diff --git a/image_output.png b/image_output.png
deleted file mode 100644
index c216f4dff..000000000
Binary files a/image_output.png and /dev/null differ
diff --git a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773652593.yaml b/tests/e2e/stage_configs/qwen2_5_omni_ci_1773652593.yaml
deleted file mode 100644
index b68bdb71a..000000000
--- a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773652593.yaml
+++ /dev/null
@@ -1,48 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {process: true, devices: '0', max_batch_size: 1}
-  engine_args: {model_stage: thinker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.9,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0}
-  is_comprehension: true
-  final_output: true
-  final_output_type: text
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 128,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-- stage_id: 1
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: talker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.4,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: latent}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen2_5_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_p: 0.8
-    top_k: 40
-    max_tokens: 4096
-    seed: 42
-    detokenize: true
-    repetition_penalty: 1.05
-    stop_token_ids: [8294]
-- stage_id: 2
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: audio, max_num_batched_tokens: 4096, max_model_len: 4096}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 4096,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-runtime:
-  enabled: true
-  defaults: {window_size: -1, max_inflight: 1}
-  edges:
-  - {from: 0, to: 1, window_size: -1}
-  - {from: 1, to: 2, window_size: -1}
diff --git a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773653262.yaml b/tests/e2e/stage_configs/qwen2_5_omni_ci_1773653262.yaml
deleted file mode 100644
index b68bdb71a..000000000
--- a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773653262.yaml
+++ /dev/null
@@ -1,48 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {process: true, devices: '0', max_batch_size: 1}
-  engine_args: {model_stage: thinker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.9,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0}
-  is_comprehension: true
-  final_output: true
-  final_output_type: text
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 128,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-- stage_id: 1
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: talker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.4,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: latent}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen2_5_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_p: 0.8
-    top_k: 40
-    max_tokens: 4096
-    seed: 42
-    detokenize: true
-    repetition_penalty: 1.05
-    stop_token_ids: [8294]
-- stage_id: 2
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: audio, max_num_batched_tokens: 4096, max_model_len: 4096}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 4096,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-runtime:
-  enabled: true
-  defaults: {window_size: -1, max_inflight: 1}
-  edges:
-  - {from: 0, to: 1, window_size: -1}
-  - {from: 1, to: 2, window_size: -1}
diff --git a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773717276.yaml b/tests/e2e/stage_configs/qwen2_5_omni_ci_1773717276.yaml
deleted file mode 100644
index b68bdb71a..000000000
--- a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773717276.yaml
+++ /dev/null
@@ -1,48 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {process: true, devices: '0', max_batch_size: 1}
-  engine_args: {model_stage: thinker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.9,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0}
-  is_comprehension: true
-  final_output: true
-  final_output_type: text
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 128,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-- stage_id: 1
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: talker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.4,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: latent}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen2_5_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_p: 0.8
-    top_k: 40
-    max_tokens: 4096
-    seed: 42
-    detokenize: true
-    repetition_penalty: 1.05
-    stop_token_ids: [8294]
-- stage_id: 2
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: audio, max_num_batched_tokens: 4096, max_model_len: 4096}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 4096,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-runtime:
-  enabled: true
-  defaults: {window_size: -1, max_inflight: 1}
-  edges:
-  - {from: 0, to: 1, window_size: -1}
-  - {from: 1, to: 2, window_size: -1}
diff --git a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773717938.yaml b/tests/e2e/stage_configs/qwen2_5_omni_ci_1773717938.yaml
deleted file mode 100644
index b68bdb71a..000000000
--- a/tests/e2e/stage_configs/qwen2_5_omni_ci_1773717938.yaml
+++ /dev/null
