A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 10, 12:45 UTC
[ROCm][CI] Fix attention backend test flakiness from uninitialized KV cache memory (#31928)
Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Andreas Karatzas committed
Jan 7, 2026 at 22:35 UTC
087a138963192749043e306a0b2552c6ba465d8e
1 file changed
+1
-1
tests/v1/attention/test_attention_backends.py
+1
-1
@@ -136,7 +136,7 @@ def create_and_prepopulate_kv_cache(
136
slot_mapping = common_attn_metadata.slot_mapping
137
138
# Create KV cache
139
- kv_cache = torch.empty(
139
+ kv_cache = torch.zeros(
140
2, num_blocks, block_size, num_kv_heads, head_size, dtype=dtype, device=device
141
)
142
kv_cache_flat = kv_cache.view(2, -1, num_kv_heads, head_size)