add qwen3_moe

Use getattr with default True for MPTConfig.tie_word_embeddings,
as some MPT model configs lack this attribute.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

.gitignore

@@ -0,0 +1,240 @@
+# version file generated by setuptools-scm
+/vllm/_version.py
+
+# vllm-flash-attn built from source
+vllm/vllm_flash_attn/*
+
+# OpenAI triton kernels copied from source
+vllm/third_party/triton_kernels/*
+
+# FlashMLA interface copied from source
+vllm/third_party/flashmla/flash_mla_interface.py
+
+# triton jit
+.triton
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+cmake-build-*/
+CMakeUserPresets.json
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+/.deps/
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# generated files
+**/generated/**
+
+# uv
+uv.lock
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/#use-with-ide
+.pdm.toml
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+docs/argparse
+docs/examples/*
+!docs/examples/README.md
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+.idea/
+
+# VSCode
+.vscode/
+
+# Claude
+CLAUDE.md
+.claude/
+
+# Codex
+AGENTS.md
+.codex/
+
+# Cursor
+.cursor/
+
+# DS Store
+.DS_Store
+
+# Results
+*.csv
+
+# Python pickle files
+*.pkl
+
+# Sphinx documentation
+_build/
+
+# vim swap files
+*.swo
+*.swp
+
+# hip files generated by PyTorch
+*.hip
+*_hip*
+hip_compat.h
+
+# Benchmark dataset
+benchmarks/**/*.json
+
+# Linting
+actionlint
+shellcheck*/
+
+# Ignore moe/marlin_moe gen code
+csrc/moe/marlin_moe_wna16/kernel_*
+
+# Ignore ep_kernels_workspace folder
+ep_kernels_workspace/
+
+# Allow tracked library source folders under submodules (e.g., benchmarks/lib)
+!vllm/benchmarks/lib/
+
+# Generated gRPC protobuf files (compiled at build time from vllm_engine.proto)
+vllm/grpc/vllm_engine_pb2.py
+vllm/grpc/vllm_engine_pb2_grpc.py
+vllm/grpc/vllm_engine_pb2.pyi

README.md

@@ -3,6 +3,7 @@
 # 寒武纪 mlu370 文本生成
 该模型测试框架在寒武纪mlu370 （X8/X4）加速卡上，基于vllm 推理引擎，适配了 Qwen1.5-1.8B-Chat 模型。
 
+
 * Qwen1.5-1.8B-Chat 是通义千问系列中一款约18亿参数、轻量级的中英文对话大模型，专为高效推理和多场景聊天交互设计。
 * Llama-2-7b-chat-hf：Meta 发布的 LLaMA 2 系列中 70 亿参数的对话优化版开源大模型，适合多轮聊天与通用任务。
 * ChatGLM3-6B：智谱 AI 推出的第 3 代 ChatGLM 系列中 60 亿参数的中英双语对话大模型，支持推理、代码和多任务能力。
@@ -162,5 +163,15 @@ curl http://localhost:80/v1/chat/completions \
 
 | 模型名称    | mlu370-X8首字延迟(秒) |	mlu370-X8输入处理速度(字每秒) |	mlu370-X8输出速度(字每秒) |	mlu370-X8输出质量 | Nvidia A100字延迟(秒) |	Nvidia A100输入处理速度(字每秒) |	Nvidia A100输出速度(字每秒) |	Nvidia A100输出质量 |
 | ------------------- | ------------------- | -------------------| ------------------- | ------------------- | ------------------- | -------------------  | ------------------- | -------------------  |
-| Qwen/Qwen-1_8B   |0.203	| 13493.2	 |	119.2 |	10.0 | 0.052 | 25591.5 |	165.0 |	15.0|					
-| Qwen/Qwen1.5-0.5B   |0.132	| 12366.6	 |	106.9 |	15.0 | 0.066 | 24935.4 |	151.4 |	10.0|
+| Qwen/Qwen-1_8B   |0.203	| 13493.2	 |	119.2 |	10.0 | 0.052 | 25591.5 |	165.0 |	15.0|
+| Qwen/Qwen1.5-0.5B   |0.132	| 12366.6	 |	106.9 |	15.0 | 0.066 | 24935.4 |	151.4 |	10.0|
+
+## 版本更新记录
+
+| 版本 | 日期 | 更新内容 |
+|------|------|----------|
+| v0.0.2 | 2026-02-04 | **Qwen3 模型支持**：实现 QK Normalization 架构适配，修复 rope_scaling 和 tokenizer 兼容性问题，解决张量连续性导致的 view 操作失败 |
+| v0.0.3 | 2026-02-06 | **Transformers 通用后端**：支持通过 `auto_map` 加载任意自定义 HuggingFace 模型，新增 registry 回退逻辑、Linear 返回值处理、RMSNorm 维度恢复等 |
+| v0.0.3.1 | 2026-02-06 | **CNNL Tensor 溢出修复**：解决极小模型在大显存设备上部署时 KV cache 元素数超过 int32 限制的问题，在 mlu_worker 和 cache_engine 中添加双重防护 |
+| v0.0.4 | 2026-02-10 | **Gemma3 模型支持**：新增 Gemma3ForCausalLM 模型实现（含 QK Normalization、per-layer rope 配置、layer_types 滑动窗口），修复 `patch_rope_scaling_dict` 在 rope_scaling 缺少 `rope_type` 键时崩溃的问题，更新模型注册表及 config.py 中 interleaved attention 和 dtype 自动处理逻辑 |
+| v0.0.4.1 | 2026-02-10 | **Gemma3 rope 兼容性修复**：修复新版 transformers `Gemma3TextConfig` 缺少 `rope_theta` 属性的问题，从 `rope_parameters` 字典兼容提取 rope 配置（支持 Transformers v4/v5）；修复 `rope_scaling` 嵌套字典导致 `get_rope` 缓存 unhashable 的问题；适配 MLU `forward_mlu` 接口，将 q/k 合并为单张量调用 rotary_emb 后再拆分 |

vllm-v0.6.2/vllm/config.py

@@ -226,7 +226,7 @@ class ModelConfig:
         sliding_window = getattr(self.hf_text_config, "sliding_window", None)
         has_interleaved_attention = (sliding_window is not None) and (
             isinstance(sliding_window, list) or
-            (self.hf_text_config.model_type in ["gemma2"]))
+            (self.hf_text_config.model_type in ["gemma2", "gemma3"]))
 
         if (not self.disable_sliding_window and has_interleaved_attention):
             sliding_window_len_min = get_min_sliding_window(
@@ -274,7 +274,13 @@ class ModelConfig:
         self, limit_mm_per_prompt: Optional[Mapping[str, int]]
     ) -> Optional["MultiModalConfig"]:
         architectures = getattr(self.hf_config, "architectures", [])
-        if ModelRegistry.is_multimodal_model(architectures):
+        if ModelRegistry.is_multimodal_model(
+            architectures,
+            model_path=self.model,
+            revision=self.revision,
+            trust_remote_code=self.trust_remote_code,
+            hf_config=self.hf_config,
+        ):
             return MultiModalConfig(limit_per_prompt=limit_mm_per_prompt or {})
 
         if limit_mm_per_prompt:
@@ -308,11 +314,23 @@ class ModelConfig:
 
     def _init_attention_free(self) -> bool:
         architectures = getattr(self.hf_config, "architectures", [])
-        return ModelRegistry.is_attention_free_model(architectures)
+        return ModelRegistry.is_attention_free_model(
+            architectures,
+            model_path=self.model,
+            revision=self.revision,
+            trust_remote_code=self.trust_remote_code,
+            hf_config=self.hf_config,
+        )
 
     def _init_has_inner_state(self) -> bool:
         architectures = getattr(self.hf_config, "architectures", [])
-        return ModelRegistry.model_has_inner_state(architectures)
+        return ModelRegistry.model_has_inner_state(
+            architectures,
+            model_path=self.model,
+            revision=self.revision,
+            trust_remote_code=self.trust_remote_code,
+            hf_config=self.hf_config,
+        )
 
