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 15, 13:30 UTC
Update test_flash_attn.py (#17102)
Signed-off-by: ShuaibinLi <lishuaibin@live.cn>
Happy committed
Apr 27, 2025 at 06:17 UTC
9869453c42b8295a84a5a4513b6b3683dde110b7
1 file changed
+1
-1
tests/kernels/attention/test_flash_attn.py
+1
-1
@@ -145,7 +145,7 @@ def test_flash_attn_with_paged_kv(
145
v_descale = None
146
if q_dtype is not None:
147
# QKV are drawn from N(0, 1): no need for a fp8 scaling factor
148
- maybe_quantized_query = query.to(q_dtype)
148
+ maybe_quantized_query = q.to(q_dtype)
149
maybe_quantized_key_cache = key_cache.to(q_dtype)
150
maybe_quantized_value_cache = value_cache.to(q_dtype)
151