[Feature] : Support disaggregated inference pipeline for Qwen3_TTS (#1161)

Signed-off-by: Sy03 <1370724210@qq.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
Sy03andHongsheng Liu authored and GitHub committed 2026-02-21 07:36:44 +08:00
1 parent 5295f0f81b
commit efbe4116ff
23 files changed
+2970 -2592

No files matched your search

@@ -380,9 +380,12 @@ class GPUGenerationModelRunner(OmniGPUModelRunner):
pooler_output.append(mm_payload)
else:
raise RuntimeError("Unsupported diffusion output type")
# [Omni] Copy req_id mappings to avoid async scheduling mutation.
req_ids_output_copy = self.input_batch.req_ids.copy()
req_id_to_index_output_copy = self.input_batch.req_id_to_index.copy()
output = OmniModelRunnerOutput(
req_ids=self.input_batch.req_ids,
req_id_to_index=self.input_batch.req_id_to_index,
req_ids=req_ids_output_copy,
req_id_to_index=req_id_to_index_output_copy,
sampled_token_ids=[],
logprobs=None,
prompt_logprobs_dict={},