Merge remote-tracking branch 'origin/v0.15.1-dev' into v0.15.1-dev

# Conflicts:
#	vllm/model_executor/layers/fused_moe/config.py
#	vllm/model_executor/layers/fused_moe/layer.py
#	vllm/model_executor/layers/quantization/compressed_tensors/compressed_tensors_marlin.py

vllm/model_executor/models/ouro.py

@@ -357,7 +357,7 @@ class OuroModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -482,7 +482,7 @@ class OuroForCausalLM(nn.Module, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -504,4 +504,4 @@ class OuroForCausalLM(nn.Module, SupportsLoRA):
             self,
             skip_prefixes=(["lm_head."] if self.config.tie_word_embeddings else None),
         )
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/ovis.py

@@ -525,7 +525,7 @@ class Ovis(nn.Module, SupportsMultiModal, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -552,4 +552,4 @@ class Ovis(nn.Module, SupportsMultiModal, SupportsPP):
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
         loader = AutoWeightsLoader(self)
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/ovis2_5.py

@@ -632,7 +632,7 @@ class Ovis2_5(nn.Module, SupportsMultiModal, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -659,4 +659,4 @@ class Ovis2_5(nn.Module, SupportsMultiModal, SupportsPP):
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
         loader = AutoWeightsLoader(self)
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/paddleocr_vl.py

@@ -1159,7 +1159,7 @@ class PaddleOCRVLForConditionalGeneration(nn.Module, SupportsMultiModal, Support
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -1227,4 +1227,4 @@ class PaddleOCRVLForConditionalGeneration(nn.Module, SupportsMultiModal, Support
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
         loader = AutoWeightsLoader(self)
         autoloaded_weights = loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
-        return autoloaded_weights
+        return autoloaded_weights
\ No newline at end of file

vllm/model_executor/models/paligemma.py

@@ -389,7 +389,7 @@ class PaliGemmaForConditionalGeneration(
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -425,4 +425,4 @@ class PaliGemmaForConditionalGeneration(
         return num_image_tokens
 
     def get_num_mm_connector_tokens(self, num_vision_tokens: int) -> int:
-        return num_vision_tokens
+        return num_vision_tokens
\ No newline at end of file

vllm/model_executor/models/persimmon.py

@@ -271,7 +271,7 @@ class PersimmonModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -348,7 +348,7 @@ class PersimmonForCausalLM(nn.Module, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -370,4 +370,4 @@ class PersimmonForCausalLM(nn.Module, SupportsPP):
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
         loader = AutoWeightsLoader(self)
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/phi.py

@@ -234,7 +234,7 @@ class PhiModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -340,7 +340,7 @@ class PhiForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -360,4 +360,4 @@ class PhiForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]) -> set[str]:
         loader = AutoWeightsLoader(self)
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/phi3v.py

@@ -686,7 +686,7 @@ class Phi3VForCausalLM(nn.Module, SupportsMultiModal, SupportsPP, SupportsQuant)
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -716,4 +716,4 @@ class Phi3VForCausalLM(nn.Module, SupportsMultiModal, SupportsPP, SupportsQuant)
         if "embed_tokens.weight" not in autoloaded_weights:
             self.embed_tokens = self.language_model.model.embed_tokens
             autoloaded_weights.add("embed_tokens.weight")
-        return autoloaded_weights
+        return autoloaded_weights
\ No newline at end of file

vllm/model_executor/models/phi4mm.py

@@ -1211,7 +1211,7 @@ class Phi4MMForCausalLM(nn.Module, SupportsLoRA, SupportsMultiModal):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -1248,4 +1248,4 @@ class Phi4MMForCausalLM(nn.Module, SupportsLoRA, SupportsMultiModal):
             language_model="model.",
             connector=["audio_projection_for_vision", "audio_projection"],
             tower_model=["vision_encoder", "embed_tokens_extend"],
-        )
+        )
\ No newline at end of file

vllm/model_executor/models/phimoe.py

@@ -483,7 +483,7 @@ class PhiMoEModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -649,7 +649,7 @@ class PhiMoEForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -668,4 +668,4 @@ class PhiMoEForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
         return loader.load_weights(weights)
 
     def get_expert_mapping(self) -> list[tuple[str, str, int, str]]:
-        return self.model.get_expert_mapping()
+        return self.model.get_expert_mapping()
\ No newline at end of file

vllm/model_executor/models/pixtral.py

@@ -479,7 +479,7 @@ class PixtralForConditionalGeneration(
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -1413,4 +1413,4 @@ class PixtralHFVisionModel(nn.Module):
                 weight_loader = getattr(param, "weight_loader", default_weight_loader)
                 weight_loader(param, loaded_weight)
             loaded_params.add(name)
-        return loaded_params
+        return loaded_params
\ No newline at end of file

vllm/model_executor/models/plamo2.py

@@ -775,7 +775,7 @@ class Plamo2Model(torch.nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -852,7 +852,7 @@ class Plamo2ForCausalLM(
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -996,4 +996,4 @@ class Plamo2ForCausalLM(
 
             param = params_dict[name]
             weight_loader = getattr(param, "weight_loader", default_weight_loader)
-            weight_loader(param, loaded_weight)
+            weight_loader(param, loaded_weight)
\ No newline at end of file

vllm/model_executor/models/plamo3.py

@@ -342,7 +342,7 @@ class Plamo3Model(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -412,7 +412,7 @@ class Plamo3ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -434,4 +434,4 @@ class Plamo3ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
             self,
             skip_prefixes=(["lm_head."] if self.config.tie_word_embeddings else None),
         )
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/qwen.py

@@ -243,7 +243,7 @@ class QWenModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -425,7 +425,7 @@ class QWenLMHeadModel(QWenBaseModel, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/qwen2.py

@@ -439,7 +439,7 @@ class Qwen2Model(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -659,7 +659,7 @@ class Qwen2ForCausalLM(nn.Module, SupportsLoRA, SupportsPP, SupportsEagle3):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/qwen2_5_omni_thinker.py

@@ -1298,7 +1298,7 @@ class Qwen2_5OmniThinkerForConditionalGeneration(
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -1330,4 +1330,4 @@ class Qwen2_5OmniThinkerForConditionalGeneration(
             language_model="language_model",
             connector="merger.",
             tower_model=["visual.", "audio_tower."],
-        )
+        )
\ No newline at end of file

vllm/model_executor/models/qwen2_audio.py

@@ -451,7 +451,7 @@ class Qwen2AudioForConditionalGeneration(nn.Module, SupportsMultiModal, Supports
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/qwen2_moe.py

@@ -408,7 +408,7 @@ class Qwen2MoeModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -633,7 +633,7 @@ class Qwen2MoeForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/qwen2_rm.py

@@ -79,7 +79,7 @@ class Qwen2RewardBaseModel(nn.Module, SupportsLoRA, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/qwen2_vl.py

@@ -1432,7 +1432,7 @@ class Qwen2VLForConditionalGeneration(
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,