We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
There was an error while loading. Please reload this page.
1 parent 973f5dc commit a4e2b26Copy full SHA for a4e2b26
vllm/engine/arg_utils.py
@@ -1157,6 +1157,12 @@ def create_engine_config(self,
1157
if self.enable_chunked_prefill and self.pipeline_parallel_size > 1:
1158
raise ValueError("Multi-Step Chunked-Prefill is not supported "
1159
"for pipeline-parallel-size > 1")
1160
+ from vllm.platforms import current_platform
1161
+ if current_platform.is_cpu():
1162
+ logger.warning("Multi-Step (--num-scheduler-steps > 1) is "
1163
+ "currently not supported for CPUs and has been "
1164
+ "disabled.")
1165
+ self.num_scheduler_steps = 1
1166
1167
# make sure num_lookahead_slots is set the higher value depending on
1168
# if we are using speculative decoding or multi-step
0 commit comments