     def _verify_tokenizer_mode(self) -> None:
         tokenizer_mode = self.tokenizer_mode.lower()
@@ -335,8 +353,20 @@ class ModelConfig:
         task_support: Dict[_Task, bool] = {
             # NOTE: Listed from highest to lowest priority,
             # in case the model supports multiple of them
-            "generate": ModelRegistry.is_text_generation_model(architectures),
-            "embedding": ModelRegistry.is_embedding_model(architectures),
+            "generate": ModelRegistry.is_text_generation_model(
+                architectures,
+                model_path=self.model,
+                revision=self.revision,
+                trust_remote_code=self.trust_remote_code,
+                hf_config=hf_config,
+            ),
+            "embedding": ModelRegistry.is_embedding_model(
+                architectures,
+                model_path=self.model,
+                revision=self.revision,
+                trust_remote_code=self.trust_remote_code,
+                hf_config=hf_config,
+            ),
         }
         supported_tasks_lst: List[_Task] = [
             task for task, is_supported in task_support.items() if is_supported
@@ -1824,9 +1854,9 @@ def _get_and_verify_dtype(
         dtype = dtype.lower()
         if dtype == "auto":
             if config_dtype == torch.float32:
-                if config.model_type == "gemma2":
+                if config.model_type in ("gemma2", "gemma3"):
                     logger.info(
-                        "For Gemma 2, we downcast float32 to bfloat16 instead "
+                        "For Gemma 2/3, we downcast float32 to bfloat16 instead "
                         "of float16 by default. Please specify `dtype` if you "
                         "want to use float16.")
                     torch_dtype = torch.bfloat16

vllm-v0.6.2/vllm/model_executor/layers/fused_moe/layer.py

@@ -153,23 +153,25 @@ class UnquantizedFusedMoEMethod(FusedMoEMethodBase, CustomOp):
 
     def forward_mlu(
         self,
+        layer: torch.nn.Module,
         x: torch.Tensor,
-        w1: torch.Tensor,
-        w2: torch.Tensor,
         router_logits: torch.Tensor,
         top_k: int,
         renormalize: bool,
         use_grouped_topk: bool,
-        num_expert_group: Optional[int],
-        topk_group: Optional[int],
+        topk_group: Optional[int] = None,
+        num_expert_group: Optional[int] = None,
+        custom_routing_function: Optional[Callable] = None,
     ) -> torch.Tensor:
         from vllm._mlu_ops import fused_moe
 
-        assert use_grouped_topk is False and num_expert_group is None and topk_group is None, \
-            f"Following params: use_grouped_topk, num_expert_group, topk_group are not support yet."
+        assert use_grouped_topk is False and num_expert_group is None \
+            and topk_group is None, \
+            "Following params: use_grouped_topk, num_expert_group, " \
+            "topk_group are not supported yet."
         return fused_moe(x,
                         router_logits,
-                        w1, w2,
+                        layer.w13_weight, layer.w2_weight,
                         None, None, # bias1, bias2
                         None, # residual
                         None, # input_smooth

vllm-v0.6.2/vllm/model_executor/layers/linear.py

@@ -146,6 +146,7 @@ class LinearBase(torch.nn.Module):
         skip_bias_add: If true, skip adding bias but instead return it.
         params_dtype: Data type for the parameters.
         quant_config: Quantization configure.
+        return_bias: If False, return only output tensor instead of (output, bias) tuple.
     """
 
     def __init__(
@@ -156,6 +157,7 @@ class LinearBase(torch.nn.Module):
         params_dtype: Optional[torch.dtype] = None,
         quant_config: Optional[QuantizationConfig] = None,
         prefix: str = "",
+        return_bias: bool = True,
     ):
         super().__init__()
 
@@ -163,6 +165,7 @@ class LinearBase(torch.nn.Module):
         self.input_size = input_size
         self.output_size = output_size
         self.skip_bias_add = skip_bias_add
+        self.return_bias = return_bias
         if params_dtype is None:
             params_dtype = torch.get_default_dtype()
         self.params_dtype = params_dtype
@@ -198,13 +201,15 @@ class ReplicatedLinear(LinearBase):
                  skip_bias_add: bool = False,
                  params_dtype: Optional[torch.dtype] = None,
                  quant_config: Optional[QuantizationConfig] = None,
-                 prefix: str = ""):
+                 prefix: str = "",
+                 return_bias: bool = True):
         super().__init__(input_size,
                          output_size,
                          skip_bias_add,
                          params_dtype,
                          quant_config,
-                         prefix=prefix)
+                         prefix=prefix,
+                         return_bias=return_bias)
 
         # All the linear layer supports quant method.
         assert self.quant_method is not None
@@ -238,6 +243,9 @@ class ReplicatedLinear(LinearBase):
         bias = self.bias if not self.skip_bias_add else None
         assert self.quant_method is not None
         output = self.quant_method.apply(self, x, bias)
+        
+        if not self.return_bias:
+            return output
         output_bias = self.bias if self.skip_bias_add else None
         return output, output_bias
 
@@ -281,9 +289,10 @@ class ColumnParallelLinear(LinearBase):
                  params_dtype: Optional[torch.dtype] = None,
                  quant_config: Optional[QuantizationConfig] = None,
                  output_sizes: Optional[List[int]] = None,
-                 prefix: str = ""):
+                 prefix: str = "",
+                 return_bias: bool = True):
         super().__init__(input_size, output_size, skip_bias_add, params_dtype,
-                         quant_config, prefix)
+                         quant_config, prefix, return_bias=return_bias)
 
         self.gather_output = gather_output
 
@@ -375,6 +384,9 @@ class ColumnParallelLinear(LinearBase):
             output = tensor_model_parallel_all_gather(output_parallel)
         else:
             output = output_parallel
+        
+        if not self.return_bias:
+            return output
         output_bias = self.bias if self.skip_bias_add else None
         return output, output_bias
 
@@ -418,7 +430,8 @@ class MergedColumnParallelLinear(ColumnParallelLinear):
                  skip_bias_add: bool = False,
                  params_dtype: Optional[torch.dtype] = None,
                  quant_config: Optional[QuantizationConfig] = None,
-                 prefix: str = ""):
+                 prefix: str = "",
+                 return_bias: bool = True):
         self.output_sizes = output_sizes
         tp_size = get_tensor_model_parallel_world_size()
         assert all(output_size % tp_size == 0 for output_size in output_sizes)
@@ -429,7 +442,8 @@ class MergedColumnParallelLinear(ColumnParallelLinear):
                          skip_bias_add=skip_bias_add,
                          params_dtype=params_dtype,
                          quant_config=quant_config,
-                         prefix=prefix)
+                         prefix=prefix,
+                         return_bias=return_bias)
 
     def weight_loader(self,
                       param: Parameter,
@@ -653,7 +667,8 @@ class QKVParallelLinear(ColumnParallelLinear):
                  skip_bias_add: bool = False,
                  params_dtype: Optional[torch.dtype] = None,
                  quant_config: Optional[QuantizationConfig] = None,
-                 prefix: str = ""):
+                 prefix: str = "",
+                 return_bias: bool = True):
         self.hidden_size = hidden_size
         self.head_size = head_size
         self.total_num_heads = total_num_heads
@@ -686,7 +701,8 @@ class QKVParallelLinear(ColumnParallelLinear):
                          skip_bias_add=skip_bias_add,
                          params_dtype=params_dtype,
                          quant_config=quant_config,
-                         prefix=prefix)
+                         prefix=prefix,
+                         return_bias=return_bias)
 
     def _get_shard_offset_mapping(self, loaded_shard_id: str):
         shard_offset_mapping = {
@@ -980,9 +996,10 @@ class RowParallelLinear(LinearBase):
                  params_dtype: Optional[torch.dtype] = None,
                  reduce_results: bool = True,
                  quant_config: Optional[QuantizationConfig] = None,
-                 prefix: str = ""):
+                 prefix: str = "",
+                 return_bias: bool = True):
         super().__init__(input_size, output_size, skip_bias_add, params_dtype,
-                         quant_config, prefix)
+                         quant_config, prefix, return_bias=return_bias)
 
         self.input_is_parallel = input_is_parallel
         self.reduce_results = reduce_results
@@ -1086,8 +1103,9 @@ class RowParallelLinear(LinearBase):
         else:
             output = output_parallel
 
+        if not self.return_bias:
+            return output
         output_bias = self.bias if self.skip_bias_add else None
-
         return output, output_bias
 
     def extra_repr(self) -> str:

vllm-v0.6.2/vllm/model_executor/model_loader/loader.py

@@ -89,15 +89,63 @@ def device_loading_context(module: torch.nn.Module,
 logger = init_logger(__name__)
 
