsdpa_op,
'fsdpa_op': self.fused_scaled_dot_product_attention,
'keys_fetch_func': self.k_cache.fetch_from_cache,
'values_fetch_func': self.v_cache.fetch_from_cache,
'softmax_op': self.softmax,
Expand Down