sync v0.18.0

vllm/model_executor/models/mixtral.py

@@ -338,7 +338,7 @@ class MixtralModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -574,7 +574,7 @@ class MixtralForCausalLM(nn.Module, SupportsLoRA, SupportsPP, MixtureOfExperts):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/mllama4.py

@@ -875,7 +875,7 @@ class Llama4ForConditionalGeneration(
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -1157,4 +1157,4 @@ class Llama4ForConditionalGeneration(
             return 0
         num_chunks = num_vision_tokens // (raw_patches + 1)
         patches_per_chunk = Mllama4ProcessingInfo.get_patch_per_chunk(vision_config)
-        return num_chunks * patches_per_chunk
+        return num_chunks * patches_per_chunk
\ No newline at end of file

vllm/model_executor/models/modernbert.py

@@ -54,12 +54,11 @@ class ModernBertEmbeddings(nn.Module):
         input_ids: torch.Tensor,
         inputs_embeds: torch.Tensor | None = None,
     ) -> torch.Tensor:
-        if inputs_embeds is not None:
-            return self.norm(inputs_embeds)
-        else:
+        if inputs_embeds is None:
             inputs_embeds = self.tok_embeddings(input_ids)
-            embeddings = self.norm(inputs_embeds)
-            return embeddings
+
+        embeddings = self.norm(inputs_embeds)
+        return embeddings
 
 
 class ModernBertAttention(nn.Module):

vllm/model_executor/models/molmo.py

@@ -876,7 +876,7 @@ class MolmoModel(nn.Module, SupportsQuant):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/molmo2.py

@@ -1206,7 +1206,7 @@ class Molmo2TextModel(nn.Module, SupportsQuant):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/mpt.py

@@ -253,7 +253,7 @@ class MPTModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         position_ids: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -313,7 +313,7 @@ class MPTForCausalLM(nn.Module, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/nano_nemotron_vl.py

@@ -2293,7 +2293,7 @@ class NemotronH_Nano_VL_V2(
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -2359,7 +2359,7 @@ class NemotronH_Nano_VL_V2(
                 with torch.no_grad():
                     default_weight_loader(param, w)
             elif is_vision_weights(name):
-                # Convert: vision_model.radio_model.* → radio_model.*
+                # Convert: vision_model.radio_model.* 鈫?radio_model.*
                 hf_key = name[len("vision_model.") :]  # Remove "vision_model." prefix
                 vision_weights.append((hf_key, w))
             elif is_sound_weights(name):
@@ -2419,4 +2419,4 @@ class NemotronH_Nano_VL_V2(
 
     @classmethod
     def get_mamba_state_copy_func(cls):
-        return NemotronHForCausalLM.get_mamba_state_copy_func()
+        return NemotronHForCausalLM.get_mamba_state_copy_func()
\ No newline at end of file

vllm/model_executor/models/nemotron.py

@@ -477,7 +477,7 @@ class NemotronForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/nemotron_h.py

@@ -608,7 +608,7 @@ class NemotronHModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -933,7 +933,7 @@ class NemotronHForCausalLM(
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/nemotron_nas.py

@@ -453,7 +453,7 @@ class DeciLMForCausalLM(nn.Module, SupportsLoRA, SupportsPP, HasNoOps):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/nemotron_parse.py

@@ -290,7 +290,7 @@ class MBartDecoderNoPos(nn.Module):
 
     def forward(
         self,
-        decoder_input_ids: torch.Tensor,
+        decoder_input_ids: torch.Tensor | None,
         *,
         encoder_hidden_states: torch.Tensor | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -901,7 +901,7 @@ class NemotronParseForConditionalGeneration(nn.Module, SupportsMultiModal):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         encoder_outputs: list[torch.Tensor] | None = None,
         **kwargs,

vllm/model_executor/models/nemotron_vl.py

@@ -644,7 +644,7 @@ class LlamaNemotronVLChatModel(nn.Module, SupportsMultiModal, SupportsPP, Suppor
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,
@@ -941,4 +941,4 @@ class LlamaNemotronVLForSequenceClassification(
 
             loaded_weights.add(name)
 
-        return loaded_weights
+        return loaded_weights
\ No newline at end of file

vllm/model_executor/models/olmo.py

@@ -271,7 +271,7 @@ class OlmoModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -382,7 +382,7 @@ class OlmoForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/olmo2.py

@@ -309,7 +309,7 @@ class Olmo2Model(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -424,7 +424,7 @@ class Olmo2ForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/olmoe.py

@@ -300,7 +300,7 @@ class OlmoeModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -476,7 +476,7 @@ class OlmoeForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/openpangu.py

@@ -1062,7 +1062,7 @@ class OpenPanguModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -1292,7 +1292,7 @@ class OpenPanguModelBase(nn.Module, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/openpangu_mtp.py

@@ -104,7 +104,7 @@ class OpenPanguMTP(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         hidden_states: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,

vllm/model_executor/models/opt.py

@@ -267,7 +267,7 @@ class OPTDecoder(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -316,7 +316,7 @@ class OPTModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -399,7 +399,7 @@ class OPTForCausalLM(nn.Module, SupportsPP, SupportsLoRA):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/orion.py

@@ -253,7 +253,7 @@ class OrionModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None,
         inputs_embeds: torch.Tensor | None = None,
@@ -343,7 +343,7 @@ class OrionForCausalLM(nn.Module, SupportsPP):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,

vllm/model_executor/models/ouro.py

@@ -357,7 +357,7 @@ class OuroModel(nn.Module):
 
     def forward(
         self,
-        input_ids: torch.Tensor,
+        input_ids: torch.Tensor | None,
         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,
+        input_ids: torch.Tensor | None,
         positions: torch.Tensor,
         intermediate_tensors: IntermediateTensors | None = None,
         inputs_embeds: torch.Tensor | None = None,