 
+def _get_device_memory_info_loader():
+    """Get device memory info for debug logging. Returns dict or None."""
+    try:
+        import torch.mlu
+        allocated = torch.mlu.memory_allocated() / (1024 ** 3)
+        reserved = torch.mlu.memory_reserved() / (1024 ** 3)
+        free, total = torch.mlu.mem_get_info()
+        return {"allocated": allocated, "reserved": reserved,
+                "free": free / (1024 ** 3), "total": total / (1024 ** 3)}
+    except Exception:
+        pass
+    try:
+        if torch.cuda.is_available():
+            allocated = torch.cuda.memory_allocated() / (1024 ** 3)
+            reserved = torch.cuda.memory_reserved() / (1024 ** 3)
+            free, total = torch.cuda.mem_get_info()
+            return {"allocated": allocated, "reserved": reserved,
+                    "free": free / (1024 ** 3), "total": total / (1024 ** 3)}
+    except Exception:
+        pass
+    return None
+
+
+def _log_mem(tag: str):
+    info = _get_device_memory_info_loader()
+    if info:
+        logger.info(
+            "[DEBUG-MEM] %s: allocated=%.2f GiB, reserved=%.2f GiB, "
+            "free=%.2f GiB, total=%.2f GiB",
+            tag, info["allocated"], info["reserved"],
+            info["free"], info["total"])
+
+
 def _initialize_model(vllm_config: VllmConfig, prefix: str = "") -> nn.Module:
     """Initialize a model with the given configurations."""
     model_config = vllm_config.model_config
     model_class, _ = get_model_architecture(model_config)
+    logger.info("[DEBUG-MEM] Model class: %s, dtype: %s",
+                model_class.__name__, model_config.dtype)
+    _log_mem("Before _initialize_model")
     signatures = inspect.signature(model_class.__init__)
     all_params = [param.name for param in signatures.parameters.values()]
     if "vllm_config" in all_params and "prefix" in all_params:
         # new-style model class
-        return model_class(vllm_config=vllm_config, prefix=prefix)
+        model = model_class(vllm_config=vllm_config, prefix=prefix)
+        _log_mem("After _initialize_model (empty weights created)")
+        # Print model parameter summary
+        total_params = 0
+        total_bytes = 0
+        for name, param in model.named_parameters():
+            total_params += param.numel()
+            total_bytes += param.numel() * param.element_size()
+        logger.info(
+            "[DEBUG-MEM] Model params: %d, "
+            "estimated size: %.2f GiB",
+            total_params, total_bytes / (1024 ** 3))
+        return model
     msg = ("vLLM model class should accept `vllm_config` and `prefix` as "
            "input arguments. Possibly you have an old-style model class"
            " registered from out of tree and it is used for new vLLM version. "
@@ -327,11 +375,14 @@ class DefaultModelLoader(BaseModelLoader):
         model_config = vllm_config.model_config
 
         target_device = torch.device(device_config.device)
+        _log_mem("load_model start, target_device=%s" % target_device)
         with set_default_torch_dtype(model_config.dtype):
             with target_device:
                 model = _initialize_model(vllm_config=vllm_config)
 
+            _log_mem("Before load_weights")
             model.load_weights(self._get_all_weights(model_config, model))
+            _log_mem("After load_weights")
 
             for _, module in model.named_modules():
                 quant_method = getattr(module, "quant_method", None)

vllm-v0.6.2/vllm/model_executor/model_loader/utils.py

@@ -32,7 +32,13 @@ def get_model_architecture(
             and "MixtralForCausalLM" in architectures):
         architectures = ["QuantMixtralForCausalLM"]
 
-    return ModelRegistry.resolve_model_cls(architectures)
+    return ModelRegistry.resolve_model_cls(
+        architectures,
+        model_path=model_config.model,
+        revision=model_config.revision,
+        trust_remote_code=model_config.trust_remote_code,
+        hf_config=model_config.hf_config,
+    )
 
 
 def get_architecture_class_name(model_config: ModelConfig) -> str:

