Revert "[Core] Replace routing replay with device cache and async D2H pipeline" (#39917) (#42434)

Signed-off-by: aoshen02 <aoshen@inferact.ai>
This commit is contained in:
aoshen02
2026-05-13 23:49:01 -07:00
committed by GitHub
parent 0d2732dd91
commit 8c79ad6580
15 changed files with 646 additions and 1409 deletions
@@ -1101,11 +1101,6 @@ class OpenAIServingChat(OpenAIServing):
token_ids=(
as_list(output.token_ids) if request.return_token_ids else None
),
routed_experts=(
output.routed_experts.tolist()
if output.routed_experts is not None
else None
),
)
choices.append(choice_data)
continue
@@ -1327,11 +1322,6 @@ class OpenAIServingChat(OpenAIServing):
token_ids=(
as_list(output.token_ids) if request.return_token_ids else None
),
routed_experts=(
output.routed_experts.tolist()
if output.routed_experts is not None
else None
),
)
choice_data = maybe_filter_parallel_tool_calls(choice_data, request)
@@ -1371,10 +1361,6 @@ class OpenAIServingChat(OpenAIServing):
request_metadata.final_usage_info = usage
prompt_routed_experts = None
if final_res.prompt_routed_experts is not None:
prompt_routed_experts = final_res.prompt_routed_experts.tolist()
# ``final_res.prompt`` is the rendered chat-templated prompt text
prompt_text = final_res.prompt if request.return_prompt_text else None
@@ -1391,7 +1377,6 @@ class OpenAIServingChat(OpenAIServing):
),
prompt_text=prompt_text,
kv_transfer_params=final_res.kv_transfer_params,
prompt_routed_experts=prompt_routed_experts,
)
# Log complete response if output logging is enabled