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 14, 05:30 UTC
[V1] reduce block size for tree attention correctness test to fix 'ou… (#22207)
Signed-off-by: Giancarlo Delfin <gdelfin@meta.com>
Giancarlo Delfin committed
Aug 4, 2025 at 19:11 UTC
5ea71ff46fe503df12f18ad41d40f5c2b18dcfcd
1 file changed
+1
-1
tests/v1/spec_decode/test_tree_attention.py
+1
-1
@@ -155,7 +155,7 @@ def test_tree_attn_correctness() -> None:
155
156
dim_per_head = 128
157
num_kv_heads = 2
158
- block_size = 128
158
+ block_size = 32
159
max_sequence_length = 8192
160
randomize_blocks = True
161
for batch_size in [1, 16, 32]: