[Model Refactoring] Rename deepseek_v4.py to model.py [4/N] (#43077)

Signed-off-by: Woosuk Kwon <woosuk@inferact.ai>

tests/models/test_deepseek_v4_mega_moe.py

@@ -6,7 +6,7 @@ from types import SimpleNamespace
 import pytest
 import torch
 
-from vllm.models.deepseek_v4.nvidia.deepseek_v4 import (
+from vllm.models.deepseek_v4.nvidia.model import (
     DeepseekV4MegaMoEExperts,
     _stage_deepseek_v4_mega_moe_inputs,
     make_deepseek_v4_expert_params_mapping,

vllm/models/deepseek_v4/__init__.py

@@ -17,11 +17,11 @@ from .quant_config import DeepseekV4FP8Config
 # default that mypy sees; the ROCm branch overrides it at runtime and is
 # kept type-compatible via ``# type: ignore[assignment]``.
 if TYPE_CHECKING or not current_platform.is_rocm():
-    from .nvidia.deepseek_v4 import DeepseekV4ForCausalLM
-    from .nvidia.deepseek_v4_mtp import DeepSeekV4MTP
+    from .nvidia.model import DeepseekV4ForCausalLM
+    from .nvidia.mtp import DeepSeekV4MTP
 else:
-    from .amd.deepseek_v4 import DeepseekV4ForCausalLM  # type: ignore[assignment]
-    from .amd.deepseek_v4_mtp import DeepSeekV4MTP  # type: ignore[assignment]
+    from .amd.model import DeepseekV4ForCausalLM  # type: ignore[assignment]
+    from .amd.mtp import DeepSeekV4MTP  # type: ignore[assignment]
 
 __all__ = [
     "DeepSeekV4MTP",

vllm/models/deepseek_v4/amd/deepseek_v4.py

@@ -1 +0,0 @@
-../nvidia/deepseek_v4.py

vllm/models/deepseek_v4/amd/deepseek_v4_mtp.py

@@ -1 +0,0 @@
-../nvidia/deepseek_v4_mtp.py

vllm/models/deepseek_v4/amd/model.py

@@ -0,0 +1 @@
+../nvidia/model.py

vllm/models/deepseek_v4/amd/mtp.py

@@ -0,0 +1 @@
+../nvidia/mtp.py

vllm/models/deepseek_v4/nvidia/mtp.py

@@ -40,7 +40,7 @@ from vllm.model_executor.models.utils import maybe_prefix
 from vllm.platforms import current_platform
 from vllm.sequence import IntermediateTensors
 
-from .deepseek_v4 import (
+from .model import (
     DeepseekV4DecoderLayer,
     make_deepseek_v4_expert_params_mapping,
 )