Fix flag contition for compile_one_hot

HabanaAI · Feb 4, 2025 · ef34cf8 · ef34cf8
1 parent 39985d7
commit ef34cf8
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/vllm/worker/hpu_model_runner.py b/vllm/worker/hpu_model_runner.py
@@ -306,7 +306,7 @@ def _set_block_mapping(self, metadata, batch_size, device, dtype):
         attn_bias = (torch.zeros_like(mask, dtype=dtype).masked_fill_(
             mask, -math.inf))
 
-        if (is_fake_hpu() and htorch.utils.internal.is_lazy())\
+        if (not is_fake_hpu() and htorch.utils.internal.is_lazy())\
               or "compile_one_hot" in enabled_flags():
             # Use one_hot directly on HPU on lazy or on t.compile when
             # build is >= 20.0.0.370