From 1e49f74ed71cce69180cd2ff577fca89a37642f9 Mon Sep 17 00:00:00 2001
From: Dan250124 <416947747@qq.com>
Date: Tue, 26 May 2026 22:33:43 +0800
Subject: [PATCH 1/2] [BugFix] Fix Qwen3-TTS Code2Wav compatibility with
 transformers >= 5.9.0

Adapt to breaking API changes in transformers masking_utils.create_causal_mask:
- input_embeds renamed to inputs_embeds
- cache_position parameter removed
Use runtime signature introspection to pass correct kwargs across versions.

Signed-off-by: Dan250124 <416947747@qq.com>
---
 .../tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py | 11 ++++++++---
 1 file changed, 8 insertions(+), 3 deletions(-)

diff --git a/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py b/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py
index 87acb600262..98ed510b6f1 100644
--- a/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py
+++ b/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py
@@ -17,7 +17,7 @@
 from collections.abc import Callable
 from dataclasses import dataclass
 from typing import Any
-
+import inspect
 import numpy as np
 import torch
 from torch import nn
@@ -565,12 +565,17 @@ def forward(
             # Prepare mask arguments
             mask_kwargs = {
                 "config": self.config,
-                "input_embeds": inputs_embeds,
                 "attention_mask": attention_mask,
-                "cache_position": cache_position,
                 "past_key_values": past_key_values,
                 "position_ids": position_ids,
             }
+            # Handle API changes across transformers versions
+            sig = inspect.signature(create_causal_mask)
+            if "input_embeds" in sig.parameters:
+                mask_kwargs["input_embeds"] = inputs_embeds
+                mask_kwargs["cache_position"] = cache_position
+            else:
+                mask_kwargs["inputs_embeds"] = inputs_embeds
             # Create the masks
             causal_mask_mapping = {
                 "full_attention": create_causal_mask(**mask_kwargs),

From 5c43fb451a19215279009cd4cdccaa0e3fd4e4f7 Mon Sep 17 00:00:00 2001
From: Dan250124 <416947747@qq.com>
Date: Tue, 26 May 2026 22:38:45 +0800
Subject: [PATCH 2/2] [Lint] Fix import ordering for ruff compliance

Move 'import inspect' to correct alphabetical position.

Signed-off-by: Dan250124 <416947747@qq.com>
---
 .../tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py          | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py b/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py
index 98ed510b6f1..99b26ea1e84 100644
--- a/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py
+++ b/vllm_omni/model_executor/models/qwen3_tts/tokenizer_12hz/modeling_qwen3_tts_tokenizer_v2.py
@@ -13,11 +13,12 @@
 # limitations under the License.
 """PyTorch Qwen3TTSTokenizerV2 model."""
 
+import inspect
 import math
 from collections.abc import Callable
 from dataclasses import dataclass
 from typing import Any
-import inspect
+
 import numpy as np
 import torch
 from torch import nn