[Frontend] [Parser] Migrate Nemotron V3 to streaming parser engine (#45755)

Signed-off-by: Ben Browning <bbrownin@redhat.com>
This commit is contained in:
Ben Browning
2026-06-16 05:31:21 +00:00
committed by GitHub
parent 259ff891be
commit f3858d5422
12 changed files with 771 additions and 172 deletions
+7
View File
@@ -110,6 +110,13 @@ class ParserEngineReasoningAdapter(ReasoningParser):
def finish_streaming(self) -> DeltaMessage | None:
return self._parser_engine.finish_streaming()
def get_streaming_fallback_content(
self,
text: str,
request: ChatCompletionRequest | ResponsesRequest,
) -> str | None:
return self._parser_engine.get_streaming_fallback_content(text, request)
def count_reasoning_tokens(self, token_ids: Sequence[int]) -> int:
return self._parser_engine.count_reasoning_tokens(token_ids)
+7
View File
@@ -519,6 +519,13 @@ class ParserEngine(Parser):
return input_ids[i + 1 :]
return input_ids
def get_streaming_fallback_content(
self,
text: str,
request: ChatCompletionRequest | ResponsesRequest,
) -> str | None:
return None
def count_reasoning_tokens(self, token_ids: Sequence[int]) -> int:
start_id = self._reasoning_start_token_id
end_id = self._reasoning_end_token_id
@@ -9,6 +9,7 @@ names so that :class:`ReasoningParserManager` and
from vllm.parser.engine.adapters import make_adapters
from vllm.parser.gemma4 import Gemma4Parser
from vllm.parser.nemotron_v3 import NemotronV3Parser
from vllm.parser.qwen3 import Qwen3Parser
(
@@ -16,6 +17,11 @@ from vllm.parser.qwen3 import Qwen3Parser
Gemma4ParserToolAdapter,
) = make_adapters(Gemma4Parser)
(
NemotronV3ParserReasoningAdapter,
NemotronV3ParserToolAdapter,
) = make_adapters(NemotronV3Parser)
(
Qwen3ParserReasoningAdapter,
Qwen3ParserToolAdapter,