From 87c94bc87943818ad039d5c916df793fbd081e6a Mon Sep 17 00:00:00 2001
From: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Date: Fri, 1 Aug 2025 13:24:46 +0100
Subject: [PATCH] Revert "Update sampling_metadata.py (#21937)" (#22088)

Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
---
 vllm/model_executor/sampling_metadata.py | 12 ++++++------
 1 file changed, 6 insertions(+), 6 deletions(-)

diff --git a/vllm/model_executor/sampling_metadata.py b/vllm/model_executor/sampling_metadata.py
index 66bcf1c4bfe50..56f0f0984bfa0 100644
--- a/vllm/model_executor/sampling_metadata.py
+++ b/vllm/model_executor/sampling_metadata.py
@@ -539,37 +539,37 @@ class SamplingTensors:
         temperatures_t = torch.tensor(
             temperatures,
             device="cpu",
-            dtype=torch.float32,
+            dtype=dtype,
             pin_memory=pin_memory,
         )
         top_ps_t = torch.tensor(
             top_ps,
             device="cpu",
-            dtype=torch.float32,
+            dtype=dtype,
             pin_memory=pin_memory,
         )
         min_ps_t = torch.tensor(
             min_ps,
             device="cpu",
-            dtype=torch.float32,
+            dtype=dtype,
             pin_memory=pin_memory,
         )
         presence_penalties_t = torch.tensor(
             presence_penalties,
             device="cpu",
-            dtype=torch.float32,
+            dtype=dtype,
             pin_memory=pin_memory,
         )
         frequency_penalties_t = torch.tensor(
             frequency_penalties,
             device="cpu",
-            dtype=torch.float32,
+            dtype=dtype,
             pin_memory=pin_memory,
         )
         repetition_penalties_t = torch.tensor(
             repetition_penalties,
             device="cpu",
-            dtype=torch.float32,
+            dtype=dtype,
             pin_memory=pin_memory,
         )
         top_ks_t = torch.tensor(