[ROCm][Bugfix] Fix ROCM_ATTN sliding-window boundary (#59550)

Signed-off-by: tangzzycc <3081129260@qq.com>
This commit is contained in:
00CC
2026-10-04 13:48:46 +00:00
committed by GitHub
parent d64f6cd08a
commit f98d1fc493
2 changed files with 61 additions and 1 deletions
@@ -5,6 +5,7 @@ import math
import random
import time
from collections.abc import Callable
from types import SimpleNamespace
import pytest
import torch
@@ -323,6 +324,63 @@ def test_contexted_kv_attention(
torch.testing.assert_close(output, output_ref, atol=atol, rtol=0)
@pytest.mark.skipif(not current_platform.is_rocm(), reason="ROCm backend")
@pytest.mark.parametrize("sliding_window", [None, 1, 4])
@pytest.mark.parametrize("query_len", [1, 3])
@torch.inference_mode()
def test_rocm_backend_sliding_window_includes_boundary_token(
sliding_window: int | None, query_len: int
) -> None:
"""A window of W must include the current token and W - 1 preceding tokens."""
from vllm.v1.attention.backends.rocm_attn import RocmAttentionImpl
from vllm.v1.attention.ops.paged_attn import PagedAttention
device, dtype = "cuda:0", torch.float16
num_heads, head_size, seq_len = 4, 128, 8
query = torch.zeros(query_len, num_heads, head_size, dtype=dtype, device=device)
key = torch.zeros(seq_len, 1, head_size, dtype=dtype, device=device)
values = torch.arange(1, seq_len + 1, dtype=dtype, device=device)
value = values[:, None, None].expand_as(key)
kv_cache = torch.zeros(2, 1, 16, head_size, dtype=dtype, device=device)
_, value_cache = PagedAttention.split_kv_cache(kv_cache, 1, head_size)
value_cache[..., :seq_len] = value.permute(1, 2, 0)
scale = torch.tensor(1.0, device=device)
metadata = SimpleNamespace(
use_cascade=False,
num_actual_tokens=query_len,
query_start_loc=torch.tensor([0, query_len], dtype=torch.int32, device=device),
seq_lens=torch.tensor([seq_len], dtype=torch.int32, device=device),
max_query_len=query_len,
max_seq_len=seq_len,
block_table=torch.zeros(1, 1, dtype=torch.int32, device=device),
causal=True,
)
impl = RocmAttentionImpl(
num_heads=num_heads,
head_size=head_size,
scale=head_size**-0.5,
num_kv_heads=1,
alibi_slopes=None,
sliding_window=sliding_window,
kv_cache_dtype="auto",
)
output = torch.empty_like(query)
impl.forward(
SimpleNamespace(_k_scale=scale, _v_scale=scale),
query,
key[-query_len:],
value[-query_len:],
kv_cache.transpose(0, 1),
metadata,
output,
)
mask = create_causal_attention_mask_for_sdpa(
[query_len], [seq_len], sliding_window or 0, device=device, dtype=dtype
)
expected = (mask.float().softmax(dim=-1) @ values.float())[:, None, None]
torch.testing.assert_close(output, expected.to(dtype).expand_as(output))
@pytest.mark.parametrize("num_heads", NUM_HEADS)
@pytest.mark.parametrize("num_queries_per_kv", NUM_QUERIES_PER_KV)
@pytest.mark.parametrize("head_size", HEAD_SIZES)
+3 -1
View File
@@ -481,7 +481,9 @@ class RocmAttentionImpl(AttentionImpl):
k_scale=layer._k_scale,
v_scale=layer._v_scale,
alibi_slopes=self.alibi_slopes,
sliding_window=self.sliding_window[0],
# self.sliding_window[0] is the FlashAttention-style left span (W - 1).
# chunked_prefill_paged_decode expects the full window length W.
sliding_window=1 + self.sliding_window[0],
sm_scale=self.scale,
output_scale=output_scale,
sinks=self.sinks,