@@ -1,48 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {process: true, devices: '0', max_batch_size: 1}
-  engine_args: {model_stage: thinker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.9,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0}
-  is_comprehension: true
-  final_output: true
-  final_output_type: text
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 128,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-- stage_id: 1
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: talker, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    max_model_len: 16384, max_num_batched_tokens: 16384, max_num_seqs: 1, gpu_memory_utilization: 0.4,
-    skip_mm_profiling: true, enforce_eager: 'true', trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: latent}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen2_5_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_p: 0.8
-    top_k: 40
-    max_tokens: 4096
-    seed: 42
-    detokenize: true
-    repetition_penalty: 1.05
-    stop_token_ids: [8294]
-- stage_id: 2
-  runtime: {process: true, devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen2_5OmniForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false,
-    engine_output_type: audio, max_num_batched_tokens: 4096, max_model_len: 4096}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 4096,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-runtime:
-  enabled: true
-  defaults: {window_size: -1, max_inflight: 1}
-  edges:
-  - {from: 0, to: 1, window_size: -1}
-  - {from: 1, to: 2, window_size: -1}
diff --git a/tests/e2e/stage_configs/qwen3_omni_ci_1773652593.yaml b/tests/e2e/stage_configs/qwen3_omni_ci_1773652593.yaml
deleted file mode 100644
index c763bb2c7..000000000
--- a/tests/e2e/stage_configs/qwen3_omni_ci_1773652593.yaml
+++ /dev/null
@@ -1,45 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {devices: '0', max_batch_size: 5}
-  engine_args: {model_stage: thinker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.9, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    distributed_executor_backend: mp, max_num_batched_tokens: 32768, max_model_len: 32768,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0, hf_config_name: thinker_config,
-    tensor_parallel_size: 1, load_format: dummy, custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker_async_chunk}
-  final_output: true
-  final_output_type: text
-  is_comprehension: true
-  default_sampling_params: {temperature: 0.4, top_p: 0.9, top_k: 1, max_tokens: 100,
-    seed: 42, ignore_eos: false, detokenize: true, repetition_penalty: 1.05}
-- stage_id: 1
-  runtime: {devices: '1', max_batch_size: 5}
-  engine_args: {model_stage: talker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, max_num_batched_tokens: 32768, max_model_len: 32768,
-    distributed_executor_backend: mp, hf_config_name: talker_config, load_format: dummy,
-    custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.talker2code2wav_async_chunk}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_k: 50
-    max_tokens: 1000
-    seed: 42
-    detokenize: false
-    repetition_penalty: 1.05
-    stop_token_ids: [2150]
-- stage_id: 2
-  runtime: {devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false, engine_output_type: audio,
-    gpu_memory_utilization: 0.1, distributed_executor_backend: mp, max_num_batched_tokens: 100000,
-    hf_config_name: thinker_config, async_scheduling: false, load_format: dummy}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 2000,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-async_chunk: true
diff --git a/tests/e2e/stage_configs/qwen3_omni_ci_1773653262.yaml b/tests/e2e/stage_configs/qwen3_omni_ci_1773653262.yaml
deleted file mode 100644
index c763bb2c7..000000000
--- a/tests/e2e/stage_configs/qwen3_omni_ci_1773653262.yaml
+++ /dev/null
@@ -1,45 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {devices: '0', max_batch_size: 5}
-  engine_args: {model_stage: thinker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.9, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    distributed_executor_backend: mp, max_num_batched_tokens: 32768, max_model_len: 32768,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0, hf_config_name: thinker_config,
-    tensor_parallel_size: 1, load_format: dummy, custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker_async_chunk}
-  final_output: true
-  final_output_type: text
-  is_comprehension: true
-  default_sampling_params: {temperature: 0.4, top_p: 0.9, top_k: 1, max_tokens: 100,
-    seed: 42, ignore_eos: false, detokenize: true, repetition_penalty: 1.05}
-- stage_id: 1
-  runtime: {devices: '1', max_batch_size: 5}
-  engine_args: {model_stage: talker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, max_num_batched_tokens: 32768, max_model_len: 32768,
-    distributed_executor_backend: mp, hf_config_name: talker_config, load_format: dummy,
-    custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.talker2code2wav_async_chunk}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_k: 50
-    max_tokens: 1000
-    seed: 42
-    detokenize: false
-    repetition_penalty: 1.05
-    stop_token_ids: [2150]
-- stage_id: 2