vllm-v0.6.2/vllm/model_executor/models/gemma3.py

@@ -0,0 +1,507 @@
+# Copyright 2024 The vLLM team.
+# Copyright 2024 Google Inc. HuggingFace Inc. team. All rights reserved.
+#
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Inference-only Gemma3 model compatible with HuggingFace weights."""
+from typing import Iterable, List, Optional, Set, Tuple, Union
+
+import torch
+from torch import nn
+
+from vllm.attention import Attention, AttentionMetadata
+from vllm.config import CacheConfig, VllmConfig
+from vllm.distributed import get_pp_group, get_tensor_model_parallel_world_size
+from vllm.logger import init_logger
+from vllm.model_executor.layers.activation import GeluAndMul
+from vllm.model_executor.layers.layernorm import GemmaRMSNorm
+from vllm.model_executor.layers.linear import (MergedColumnParallelLinear,
+                                               QKVParallelLinear,
+                                               RowParallelLinear)
+from vllm.model_executor.layers.logits_processor import LogitsProcessor
+from vllm.model_executor.layers.quantization import QuantizationConfig
+from vllm.model_executor.layers.rotary_embedding import get_rope
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
+from vllm.model_executor.layers.vocab_parallel_embedding import (
+    VocabParallelEmbedding)
+from vllm.model_executor.model_loader.weight_utils import default_weight_loader
+from vllm.model_executor.sampling_metadata import SamplingMetadata
+from vllm.sequence import IntermediateTensors
+
+from .interfaces import SupportsLoRA, SupportsPP
+from .utils import (AutoWeightsLoader, is_pp_missing_parameter,
+                    make_empty_intermediate_tensors_factory, make_layers,
+                    maybe_prefix)
+
+logger = init_logger(__name__)
+
+
+class Gemma3MLP(nn.Module):
+
+    def __init__(
+        self,
+        hidden_size: int,
+        intermediate_size: int,
+        hidden_activation: str,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.gate_up_proj = MergedColumnParallelLinear(
+            hidden_size, [intermediate_size] * 2,
+            bias=False,
+            quant_config=quant_config)
+        self.down_proj = RowParallelLinear(intermediate_size,
+                                           hidden_size,
+                                           bias=False,
+                                           quant_config=quant_config)
+        if hidden_activation != "gelu_pytorch_tanh":
+            raise ValueError(
+                "Gemma3 uses `gelu_pytorch_tanh` as the hidden activation "
+                "function. Please set `hidden_activation` to "
+                "`gelu_pytorch_tanh`.")
+        self.act_fn = GeluAndMul(approximate="tanh")
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        gate_up, _ = self.gate_up_proj(x)
+        x = self.act_fn(gate_up)
+        x, _ = self.down_proj(x)
+        return x
+
+
+class Gemma3Attention(nn.Module):
+
+    def __init__(self,
+                 layer_idx: int,
+                 config,
+                 hidden_size: int,
+                 num_heads: int,
+                 num_kv_heads: int,
+                 head_dim: int,
+                 max_position_embeddings: int,
+                 cache_config: Optional[CacheConfig] = None,
+                 quant_config: Optional[QuantizationConfig] = None,
+                 attn_logits_soft_cap: Optional[float] = None) -> None:
+        super().__init__()
+        self.layer_idx = layer_idx
+        self.config = config
+        self.hidden_size = hidden_size
+        tp_size = get_tensor_model_parallel_world_size()
+        self.total_num_heads = num_heads
+        assert self.total_num_heads % tp_size == 0
+        self.num_heads = self.total_num_heads // tp_size
+        self.total_num_kv_heads = num_kv_heads
+        if self.total_num_kv_heads >= tp_size:
+            assert self.total_num_kv_heads % tp_size == 0
+        else:
+            assert tp_size % self.total_num_kv_heads == 0
+        self.num_kv_heads = max(1, self.total_num_kv_heads // tp_size)
+        self.head_dim = head_dim
+        self.q_size = self.num_heads * self.head_dim
+        self.kv_size = self.num_kv_heads * self.head_dim
+        self.scaling = config.query_pre_attn_scalar**-0.5
+
+        # Extract rope_theta from config, compatible with both old-style
+        # (config.rope_theta) and new-style (config.rope_parameters dict).
+        rope_params = getattr(config, "rope_parameters", None)
+        if hasattr(config, "rope_theta"):
+            self.rope_theta = config.rope_theta
+        elif isinstance(rope_params, dict):
+            # Transformers v5: nested per layer_type
+            if "full_attention" in rope_params:
+                self.rope_theta = rope_params["full_attention"].get(
+                    "rope_theta", 10000.0)
+            else:
+                # Transformers v4: flat dict
+                self.rope_theta = rope_params.get("rope_theta", 10000.0)
+        else:
+            self.rope_theta = 10000.0
+
+        self.qkv_proj = QKVParallelLinear(
+            hidden_size,
+            self.head_dim,
+            self.total_num_heads,
+            self.total_num_kv_heads,
+            bias=config.attention_bias,
+            quant_config=quant_config,
+        )
+        self.o_proj = RowParallelLinear(
+            self.total_num_heads * self.head_dim,
+            hidden_size,
+            bias=config.attention_bias,
+            quant_config=quant_config,
+        )
+
+        # Gemma3 specific: QK normalization
+        self.q_norm = GemmaRMSNorm(self.head_dim, eps=config.rms_norm_eps)
+        self.k_norm = GemmaRMSNorm(self.head_dim, eps=config.rms_norm_eps)
+
+        # Determine layer type and rope config
+        layer_types = getattr(config, "layer_types", None)
+        if layer_types is not None:
+            layer_type = layer_types[layer_idx]
+            self.is_sliding = (layer_type == "sliding_attention")
+        else:
+            self.is_sliding = (layer_idx % 2 == 1
+                               and config.sliding_window is not None)
+
+        # Extract rope config, compatible with both old-style (rope_theta,
+        # rope_scaling) and new-style (rope_parameters dict) transformers.
+        rope_params = getattr(config, "rope_parameters", None)
+
+        # Set up rope based on layer type
+        if self.is_sliding:
+            # Local/sliding attention uses rope_local_base_freq
+            if hasattr(config, "rope_local_base_freq"):
+                local_base = config.rope_local_base_freq
+            elif (isinstance(rope_params, dict)
+                  and "sliding_attention" in rope_params):
+                local_base = rope_params["sliding_attention"].get(
+                    "rope_theta", self.rope_theta)
+            else:
+                local_base = self.rope_theta
+            self.rotary_emb = get_rope(
+                self.head_dim,
+                rotary_dim=self.head_dim,
+                max_position=max_position_embeddings,
+                base=local_base,
+                is_neox_style=True,
+            )
+        else:
+            # Global attention: extract rope_base and rope_scaling.
+            # Prioritize rope_parameters dict (newer transformers) to
+            # avoid passing nested dicts that are unhashable.
+            rope_scaling = None
+            rope_base = self.rope_theta
+            if isinstance(rope_params, dict):
+                # Transformers v5: per layer_type sub-dicts
+                if "full_attention" in rope_params:
+                    rp = rope_params["full_attention"]
+                else:
+                    # Transformers v4: flat dict
+                    rp = rope_params
+                rope_base = rp.get("rope_theta", self.rope_theta)
+                rtype = rp.get("rope_type", None)
+                if rtype and rtype != "default":
+                    rope_scaling = {
+                        k: v for k, v in rp.items()
+                        if k not in ("rope_theta",)
+                    }
+            else:
+                # Fallback: old-style config.rope_scaling (flat dict)
+                rope_scaling = getattr(config, "rope_scaling", None)
+            self.rotary_emb = get_rope(
+                self.head_dim,
+                rotary_dim=self.head_dim,
+                max_position=max_position_embeddings,
+                base=rope_base,
+                is_neox_style=True,
+                rope_scaling=rope_scaling,
+            )
+
+        # NOTE: Like Gemma2, vLLM currently ignores sliding window
+        # and uses global attention for all layers.
+        self.attn = Attention(self.num_heads,
+                              self.head_dim,
+                              self.scaling,
+                              num_kv_heads=self.num_kv_heads,
+                              cache_config=cache_config,
+                              quant_config=quant_config,
+                              logits_soft_cap=attn_logits_soft_cap)
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        kv_cache: torch.Tensor,
+        attn_metadata: AttentionMetadata,
+    ) -> torch.Tensor:
+        qkv, _ = self.qkv_proj(hidden_states)
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size],
+                             dim=-1)
+
+        # Gemma3 specific: apply QK normalization
+        q = q.unflatten(-1, (self.num_heads, self.head_dim))
+        q = self.q_norm(q)
+        q = q.flatten(-2, -1)
+        k = k.unflatten(-1, (self.num_kv_heads, self.head_dim))
+        k = self.k_norm(k)
+        k = k.flatten(-2, -1)
+
+        # MLU rotary_emb expects a single concatenated tensor, not
+        # separate q and k (forward_mlu signature differs from forward_native).
+        qk = torch.cat([q, k], dim=-1)
+        self.rotary_emb(positions,
+                        qk.view(-1, self.num_heads + self.num_kv_heads,
+                                self.head_dim))
+        q, k = qk.split([self.q_size, self.kv_size], dim=-1)
+        attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
+        output, _ = self.o_proj(attn_output)
+        return output
+
+
+class Gemma3DecoderLayer(nn.Module):
+
+    def __init__(
+        self,
+        layer_idx: int,
+        config,
+        cache_config: Optional[CacheConfig] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        self.self_attn = Gemma3Attention(
+            layer_idx=layer_idx,
+            config=config,
+            hidden_size=self.hidden_size,
+            num_heads=config.num_attention_heads,
+            num_kv_heads=config.num_key_value_heads,
+            head_dim=config.head_dim,
+            max_position_embeddings=config.max_position_embeddings,
+            cache_config=cache_config,
+            quant_config=quant_config,
+            # Gemma3 does not use attn logit softcapping
+            attn_logits_soft_cap=getattr(config,
+                                         "attn_logit_softcapping", None),
+        )
+        self.hidden_size = config.hidden_size
+        self.mlp = Gemma3MLP(
+            hidden_size=self.hidden_size,
+            intermediate_size=config.intermediate_size,
+            hidden_activation=config.hidden_activation,
+            quant_config=quant_config,
+        )
+        self.input_layernorm = GemmaRMSNorm(config.hidden_size,
+                                            eps=config.rms_norm_eps)
+        self.post_attention_layernorm = GemmaRMSNorm(config.hidden_size,
+                                                     eps=config.rms_norm_eps)
+        self.pre_feedforward_layernorm = GemmaRMSNorm(config.hidden_size,
+                                                      eps=config.rms_norm_eps)
+        self.post_feedforward_layernorm = GemmaRMSNorm(config.hidden_size,
+                                                       eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        kv_cache: torch.Tensor,
+        attn_metadata: AttentionMetadata,
+        residual: Optional[torch.Tensor],
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        if residual is None:
+            residual = hidden_states
+            hidden_states = self.input_layernorm(hidden_states)
+        else:
+            hidden_states, residual = self.input_layernorm(
+                hidden_states, residual)
+        hidden_states = self.self_attn(
+            positions=positions,
+            hidden_states=hidden_states,
+            kv_cache=kv_cache,
+            attn_metadata=attn_metadata,
+        )
+        hidden_states = self.post_attention_layernorm(hidden_states)
+
+        hidden_states, residual = self.pre_feedforward_layernorm(
+            hidden_states, residual)
+        hidden_states = self.mlp(hidden_states)
+        hidden_states = self.post_feedforward_layernorm(hidden_states)
+        return hidden_states, residual
+
+
+class Gemma3Model(nn.Module):
+
+    def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
+        super().__init__()
+        config = vllm_config.model_config.hf_config
+        cache_config = vllm_config.cache_config
+        quant_config = vllm_config.quant_config
+        self.config = config
+
+        self.embed_tokens = VocabParallelEmbedding(
+            config.vocab_size,
+            config.hidden_size,
+        )
+        self.start_layer, self.end_layer, self.layers = make_layers(
+            config.num_hidden_layers,
+            lambda prefix: Gemma3DecoderLayer(
+                int(prefix.split(".")[-1]),
+                config, cache_config, quant_config),
+            prefix=f"{prefix}.layers")
+        self.norm = GemmaRMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+        normalizer = self.config.hidden_size**0.5
+        self.register_buffer("normalizer", torch.tensor(normalizer))
+        self.make_empty_intermediate_tensors = (
+            make_empty_intermediate_tensors_factory(
+                ["hidden_states", "residual"], config.hidden_size))
+
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor],
+        positions: torch.Tensor,
+        kv_caches: List[torch.Tensor],
+        attn_metadata: AttentionMetadata,
+        intermediate_tensors: Optional[IntermediateTensors],
+        inputs_embeds: Optional[torch.Tensor] = None,
+    ) -> Union[torch.Tensor, IntermediateTensors]:
+        if get_pp_group().is_first_rank:
+            if inputs_embeds is not None:
+                hidden_states = inputs_embeds
+            else:
+                hidden_states = self.embed_tokens(input_ids)
+            hidden_states *= self.normalizer
+            residual = None
+        else:
+            assert intermediate_tensors is not None
+            hidden_states = intermediate_tensors["hidden_states"]
+            residual = intermediate_tensors["residual"]
+        for i in range(self.start_layer, self.end_layer):
+            layer = self.layers[i]
+            hidden_states, residual = layer(
+                positions,
+                hidden_states,
+                kv_caches[i - self.start_layer],
+                attn_metadata,
+                residual,
+            )
+        if not get_pp_group().is_last_rank:
+            return IntermediateTensors({
+                "hidden_states": hidden_states,
+                "residual": residual
+            })
+        hidden_states, _ = self.norm(hidden_states, residual)
+        return hidden_states
+
+    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
+        stacked_params_mapping = [
+            # (param_name, shard_name, shard_id)
+            ("qkv_proj", "q_proj", "q"),
+            ("qkv_proj", "k_proj", "k"),
+            ("qkv_proj", "v_proj", "v"),
+            ("gate_up_proj", "gate_proj", 0),
+            ("gate_up_proj", "up_proj", 1),
+        ]
+        params_dict = dict(self.named_parameters())
+        loaded_params: Set[str] = set()
+        for name, loaded_weight in weights:
+            for (param_name, shard_name, shard_id) in stacked_params_mapping:
+                if shard_name not in name:
+                    continue
+                name = name.replace(shard_name, param_name)
+                if name.endswith(".bias") and name not in params_dict:
+                    continue
+                if is_pp_missing_parameter(name, self):
+                    continue
+                param = params_dict[name]
+                weight_loader = param.weight_loader
+                weight_loader(param, loaded_weight, shard_id)
+                break
+            else:
+                if name.endswith(".bias") and name not in params_dict:
+                    continue
+                if is_pp_missing_parameter(name, self):
+                    continue
+                if name not in params_dict:
+                    continue
+                param = params_dict[name]
+                weight_loader = getattr(param, "weight_loader",
+                                        default_weight_loader)
+                weight_loader(param, loaded_weight)
+            loaded_params.add(name)
+
+        unloaded_params = params_dict.keys() - loaded_params
+        if unloaded_params:
+            logger.warning(
+                "Some weights are not initialized from checkpoints: %s",
+                unloaded_params)
+
+
+class Gemma3ForCausalLM(nn.Module, SupportsLoRA, SupportsPP):
+    packed_modules_mapping = {
+        "qkv_proj": [
+            "q_proj",
+            "k_proj",
+            "v_proj",
+        ],
+        "gate_up_proj": [
+            "gate_proj",
+            "up_proj",
+        ],
+    }
+
+    # LoRA specific attributes
+    supported_lora_modules = [
+        "qkv_proj",
+        "o_proj",
+        "gate_up_proj",
+        "down_proj",
+    ]
+    embedding_modules = {}
+    embedding_padding_modules = []
+
+    def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
+        config = vllm_config.model_config.hf_config
+        quant_config = vllm_config.quant_config
+        lora_config = vllm_config.lora_config
+        del lora_config  # Unused.
+        super().__init__()
+        self.config = config
+        self.quant_config = quant_config
+        self.model = Gemma3Model(vllm_config=vllm_config,
+                                 prefix=maybe_prefix(prefix, "model"))
+
+        # Gemma3 may or may not have final_logit_softcapping
+        soft_cap = getattr(config, "final_logit_softcapping", None)
+        self.logits_processor = LogitsProcessor(
+            config.vocab_size, soft_cap=soft_cap)
+        self.sampler = get_sampler()
+        self.make_empty_intermediate_tensors = (
+            self.model.make_empty_intermediate_tensors)
+
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        kv_caches: List[torch.Tensor],
+        attn_metadata: AttentionMetadata,
+        intermediate_tensors: Optional[IntermediateTensors] = None,
+    ) -> Union[torch.Tensor, IntermediateTensors]:
+        hidden_states = self.model(input_ids, positions, kv_caches,
+                                   attn_metadata, intermediate_tensors)
+        return hidden_states
+
+    def compute_logits(
+        self,
+        hidden_states: torch.Tensor,
+        sampling_metadata: SamplingMetadata,
+    ) -> Optional[torch.Tensor]:
+        logits = self.logits_processor(self.model.embed_tokens,
+                                       hidden_states, sampling_metadata)
+        return logits
+
+    def sample(
+        self,
+        logits: torch.Tensor,
+        sampling_metadata: SamplingMetadata,
+    ) -> Optional[SamplerOutput]:
+        next_tokens = self.sampler(logits, sampling_metadata)
+        return next_tokens
+
+    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
+        loader = AutoWeightsLoader(
+            self,
+            skip_prefixes=(["lm_head."]
+                           if self.config.tie_word_embeddings else None),
+        )
+        loader.load_weights(weights)

vllm-v0.6.2/vllm/model_executor/models/llama.py

@@ -26,6 +26,10 @@ import torch
 from torch import nn
 from transformers import LlamaConfig
 
+from vllm.logger import init_logger
+
+logger = init_logger(__name__)
+
 from vllm.attention import Attention, AttentionMetadata
 from vllm.compilation.decorators import support_torch_compile
 from vllm.config import CacheConfig, VllmConfig
@@ -404,6 +408,12 @@ class LlamaModel(nn.Module):
                 if is_pp_missing_parameter(name, self):
                     continue
 
+                if name not in params_dict:
+                    logger.warning(
+                        "Skipping weight %s not present in the model",
+                        name)
+                    continue
+
                 param = params_dict[name]
                 weight_loader = getattr(param, "weight_loader",
                                         default_weight_loader)

vllm-v0.6.2/vllm/model_executor/models/mpt.py

@@ -272,7 +272,7 @@ class MPTForCausalLM(nn.Module, SupportsPP):
         config = vllm_config.model_config.hf_config
         quant_config = vllm_config.quant_config
         self.config = config
-        assert config.tie_word_embeddings
+        assert getattr(config, "tie_word_embeddings", True)
         self.quant_config = quant_config
 
         self.transformer = MPTModel(vllm_config=vllm_config,

vllm-v0.6.2/vllm/model_executor/models/qwen3_moe.py

@@ -0,0 +1,556 @@
+# Adapted from
+# https://github.com/huggingface/transformers/blob/v4.28.0/src/transformers/models/qwen2_moe/modeling_qwen2_moe.py
+# Copyright 2024 The Qwen team.
+# Copyright 2023 The vLLM team.
+# Copyright 2022 EleutherAI and the HuggingFace Inc. team. All rights reserved.
+#
+# This code is based on EleutherAI's GPT-NeoX library and the GPT-NeoX
+# and OPT implementations in this library. It has been modified from its
+# original forms to accommodate minor architectural differences compared
+# to GPT-NeoX and OPT used by the Meta AI team that trained the model.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Inference-only Qwen3MoE model compatible with HuggingFace weights."""
+from typing import Any, Dict, Iterable, List, Optional, Tuple, Union
+
+import torch
+import torch.nn.functional as F
+from torch import nn
+from transformers import PretrainedConfig
+
+from vllm.attention import Attention, AttentionMetadata
+from vllm.compilation.decorators import support_torch_compile
+from vllm.config import CacheConfig, VllmConfig
+from vllm.distributed import (get_pp_group,
+                              get_tensor_model_parallel_world_size,
+                              tensor_model_parallel_all_reduce)
+from vllm.model_executor.layers.activation import SiluAndMul
+from vllm.model_executor.layers.fused_moe import FusedMoE
+from vllm.model_executor.layers.layernorm import RMSNorm
+from vllm.model_executor.layers.linear import (MergedColumnParallelLinear,
+                                               QKVParallelLinear,
+                                               ReplicatedLinear,
+                                               RowParallelLinear)
+from vllm.model_executor.layers.logits_processor import LogitsProcessor
+from vllm.model_executor.layers.quantization import QuantizationConfig
+from vllm.model_executor.layers.rotary_embedding import get_rope
+from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
+from vllm.model_executor.layers.vocab_parallel_embedding import (
+    ParallelLMHead, VocabParallelEmbedding)
+from vllm.model_executor.model_loader.weight_utils import default_weight_loader
+from vllm.model_executor.sampling_metadata import SamplingMetadata
+from vllm.sequence import IntermediateTensors
+from vllm.utils import print_warning_once
+
+from .interfaces import SupportsPP
+from .utils import (is_pp_missing_parameter,
+                    make_empty_intermediate_tensors_factory, make_layers,
+                    maybe_prefix)
+
+
+class Qwen3MoeMLP(nn.Module):
+
+    def __init__(
+        self,
+        hidden_size: int,
+        intermediate_size: int,
+        hidden_act: str,
+        quant_config: Optional[QuantizationConfig] = None,
+        reduce_results: bool = True,
+    ) -> None:
+        super().__init__()
+        self.gate_up_proj = MergedColumnParallelLinear(
+            hidden_size, [intermediate_size] * 2,
+            bias=False,
+            quant_config=quant_config)
+        self.down_proj = RowParallelLinear(intermediate_size,
+                                           hidden_size,
+                                           bias=False,
+                                           quant_config=quant_config,
+                                           reduce_results=reduce_results)
+        if hidden_act != "silu":
+            raise ValueError(f"Unsupported activation: {hidden_act}. "
+                             "Only silu is supported for now.")
+        self.act_fn = SiluAndMul()
+
+    def forward(self, x):
+        gate_up, _ = self.gate_up_proj(x)
+        x = self.act_fn(gate_up)
+        x, _ = self.down_proj(x)
+        return x
+
+
+class Qwen3MoeSparseMoeBlock(nn.Module):
+
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+    ):
+        super().__init__()
+        self.tp_size = get_tensor_model_parallel_world_size()
+
+        if self.tp_size > config.num_experts:
+            raise ValueError(
+                f"Tensor parallel size {self.tp_size} is greater than "
+                f"the number of experts {config.num_experts}.")
+
+        self.experts = FusedMoE(num_experts=config.num_experts,
+                                top_k=config.num_experts_per_tok,
+                                hidden_size=config.hidden_size,
+                                intermediate_size=config.moe_intermediate_size,
+                                reduce_results=False,
+                                renormalize=config.norm_topk_prob,
+                                quant_config=quant_config)
+
+        self.gate = ReplicatedLinear(config.hidden_size,
+                                     config.num_experts,
+                                     bias=False,
+                                     quant_config=None)
+
+        shared_expert_intermediate_size = getattr(
+            config, "shared_expert_intermediate_size", 0)
+        if shared_expert_intermediate_size > 0:
+            self.shared_expert = Qwen3MoeMLP(
+                hidden_size=config.hidden_size,
+                intermediate_size=shared_expert_intermediate_size,
+                hidden_act=config.hidden_act,
+                quant_config=quant_config,
+                reduce_results=False,
+            )
+        else:
+            self.shared_expert = None
+
+        # Qwen3Moe uses ReplicatedLinear for shared_expert_gate
+        # (unlike Qwen2Moe which uses torch.nn.Linear)
+        self.shared_expert_gate = ReplicatedLinear(config.hidden_size,
+                                                   1,
+                                                   bias=False,
+                                                   quant_config=None)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        # NOTE: hidden_states can have either 1D or 2D shape.
+        orig_shape = hidden_states.shape
+        hidden_dim = hidden_states.shape[-1]
+        hidden_states = hidden_states.view(-1, hidden_dim)
+        shared_output = None
+        if self.shared_expert is not None:
+            shared_output = self.shared_expert(hidden_states)
+            if self.shared_expert_gate is not None:
+                shared_output = F.sigmoid(
+                    self.shared_expert_gate(hidden_states)[0]
+                ) * shared_output
+
+        # router_logits: (num_tokens, n_experts)
+        router_logits, _ = self.gate(hidden_states)
+        final_hidden_states = self.experts(hidden_states=hidden_states,
+                                           router_logits=router_logits)
+        if shared_output is not None:
+            final_hidden_states = final_hidden_states + shared_output
+        if self.tp_size > 1:
+            final_hidden_states = tensor_model_parallel_all_reduce(
+                final_hidden_states)
+
+        return final_hidden_states.view(orig_shape)
+
+
+class Qwen3MoeAttention(nn.Module):
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        num_kv_heads: int,
+        rope_theta: float = 10000,
+        rope_scaling: Optional[Dict[str, Any]] = None,
+        max_position_embeddings: int = 8192,
+        rms_norm_eps: float = 1e-06,
+        qkv_bias: bool = False,
+        cache_config: Optional[CacheConfig] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.hidden_size = hidden_size
+        tp_size = get_tensor_model_parallel_world_size()
+        self.total_num_heads = num_heads
+        assert self.total_num_heads % tp_size == 0
+        self.num_heads = self.total_num_heads // tp_size
+        self.total_num_kv_heads = num_kv_heads
+        if self.total_num_kv_heads >= tp_size:
+            assert self.total_num_kv_heads % tp_size == 0
+        else:
+            assert tp_size % self.total_num_kv_heads == 0
+        self.num_kv_heads = max(1, self.total_num_kv_heads // tp_size)
+        self.head_dim = hidden_size // self.total_num_heads
+        self.q_size = self.num_heads * self.head_dim
+        self.kv_size = self.num_kv_heads * self.head_dim
+        self.scaling = self.head_dim**-0.5
+        self.rope_theta = rope_theta
+        self.max_position_embeddings = max_position_embeddings
+
+        self.qkv_proj = QKVParallelLinear(
+            hidden_size,
+            self.head_dim,
+            self.total_num_heads,
+            self.total_num_kv_heads,
+            bias=qkv_bias,
+            quant_config=quant_config,
+        )
+
+        self.o_proj = RowParallelLinear(
+            self.total_num_heads * self.head_dim,
+            hidden_size,
+            bias=False,
+            quant_config=quant_config,
+        )
+
+        self.rotary_emb = get_rope(
+            self.head_dim,
+            rotary_dim=self.head_dim,
+            max_position=max_position_embeddings,
+            base=rope_theta,
+            rope_scaling=rope_scaling,
+        )
+        self.attn = Attention(self.num_heads,
+                              self.head_dim,
+                              self.scaling,
+                              num_kv_heads=self.num_kv_heads,
+                              cache_config=cache_config,
+                              quant_config=quant_config)
+
+        # Qwen3 specific: QK normalization
+        self.q_norm = RMSNorm(self.head_dim, eps=rms_norm_eps)
+        self.k_norm = RMSNorm(self.head_dim, eps=rms_norm_eps)
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        kv_cache: torch.Tensor,
+        attn_metadata: AttentionMetadata,
+    ) -> torch.Tensor:
+        qkv, _ = self.qkv_proj(hidden_states)
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size],
+                             dim=-1)
+
+        # Qwen3 specific: Apply QK normalization before rotary embedding
+        # Use .contiguous() to ensure memory layout is compatible with
+        # MLU's RMSNorm which uses .view() internally.
+        q_shape = q.shape
+        q_by_head = q.view(*q.shape[:-1], q.shape[-1] // self.head_dim,
+                           self.head_dim).contiguous()
+        q_by_head = self.q_norm(q_by_head)
+        q = q_by_head.reshape(q_shape)
+
+        k_shape = k.shape
+        k_by_head = k.view(*k.shape[:-1], k.shape[-1] // self.head_dim,
+                           self.head_dim).contiguous()
+        k_by_head = self.k_norm(k_by_head)
+        k = k_by_head.reshape(k_shape)
+
+        # MLU rotary_emb expects a single concatenated 3D tensor, not
+        # separate q and k (forward_mlu signature differs from forward_native).
+        qk = torch.cat([q, k], dim=-1)
+        self.rotary_emb(positions,
+                        qk.view(-1, self.num_heads + self.num_kv_heads,
+                                self.head_dim))
+        q, k = qk.split([self.q_size, self.kv_size], dim=-1)
+        attn_output = self.attn(q, k, v, kv_cache, attn_metadata)
+        output, _ = self.o_proj(attn_output)
+        return output
+
+
+class Qwen3MoeDecoderLayer(nn.Module):
+
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        layer_idx: int,
+        cache_config: Optional[CacheConfig] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.hidden_size = config.hidden_size
+        rope_theta = getattr(config, "rope_theta", 10000)
+        rope_scaling = getattr(config, "rope_scaling", None)
+        max_position_embeddings = getattr(config, "max_position_embeddings",
+                                          8192)
+        self.self_attn = Qwen3MoeAttention(
+            hidden_size=self.hidden_size,
+            num_heads=config.num_attention_heads,
+            num_kv_heads=config.num_key_value_heads,
+            rope_theta=rope_theta,
+            rope_scaling=rope_scaling,
+            max_position_embeddings=max_position_embeddings,
+            rms_norm_eps=config.rms_norm_eps,
+            qkv_bias=getattr(config, "attention_bias", False),
+            cache_config=cache_config,
+            quant_config=quant_config,
+        )
+
+        # Note: Qwen3MoE may not have `mlp_only_layers` in the config.
+        mlp_only_layers = ([] if not hasattr(config, "mlp_only_layers") else
+                           config.mlp_only_layers)
+        if (layer_idx not in mlp_only_layers) and (
+                config.num_experts > 0 and
+            (layer_idx + 1) % config.decoder_sparse_step == 0):
+            self.mlp = Qwen3MoeSparseMoeBlock(config=config,
+                                              quant_config=quant_config)
+        else:
+            self.mlp = Qwen3MoeMLP(
+                hidden_size=config.hidden_size,
+                intermediate_size=config.intermediate_size,
+                hidden_act=config.hidden_act,
+                quant_config=quant_config,
+            )
+        self.input_layernorm = RMSNorm(config.hidden_size,
+                                       eps=config.rms_norm_eps)
+        self.post_attention_layernorm = RMSNorm(config.hidden_size,
+                                                eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        kv_cache: torch.Tensor,
+        attn_metadata: AttentionMetadata,
+        residual: Optional[torch.Tensor],
+    ) -> torch.Tensor:
+        # Self Attention
+        if residual is None:
+            residual = hidden_states
+            hidden_states = self.input_layernorm(hidden_states)
+        else:
+            hidden_states, residual = self.input_layernorm(
+                hidden_states, residual)
+        hidden_states = self.self_attn(
+            positions=positions,
+            hidden_states=hidden_states,
+            kv_cache=kv_cache,
+            attn_metadata=attn_metadata,
+        )
+
+        # Fully Connected
+        hidden_states, residual = self.post_attention_layernorm(
+            hidden_states, residual)
+        hidden_states = self.mlp(hidden_states)
+        return hidden_states, residual
+
+
+@support_torch_compile
+class Qwen3MoeModel(nn.Module):
+
+    def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
+        super().__init__()
+
+        config = vllm_config.model_config.hf_config
+        cache_config = vllm_config.cache_config
+        quant_config = vllm_config.quant_config
+
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = VocabParallelEmbedding(
+            config.vocab_size,
+            config.hidden_size,
+        )
+        self.start_layer, self.end_layer, self.layers = make_layers(
+            config.num_hidden_layers,
+            lambda prefix: Qwen3MoeDecoderLayer(config=config,
+                                                layer_idx=int(
+                                                    prefix.split(".")[-1]),
+                                                cache_config=cache_config,
+                                                quant_config=quant_config),
+            prefix=f"{prefix}.layers",
+        )
+        self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+        self.make_empty_intermediate_tensors = (
+            make_empty_intermediate_tensors_factory(
+                ["hidden_states", "residual"], config.hidden_size))
+
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        kv_caches: List[torch.Tensor],
+        attn_metadata: AttentionMetadata,
+        intermediate_tensors: Optional[IntermediateTensors] = None,
+    ) -> Union[torch.Tensor, IntermediateTensors]:
+        if get_pp_group().is_first_rank:
+            hidden_states = self.embed_tokens(input_ids)
+            residual = None
+        else:
+            assert intermediate_tensors is not None
+            hidden_states = intermediate_tensors["hidden_states"]
+            residual = intermediate_tensors["residual"]
+        for i in range(self.start_layer, self.end_layer):
+            layer = self.layers[i]
+            hidden_states, residual = layer(positions, hidden_states,
+                                            kv_caches[i - self.start_layer],
+                                            attn_metadata, residual)
+        if not get_pp_group().is_last_rank:
+            return IntermediateTensors({
+                "hidden_states": hidden_states,
+                "residual": residual
+            })
+        hidden_states, _ = self.norm(hidden_states, residual)
+        return hidden_states
+
+
+class Qwen3MoeForCausalLM(nn.Module, SupportsPP):
+
+    fall_back_to_pt_during_load = False
+
+    def __init__(self, *, vllm_config: VllmConfig, prefix: str = ""):
+        super().__init__()
+        config = vllm_config.model_config.hf_config
+        quant_config = vllm_config.quant_config
+        self.config = config
+        self.quant_config = quant_config
+        self.model = Qwen3MoeModel(vllm_config=vllm_config,
+                                   prefix=maybe_prefix(prefix, "model"))
+        self.lm_head = ParallelLMHead(config.vocab_size,
+                                      config.hidden_size,
+                                      quant_config=quant_config)
+        if self.config.tie_word_embeddings:
+            self.lm_head.weight = self.model.embed_tokens.weight
+        self.logits_processor = LogitsProcessor(config.vocab_size)
+        self.sampler = get_sampler()
+        self.make_empty_intermediate_tensors = (
+            self.model.make_empty_intermediate_tensors)
+
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        kv_caches: List[torch.Tensor],
+        attn_metadata: AttentionMetadata,
+        intermediate_tensors: Optional[IntermediateTensors] = None,
+    ) -> Union[torch.Tensor, IntermediateTensors]:
+        hidden_states = self.model(input_ids, positions, kv_caches,
+                                   attn_metadata, intermediate_tensors)
+        return hidden_states
+
+    def compute_logits(
+        self,
+        hidden_states: torch.Tensor,
+        sampling_metadata: SamplingMetadata,
+    ) -> Optional[torch.Tensor]:
+        logits = self.logits_processor(self.lm_head, hidden_states,
+                                       sampling_metadata)
+        return logits
+
+    def sample(
+        self,
+        logits: Optional[torch.Tensor],
+        sampling_metadata: SamplingMetadata,
+    ) -> Optional[SamplerOutput]:
+        next_tokens = self.sampler(logits, sampling_metadata)
+        return next_tokens
+
+    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
+        stacked_params_mapping = [
+            # (param_name, shard_name, shard_id)
+            ("qkv_proj", "q_proj", "q"),
+            ("qkv_proj", "k_proj", "k"),
+            ("qkv_proj", "v_proj", "v"),
+            ("gate_up_proj", "gate_proj", 0),
+            ("gate_up_proj", "up_proj", 1),
+        ]
+
+        # Params for weights, fp8 weight scales, fp8 activation scales
+        # (param_name, weight_name, expert_id, shard_id)
+        expert_params_mapping = FusedMoE.make_expert_params_mapping(
+            ckpt_gate_proj_name="gate_proj",
+            ckpt_down_proj_name="down_proj",
+            ckpt_up_proj_name="up_proj",
+            num_experts=self.config.num_experts)
+
+        params_dict = dict(self.named_parameters())
+        for name, loaded_weight in weights:
+            if "rotary_emb.inv_freq" in name:
+                continue
+            for (param_name, weight_name, shard_id) in stacked_params_mapping:
+                # Skip non-stacked layers and experts (experts handled below).
+                if weight_name not in name:
+                    continue
+                # We have mlp.experts[0].gate_proj in the checkpoint.
+                # Since we handle the experts below in expert_params_mapping,
+                # we need to skip here BEFORE we update the name, otherwise
+                # name will be updated to mlp.experts[0].gate_up_proj, which
+                # will then be updated below in expert_params_mapping
+                # for mlp.experts[0].gate_gate_up_proj, which breaks load.
+                if "mlp.experts" in name:
+                    continue
+                name = name.replace(weight_name, param_name)
+                # Skip loading extra bias for GPTQ models.
+                if ((name.endswith(".bias") or name.endswith("_bias"))
+                        and name not in params_dict):
+                    continue
+                # Skip layers on other devices.
+                if is_pp_missing_parameter(name, self):
+                    continue
+                if name not in params_dict:
+                    continue
+
+                param = params_dict[name]
+                weight_loader = param.weight_loader
+                weight_loader(param, loaded_weight, shard_id)
+                break
+            else:
+                for mapping in expert_params_mapping:
+                    param_name, weight_name, expert_id, shard_id = mapping
+                    if weight_name not in name:
+                        continue
+                    name = name.replace(weight_name, param_name)
+                    # Skip layers on other devices.
+                    if is_pp_missing_parameter(name, self):
+                        continue
+                    # Skip loading extra bias for GPTQ models.
+                    if ((name.endswith(".bias") or name.endswith("_bias"))
+                            and name not in params_dict):
+                        continue
+                    param = params_dict[name]
+                    weight_loader = param.weight_loader
+                    weight_loader(param,
+                                  loaded_weight,
+                                  name,
+                                  shard_id=shard_id,
+                                  expert_id=expert_id)
+                    break
+                else:
+                    # Skip loading extra bias for GPTQ models.
+                    if ((name.endswith(".bias") or name.endswith("_bias"))
+                            and name not in params_dict):
+                        continue
+                    # Skip layers on other devices.
+                    if is_pp_missing_parameter(name, self):
+                        continue
+                    # Remapping the name of FP8 kv-scale.
+                    if name.endswith("kv_scale"):
+                        remapped_kv_scale_name = name.replace(
+                            ".kv_scale", ".attn.kv_scale")
+                        if remapped_kv_scale_name not in params_dict:
+                            print_warning_once(
+                                "Found kv scale in the checkpoint "
+                                f"(e.g. {name}), but not found the expected "
+                                f"name in the model "
+                                f"(e.g. {remapped_kv_scale_name}). "
+                                "kv-scale is not loaded.")
+                            continue
+                        else:
+                            name = remapped_kv_scale_name
+                    param = params_dict[name]
+                    weight_loader = getattr(param, "weight_loader",
+                                            default_weight_loader)
+                    weight_loader(param, loaded_weight)

