Refactor model mode initialization and enhance error handling in AsyncOmni. Update worker type retrieval in omni_stage and add audio token handling in Qwen3OmniMoe classes for improved robustness.

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>
This commit is contained in:
tzhouam committed 2026-03-09 06:17:08 +00:00
1 parent 9cdbcec44b
commit e6d4cdd71a
4 files changed
+25 -6

No files matched your search

@@ -39,7 +39,7 @@ class OmniChunkTransferAdapter(OmniTransferAdapterBase):
self.scheduler_max_num_seqs = vllm_config.scheduler_config.max_num_seqs
self.connector = self.create_connector(model_config)
super().__init__(model_config)
self.model_mode = getattr(model_config, "worker_type", "ar")
self.model_mode = getattr(model_config, "worker_type", None) or "ar"
# State specific to Chunk management
self.custom_process_next_stage_input_func = None
custom_process_next_stage_input_func = getattr(model_config, "custom_process_next_stage_input_func", None)