diff --git a/server/Makefile-vllm b/server/Makefile-vllm
index 5e60b9b0..62fa413f 100644
--- a/server/Makefile-vllm
+++ b/server/Makefile-vllm
@@ -18,7 +18,6 @@ vllm-rocm:
 
 build-vllm-rocm: vllm-rocm
 	cd vllm && git fetch && git checkout ca6913b3c2ffacdcb7d15e914dc34adbc6c89479
-	cd vllm && patch /opt/rocm/include/hip/amd_detail/amd_hip_bf16.h ./rocm_patch/rocm_bf16.patch
 	cd vllm && PYTORCH_ROCM_ARCH="gfx90a;gfx942" python setup.py install
 
 install-vllm-rocm: build-vllm-rocm
diff --git a/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py b/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
index 0a9d57e6..9a21d043 100644
--- a/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
+++ b/server/text_generation_server/models/custom_modeling/flash_llama_modeling.py
@@ -244,7 +244,7 @@ class LlamaMLP(nn.Module):
         )
 
     def forward(self, hidden_states):
-        if False and IS_ROCM_SYSTEM and self.hidden_act == "silu" and hidden_states.shape[0] == 1:
+        if IS_ROCM_SYSTEM and self.hidden_act == "silu" and hidden_states.shape[0] == 1:
             out = torch.empty(
                 hidden_states.shape[0],
                 self.intermediate_size,
diff --git a/server/text_generation_server/utils/layers.py b/server/text_generation_server/utils/layers.py
index 330b7408..6635be56 100644
--- a/server/text_generation_server/utils/layers.py
+++ b/server/text_generation_server/utils/layers.py
@@ -366,7 +366,7 @@ class FastLinearROCm(nn.Module):
         weight = self.weight
         bias = self.bias
 
-        if False and IS_ROCM_SYSTEM and inp.numel() // inp.size(-1) == 1:
+        if IS_ROCM_SYSTEM and inp.numel() // inp.size(-1) == 1:
             batched = False
 
             if inp.dim() == 3: