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/interfaces.py

@@ -44,10 +44,6 @@ else:
     _ProcessorFactories = object
     IntermediateTensors = object
 
-if TYPE_CHECKING:
-    from vllm.config import LoRAConfig, MultiModalConfig, SchedulerConfig
-    from vllm.sequence import IntermediateTensors
-
 logger = init_logger(__name__)
 
 MultiModalEmbeddings: TypeAlias = list[Tensor] | Tensor | tuple[Tensor, ...]
@@ -607,8 +603,6 @@ class SupportsPP(Protocol):
 
     def forward(
         self,
-        input_ids: Tensor | None,
-        positions: Tensor,
         *,
         intermediate_tensors: IntermediateTensors | None,
     ) -> IntermediateTensors | None:
@@ -637,8 +631,6 @@ class _SupportsPPType(Protocol):
 
     def forward(
         self,
-        input_ids: Tensor | None,
-        positions: Tensor,
         *,
         intermediate_tensors: IntermediateTensors | None,
     ) -> Tensor | IntermediateTensors: ...
@@ -1339,4 +1331,4 @@ def supports_xdrope(model: object) -> TypeIs[SupportsXDRoPE]: ...
 def supports_xdrope(
     model: type[object] | object,
 ) -> TypeIs[type[SupportsXDRoPE]] | TypeIs[SupportsXDRoPE]:
-    return isinstance(model, SupportsXDRoPE)
+    return isinstance(model, SupportsXDRoPE)
\ No newline at end of file

vllm/model_executor/models/intern_vit.py

@@ -33,8 +33,6 @@ from vllm.model_executor.layers.linear import (
 )
 from vllm.model_executor.layers.quantization import QuantizationConfig
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
-import vllm.envs as envs
-
 
 from .vision import run_dp_sharded_vision_model
 
@@ -457,4 +455,4 @@ class InternVisionModel(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/internlm2.py

@@ -284,7 +284,7 @@ class InternLM2Model(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,
@@ -350,7 +350,7 @@ class InternLM2ForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
 
     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,
@@ -446,7 +446,7 @@ class InternLM2ForRewardModel(InternLM2ForCausalLM):
 
     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,
@@ -456,4 +456,4 @@ class InternLM2ForRewardModel(InternLM2ForCausalLM):
         )
         hidden_states = hidden_states.to(self.head_dtype)
         logits = self.v_head(hidden_states)
-        return logits
+        return logits
\ No newline at end of file

vllm/model_executor/models/internlm2_ve.py

@@ -101,7 +101,7 @@ class InternLM2VEModel(InternLM2Model):
 
     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,
@@ -136,4 +136,4 @@ class InternLM2VEForCausalLM(InternLM2ForCausalLM):
     def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
         super().__init__(
             vllm_config=vllm_config, prefix=prefix, model_type=InternLM2VEModel
-        )
+        )
\ No newline at end of file

vllm/model_executor/models/interns1.py

@@ -782,7 +782,7 @@ class InternS1ForConditionalGeneration(
 
     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,
@@ -819,4 +819,4 @@ class InternS1ForConditionalGeneration(
             language_model="language_model",
             connector="multi_modal_projector",
             tower_model="vision_tower",
-        )
+        )
\ No newline at end of file

vllm/model_executor/models/internvl.py

@@ -1371,7 +1371,7 @@ class InternVLChatModel(nn.Module, SupportsMultiModal, 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,
@@ -1442,4 +1442,4 @@ class InternVLChatModel(nn.Module, SupportsMultiModal, SupportsPP, SupportsLoRA)
             return 0
 
         num_patches = num_vision_tokens // (self.patch_tokens + 1)
-        return num_patches * self.num_image_token
+        return num_patches * self.num_image_token
\ No newline at end of file

vllm/model_executor/models/iquest_loopcoder.py

@@ -438,7 +438,7 @@ class IQuestLoopCoderModel(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,
@@ -570,7 +570,7 @@ class IQuestLoopCoderForCausalLM(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,
@@ -592,4 +592,4 @@ class IQuestLoopCoderForCausalLM(nn.Module):
             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/isaac.py

@@ -1450,7 +1450,7 @@ class IsaacForConditionalGeneration(
 
     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,
@@ -1479,4 +1479,4 @@ class IsaacForConditionalGeneration(
             language_model="language_model",
             connector="vision_embedding.linear_fc2",  # The final linear layer
             tower_model="vision_embedding",
-        )
+        )
\ No newline at end of file

vllm/model_executor/models/jais.py

@@ -280,7 +280,7 @@ class JAISModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor | None,
+        input_ids: torch.Tensor,
         position_ids: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -344,7 +344,7 @@ class JAISLMHeadModel(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,
@@ -394,4 +394,4 @@ class JAISLMHeadModel(nn.Module, SupportsPP):
             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/jais2.py

@@ -483,7 +483,7 @@ class Jais2ForCausalLM(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,
@@ -505,4 +505,4 @@ class Jais2ForCausalLM(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/jamba.py

@@ -348,7 +348,7 @@ class JambaModel(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,
@@ -516,7 +516,7 @@ class JambaForCausalLM(
 
     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,
@@ -602,4 +602,4 @@ class JambaForSequenceClassification(JambaForCausalLM):
         pooler_config = vllm_config.model_config.pooler_config
         assert pooler_config is not None
 
-        self.pooler = DispatchPooler.for_seq_cls(pooler_config, classifier=self.score)
+        self.pooler = DispatchPooler.for_seq_cls(pooler_config, classifier=self.score)
\ No newline at end of file

vllm/model_executor/models/jina_vl.py

@@ -125,7 +125,7 @@ class JinaVLForSequenceClassification(
 
     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,
@@ -142,4 +142,4 @@ class JinaVLForSequenceClassification(
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]):
         loader = AutoWeightsLoader(self)
-        return loader.load_weights(weights, mapper=self.weight_mapper)
+        return loader.load_weights(weights, mapper=self.weight_mapper)
\ No newline at end of file

vllm/model_executor/models/kanana_v.py

@@ -732,7 +732,7 @@ class KananaVForConditionalGeneration(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,
@@ -755,4 +755,4 @@ class KananaVForConditionalGeneration(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/keye.py

@@ -1438,7 +1438,7 @@ class BaseKeyeModule(nn.Module, 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,
@@ -1692,4 +1692,4 @@ class KeyeForConditionalGeneration(
         llm_positions = torch.cat(llm_pos_ids_list, dim=1).reshape(3, -1)
         mrope_position_delta = (llm_positions.max() + 1 - len(input_tokens)).item()
 
-        return llm_positions, mrope_position_delta
+        return llm_positions, mrope_position_delta
\ No newline at end of file

vllm/model_executor/models/kimi_linear.py

@@ -506,7 +506,7 @@ class KimiLinearForCausalLM(
 
     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,
@@ -666,4 +666,4 @@ def get_spec_layer_idx_from_weight_name(
         for i in range(config.num_nextn_predict_layers):
             if weight_name.startswith(f"model.layers.{layer_idx + i}."):
                 return layer_idx + i
-    return None
+    return None
\ No newline at end of file

vllm/model_executor/models/kimi_vl.py

@@ -389,7 +389,7 @@ class KimiVLForConditionalGeneration(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,
@@ -412,4 +412,4 @@ class KimiVLForConditionalGeneration(nn.Module, SupportsMultiModal, SupportsPP):
 
     def load_weights(self, weights: Iterable[tuple[str, torch.Tensor]]):
         loader = AutoWeightsLoader(self)
-        return loader.load_weights(weights)
+        return loader.load_weights(weights)
\ No newline at end of file

vllm/model_executor/models/lfm2.py

@@ -342,7 +342,7 @@ class Lfm2Model(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,
@@ -503,7 +503,7 @@ class Lfm2ForCausalLM(
 
     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,
@@ -523,4 +523,4 @@ class Lfm2ForCausalLM(
             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/lfm2_moe.py

@@ -457,7 +457,7 @@ class Lfm2MoeModel(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,
@@ -730,7 +730,7 @@ class Lfm2MoeForCausalLM(
 
     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,
@@ -753,4 +753,4 @@ class Lfm2MoeForCausalLM(
         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/lfm2_vl.py

@@ -769,7 +769,7 @@ class Lfm2VLForConditionalGeneration(
 
     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,
@@ -804,4 +804,4 @@ class Lfm2VLForConditionalGeneration(
             language_model="language_model",
             connector="multi_modal_projector",
             tower_model="vision_tower",
-        )
+        )
\ No newline at end of file

vllm/model_executor/models/llama.py

@@ -651,7 +651,7 @@ class LlamaForCausalLM(
 
     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,