[Bugfix] Ming-omni-tts: generalize dict checks to Mapping (#4397)

Signed-off-by: akshatvishu <akshatnayak197@gmail.com>

tests/e2e/offline_inference/test_ming_tts.py

@@ -5,6 +5,8 @@
 
 import asyncio
 import uuid
+from collections.abc import Mapping
+from typing import Any
 
 import numpy as np
 import pytest
@@ -102,8 +104,8 @@ def _flatten_audio(audio) -> torch.Tensor:
     return torch.as_tensor(audio, dtype=torch.float32).reshape(-1).cpu()
 
 
-def _extract_audio(multimodal_output: dict) -> torch.Tensor:
-    assert isinstance(multimodal_output, dict), f"Expected dict, got {type(multimodal_output)}"
+def _extract_audio(multimodal_output: Mapping[str, Any]) -> torch.Tensor:
+    assert isinstance(multimodal_output, Mapping), f"Expected Mapping, got {type(multimodal_output)}"
     audio = multimodal_output.get("audio", multimodal_output.get("model_outputs"))
     assert audio is not None, f"No audio output found, keys={list(multimodal_output.keys())}"
     waveform = _flatten_audio(audio)
@@ -112,7 +114,7 @@ def _extract_audio(multimodal_output: dict) -> torch.Tensor:
     return waveform
 
 
-def _extract_sample_rate(multimodal_output: dict) -> int:
+def _extract_sample_rate(multimodal_output: Mapping[str, Any]) -> int:
     sample_rate = multimodal_output.get("sr")
     if sample_rate is None:
         raise RuntimeError("Expected multimodal_output['sr']")
@@ -133,29 +135,29 @@ def _extract_final_audio_outputs(outputs):
         if request_output is None:
             continue
         multimodal_output = getattr(request_output, "multimodal_output", None)
-        if isinstance(multimodal_output, dict):
+        if isinstance(multimodal_output, Mapping):
             final_outputs.append(item)
             continue
         completions = getattr(request_output, "outputs", None) or []
-        if any(isinstance(getattr(completion, "multimodal_output", None), dict) for completion in completions):
+        if any(isinstance(getattr(completion, "multimodal_output", None), Mapping) for completion in completions):
             final_outputs.append(item)
     return final_outputs
 
 
-def _extract_multimodal_output(output) -> dict:
+def _extract_multimodal_output(output) -> Mapping[str, Any]:
     multimodal_output = getattr(output, "multimodal_output", None)
-    if isinstance(multimodal_output, dict):
+    if isinstance(multimodal_output, Mapping):
         return multimodal_output
 
     request_output = getattr(output, "request_output", None)
     if request_output is not None:
         multimodal_output = getattr(request_output, "multimodal_output", None)
-        if isinstance(multimodal_output, dict):
+        if isinstance(multimodal_output, Mapping):
             return multimodal_output
         completions = getattr(request_output, "outputs", None) or []
         for completion in completions:
             multimodal_output = getattr(completion, "multimodal_output", None)
-            if isinstance(multimodal_output, dict):
+            if isinstance(multimodal_output, Mapping):
                 return multimodal_output
 
     raise AssertionError("No multimodal audio output found in Ming generate results")

vllm_omni/model_executor/stage_input_processors/ming_tts.py

@@ -2,6 +2,7 @@
 # SPDX-FileCopyrightText: Copyright contributors to the vLLM project
 from __future__ import annotations
 
+from collections.abc import Mapping
 from typing import Any
 
 import torch
@@ -33,8 +34,8 @@ MING_FINAL_DECODE_STEP_KEY = "ming_final_decode_step"
 MING_STOP_REASON_BY_CODE = {code: reason for reason, code in MING_STOP_REASON_CODES.items()}
 
 
-def _extract_last_patch(pooling_output: dict[str, Any] | None) -> torch.Tensor | None:
-    if not isinstance(pooling_output, dict):
+def _extract_last_patch(pooling_output: Mapping[str, Any] | None) -> torch.Tensor | None:
+    if not isinstance(pooling_output, Mapping):
         return None
     has_patch = pooling_output.get("ming_has_patch")
     patch = pooling_output.get("ming_latent_patch")
@@ -54,8 +55,8 @@ def _extract_last_patch(pooling_output: dict[str, Any] | None) -> torch.Tensor |
     return patch.to(torch.float32).cpu()
 
 
-def _extract_all_patches(pooling_output: dict[str, Any] | None) -> torch.Tensor | None:
-    if not isinstance(pooling_output, dict):
+def _extract_all_patches(pooling_output: Mapping[str, Any] | None) -> torch.Tensor | None:
+    if not isinstance(pooling_output, Mapping):
         return None
     has_patch = pooling_output.get("ming_has_patch")
     patch = pooling_output.get("ming_latent_patch")
@@ -78,8 +79,8 @@ def _extract_all_patches(pooling_output: dict[str, Any] | None) -> torch.Tensor
     return patch.to(torch.float32).cpu()
 
 
-def _extract_last_value(pooling_output: dict[str, Any] | None, key: str) -> Any:
-    if not isinstance(pooling_output, dict):
+def _extract_last_value(pooling_output: Mapping[str, Any] | None, key: str) -> Any:
+    if not isinstance(pooling_output, Mapping):
         return None
     value = pooling_output.get(key)
     if value is None:
@@ -118,7 +119,7 @@ def _decode_stop_reason(value: Any) -> str | None:
 def _get_async_chunk_config(transfer_manager: Any) -> tuple[int, int]:
     connector = getattr(transfer_manager, "connector", None)
     raw_cfg = getattr(connector, "config", {}) or {}
-    cfg = raw_cfg.get("extra", raw_cfg) if isinstance(raw_cfg, dict) else {}
+    cfg = raw_cfg.get("extra", raw_cfg) if isinstance(raw_cfg, Mapping) else {}
     if "latent_chunk_size" not in cfg:
         logger.warning(
             "Ming async chunk config missing latent_chunk_size, using fallback value %s",