Skip to content

[Bugfix][ROCm] Use chunked_prefill_paged_decode as fallback for V1 attention on ROCm #18093

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 2 commits into from
May 16, 2025
Merged

[Bugfix][ROCm] Use chunked_prefill_paged_decode as fallback for V1 attention on ROCm #18093

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6ee8222
fallback v1 FA to chunked_prefill_paged_decode
kliuae May 13, 2025
93952cb
fallback v1 FA to chunked_prefill_paged_decode
kliuae May 13, 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
109 changes: 77 additions & 32 deletions vllm/v1/attention/backends/triton_attn.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,9 @@
from vllm import _custom_ops as ops
from vllm.attention.backends.abstract import (AttentionBackend, AttentionImpl,
AttentionMetadata, AttentionType)
from vllm.attention.ops.chunked_prefill_paged_decode import (
chunked_prefill_paged_decode)
from vllm.attention.ops.paged_attn import PagedAttention
from vllm.attention.ops.triton_unified_attention import unified_attention
from vllm.logger import init_logger
from vllm.platforms import current_platform
Expand Down Expand Up @@ -149,19 +152,40 @@ def forward(
# Whenever making a change in this method, please benchmark the
# performance to make sure it does not introduce any overhead.

num_queries_per_kv = query.shape[1] // key.shape[1]
use_prefill_decode_attn = (num_queries_per_kv &
(num_queries_per_kv - 1)) != 0

num_actual_tokens = attn_metadata.num_actual_tokens

key_cache, value_cache = kv_cache.unbind(0)
torch.ops._C_cache_ops.reshape_and_cache_flash(
key,
value,
key_cache,
value_cache,
attn_metadata.slot_mapping,
self.kv_cache_dtype,
layer._k_scale,
layer._v_scale,
)
if use_prefill_decode_attn:
key_cache, value_cache = PagedAttention.split_kv_cache(
kv_cache, self.num_kv_heads, self.head_size)

# Reshape the input keys and values and store them in the cache.
PagedAttention.write_to_paged_cache(
key,
value,
key_cache,
value_cache,
attn_metadata.slot_mapping,
self.kv_cache_dtype,
layer._k_scale,
layer._v_scale,
)

else:
key_cache, value_cache = kv_cache.unbind(0)
torch.ops._C_cache_ops.reshape_and_cache_flash(
key,
value,
key_cache,
value_cache,
attn_metadata.slot_mapping,
self.kv_cache_dtype,
layer._k_scale,
layer._v_scale,
)

if self.kv_cache_dtype.startswith("fp8"):
key_cache = key_cache.view(self.fp8_dtype)
Expand Down Expand Up @@ -196,26 +220,47 @@ def forward(
max_seqlen_k = attn_metadata.max_seq_len
block_table = attn_metadata.block_table

descale_shape = (cu_seqlens_q.shape[0] - 1, key.shape[1])

unified_attention(
q=query[:num_actual_tokens],
k=key_cache,
v=value_cache,
out=output[:num_actual_tokens],
cu_seqlens_q=cu_seqlens_q,
max_seqlen_q=max_seqlen_q,
seqused_k=seqused_k,
max_seqlen_k=max_seqlen_k,
softmax_scale=self.scale,
causal=True,
alibi_slopes=self.alibi_slopes,
window_size=self.sliding_window,
block_table=block_table,
softcap=self.logits_soft_cap,
q_descale=None, # Not supported
k_descale=layer._k_scale.expand(descale_shape),
v_descale=layer._v_scale.expand(descale_shape),
)
if use_prefill_decode_attn:
# Compute attention and update output up to `num_actual_tokens`.
chunked_prefill_paged_decode(query=query[:num_actual_tokens],
key=key[:num_actual_tokens],
value=value[:num_actual_tokens],
output=output[:num_actual_tokens],
kv_cache_dtype=self.kv_cache_dtype,
key_cache=key_cache,
value_cache=value_cache,
block_table=block_table,
query_start_loc=cu_seqlens_q,
seq_lens=seqused_k,
max_seq_len=max_seqlen_k,
max_query_len=max_seqlen_q,
k_scale=layer._k_scale,
v_scale=layer._v_scale,
alibi_slopes=self.alibi_slopes,
sliding_window=self.sliding_window[0],
sm_scale=self.scale)

else:
descale_shape = (cu_seqlens_q.shape[0] - 1, key.shape[1])

unified_attention(
q=query[:num_actual_tokens],
k=key_cache,
v=value_cache,
out=output[:num_actual_tokens],
cu_seqlens_q=cu_seqlens_q,
max_seqlen_q=max_seqlen_q,
seqused_k=seqused_k,
max_seqlen_k=max_seqlen_k,
softmax_scale=self.scale,
causal=True,
alibi_slopes=self.alibi_slopes,
window_size=self.sliding_window,
block_table=block_table,
softcap=self.logits_soft_cap,
q_descale=None, # Not supported
k_descale=layer._k_scale.expand(descale_shape),
v_descale=layer._v_scale.expand(descale_shape),
)

return output