-  runtime: {devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false, engine_output_type: audio,
-    gpu_memory_utilization: 0.1, distributed_executor_backend: mp, max_num_batched_tokens: 100000,
-    hf_config_name: thinker_config, async_scheduling: false, load_format: dummy}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 2000,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-async_chunk: true
diff --git a/tests/e2e/stage_configs/qwen3_omni_ci_1773717276.yaml b/tests/e2e/stage_configs/qwen3_omni_ci_1773717276.yaml
deleted file mode 100644
index c763bb2c7..000000000
--- a/tests/e2e/stage_configs/qwen3_omni_ci_1773717276.yaml
+++ /dev/null
@@ -1,45 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {devices: '0', max_batch_size: 5}
-  engine_args: {model_stage: thinker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.9, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    distributed_executor_backend: mp, max_num_batched_tokens: 32768, max_model_len: 32768,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0, hf_config_name: thinker_config,
-    tensor_parallel_size: 1, load_format: dummy, custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker_async_chunk}
-  final_output: true
-  final_output_type: text
-  is_comprehension: true
-  default_sampling_params: {temperature: 0.4, top_p: 0.9, top_k: 1, max_tokens: 100,
-    seed: 42, ignore_eos: false, detokenize: true, repetition_penalty: 1.05}
-- stage_id: 1
-  runtime: {devices: '1', max_batch_size: 5}
-  engine_args: {model_stage: talker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, max_num_batched_tokens: 32768, max_model_len: 32768,
-    distributed_executor_backend: mp, hf_config_name: talker_config, load_format: dummy,
-    custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.talker2code2wav_async_chunk}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_k: 50
-    max_tokens: 1000
-    seed: 42
-    detokenize: false
-    repetition_penalty: 1.05
-    stop_token_ids: [2150]
-- stage_id: 2
-  runtime: {devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false, engine_output_type: audio,
-    gpu_memory_utilization: 0.1, distributed_executor_backend: mp, max_num_batched_tokens: 100000,
-    hf_config_name: thinker_config, async_scheduling: false, load_format: dummy}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 2000,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-async_chunk: true
diff --git a/tests/e2e/stage_configs/qwen3_omni_ci_1773717938.yaml b/tests/e2e/stage_configs/qwen3_omni_ci_1773717938.yaml
deleted file mode 100644
index c763bb2c7..000000000
--- a/tests/e2e/stage_configs/qwen3_omni_ci_1773717938.yaml
+++ /dev/null
@@ -1,45 +0,0 @@
-stage_args:
-- stage_id: 0
-  runtime: {devices: '0', max_batch_size: 5}
-  engine_args: {model_stage: thinker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.9, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    distributed_executor_backend: mp, max_num_batched_tokens: 32768, max_model_len: 32768,
-    enable_prefix_caching: false, mm_processor_cache_gb: 0, hf_config_name: thinker_config,
-    tensor_parallel_size: 1, load_format: dummy, custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker_async_chunk}
-  final_output: true
-  final_output_type: text
-  is_comprehension: true
-  default_sampling_params: {temperature: 0.4, top_p: 0.9, top_k: 1, max_tokens: 100,
-    seed: 42, ignore_eos: false, detokenize: true, repetition_penalty: 1.05}
-- stage_id: 1
-  runtime: {devices: '1', max_batch_size: 5}
-  engine_args: {model_stage: talker, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: ar, scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler,
-    gpu_memory_utilization: 0.5, enforce_eager: false, trust_remote_code: true, engine_output_type: latent,
-    enable_prefix_caching: false, max_num_batched_tokens: 32768, max_model_len: 32768,
-    distributed_executor_backend: mp, hf_config_name: talker_config, load_format: dummy,
-    custom_process_next_stage_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.talker2code2wav_async_chunk}
-  engine_input_source: [0]
-  custom_process_input_func: vllm_omni.model_executor.stage_input_processors.qwen3_omni.thinker2talker
-  default_sampling_params:
-    temperature: 0.9
-    top_k: 50
-    max_tokens: 1000
-    seed: 42
-    detokenize: false
-    repetition_penalty: 1.05
-    stop_token_ids: [2150]
-- stage_id: 2
-  runtime: {devices: '1', max_batch_size: 1}
-  engine_args: {model_stage: code2wav, model_arch: Qwen3OmniMoeForConditionalGeneration,
-    worker_type: generation, scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler,
-    enforce_eager: true, trust_remote_code: true, enable_prefix_caching: false, engine_output_type: audio,
-    gpu_memory_utilization: 0.1, distributed_executor_backend: mp, max_num_batched_tokens: 100000,
-    hf_config_name: thinker_config, async_scheduling: false, load_format: dummy}
-  engine_input_source: [1]
-  final_output: true
-  final_output_type: audio
-  default_sampling_params: {temperature: 0.0, top_p: 1.0, top_k: -1, max_tokens: 2000,
-    seed: 42, detokenize: true, repetition_penalty: 1.1}
-async_chunk: true