vllm-v0.6.2/vllm/model_executor/models/registry.py

@@ -16,9 +16,11 @@ from typing import (AbstractSet, Callable, Dict, List, Optional, Tuple, Type,
 
 import cloudpickle
 import torch.nn as nn
+import transformers
 
 from vllm.logger import init_logger
 from vllm.platforms import current_platform
+from vllm.transformers_utils.dynamic_module import try_get_class_from_dynamic_module
 
 from .interfaces import (has_inner_state, is_attention_free,
                          supports_multimodal, supports_pp)
@@ -26,6 +28,9 @@ from .interfaces_base import is_embedding_model, is_text_generation_model
 
 logger = init_logger(__name__)
 
+# Cache for architectures that have already been logged
+_logged_transformers_architectures: set = set()
+
 # yapf: disable
 _TEXT_GENERATION_MODELS = {
     # [Decoder-only]
@@ -47,6 +52,7 @@ _TEXT_GENERATION_MODELS = {
     "FalconForCausalLM": ("falcon", "FalconForCausalLM"),
     "GemmaForCausalLM": ("gemma", "GemmaForCausalLM"),
     "Gemma2ForCausalLM": ("gemma2", "Gemma2ForCausalLM"),
+    "Gemma3ForCausalLM": ("gemma3", "Gemma3ForCausalLM"),
     "GPT2LMHeadModel": ("gpt2", "GPT2LMHeadModel"),
     "GPTBigCodeForCausalLM": ("gpt_bigcode", "GPTBigCodeForCausalLM"),
     "GPTJForCausalLM": ("gpt_j", "GPTJForCausalLM"),
@@ -85,6 +91,7 @@ _TEXT_GENERATION_MODELS = {
     "Qwen2ForCausalLM": ("qwen2", "Qwen2ForCausalLM"),
     "Qwen2MoeForCausalLM": ("qwen2_moe", "Qwen2MoeForCausalLM"),
     "Qwen3ForCausalLM": ("qwen3", "Qwen3ForCausalLM"),
+    "Qwen3MoeForCausalLM": ("qwen3_moe", "Qwen3MoeForCausalLM"),
     "RWForCausalLM": ("falcon", "FalconForCausalLM"),
     "StableLMEpochForCausalLM": ("stablelm", "StablelmForCausalLM"),
     "StableLmForCausalLM": ("stablelm", "StablelmForCausalLM"),
@@ -157,6 +164,13 @@ _SPECULATIVE_DECODING_MODELS = {
     "MedusaModel": ("medusa", "Medusa"),
     "MLPSpeculatorPreTrainedModel": ("mlp_speculator", "MLPSpeculator"),
 }
+
+# Transformers backend models - wrapper classes for custom HuggingFace models
+# These provide the vLLM interface for models loaded via auto_map
+_TRANSFORMERS_BACKEND_MODELS = {
+    # Text generation models
+    "TransformersForCausalLM": ("transformers", "TransformersForCausalLM"),
+}
 # yapf: enable
 
 _VLLM_MODELS = {
@@ -164,6 +178,7 @@ _VLLM_MODELS = {
     **_EMBEDDING_MODELS,
     **_MULTIMODAL_MODELS,
     **_SPECULATIVE_DECODING_MODELS,
+    **_TRANSFORMERS_BACKEND_MODELS,
 }
 
 # Models not supported by ROCm.
@@ -369,6 +384,94 @@ class _ModelRegistry:
 
         return _try_inspect_model_cls(model_arch, self.models[model_arch])
 
+    def _try_resolve_transformers(
+        self,
+        architecture: str,
+        model_path: str,
+        revision: Optional[str],
+        trust_remote_code: bool,
+        hf_config: Optional[object] = None,
+    ) -> Optional[str]:
+        """
+        Try to resolve a model architecture using the Transformers backend.
+        This allows loading custom models that define their own implementation
+        via the `auto_map` field in config.json.
+        
+        Returns the vLLM wrapper architecture name (e.g. "TransformersForCausalLM")
+        if the model can be loaded via auto_map, None otherwise.
+        """
+        # If architecture is already a transformers backend model, return it
+        if architecture in _TRANSFORMERS_BACKEND_MODELS:
+            return architecture
+        
+        # Check if architecture exists in transformers library
+        model_module = getattr(transformers, architecture, None)
+        if model_module is not None:
+            # Model exists in transformers, can use TransformersForCausalLM wrapper
+            # Only log once per architecture to avoid spam
+            if architecture not in _logged_transformers_architectures:
+                _logged_transformers_architectures.add(architecture)
+                logger.info(
+                    "Architecture %s found in transformers library, "
+                    "using TransformersForCausalLM wrapper",
+                    architecture
+                )
+            return "TransformersForCausalLM"
+        
+        # Get auto_map from hf_config
+        auto_map: Dict[str, str] = {}
+        if hf_config is not None:
+            auto_map = getattr(hf_config, "auto_map", None) or {}
+        
+        if not auto_map:
+            return None
+        
+        # Try to load from auto_map to verify it works
+        # First, ensure config class is loaded
+        for name, module in auto_map.items():
+            if name.startswith("AutoConfig"):
+                try_get_class_from_dynamic_module(
+                    module,
+                    model_path,
+                    trust_remote_code=trust_remote_code,
+                    revision=revision,
+                    warn_on_fail=False,
+                )
+        
+        # Check if auto_map has a model class we can use
+        # Priority: AutoModelForCausalLM > AutoModelForSeq2SeqLM > AutoModel
+        auto_model_keys = sorted(
+            [k for k in auto_map.keys() if k.startswith("AutoModel")],
+            key=lambda x: (0 if "ForCausalLM" in x else (1 if "ForSeq2Seq" in x else 2))
+        )
+        
+        for name in auto_model_keys:
+            module = auto_map[name]
+            model_cls = try_get_class_from_dynamic_module(
+                module,
+                model_path,
+                trust_remote_code=trust_remote_code,
+                revision=revision,
+                warn_on_fail=True,
+            )
+            if model_cls is not None:
+                # Only log once per model class to avoid spam
+                log_key = f"{model_cls.__name__}_{name}"
+                if not hasattr(self, '_logged_custom_models'):
+                    self._logged_custom_models = set()
+                if log_key not in self._logged_custom_models:
+                    logger.info(
+                        "Found custom model class %s from auto_map[%s], "
+                        "using TransformersForCausalLM wrapper",
+                        model_cls.__name__,
+                        name
+                    )
+                    self._logged_custom_models.add(log_key)
+                # Return the wrapper architecture, not the actual class
+                return "TransformersForCausalLM"
+        
+        return None
+
     def _normalize_archs(
         self,
         architectures: Union[str, List[str]],
@@ -383,6 +486,10 @@ class _ModelRegistry:
     def inspect_model_cls(
         self,
         architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
     ) -> _ModelInfo:
         architectures = self._normalize_archs(architectures)
 
@@ -391,11 +498,27 @@ class _ModelRegistry:
             if model_info is not None:
                 return model_info
 
+        # Fallback: try to resolve using transformers backend (auto_map)
+        if model_path and trust_remote_code and hf_config:
+            for arch in architectures:
+                wrapper_arch = self._try_resolve_transformers(
+                    arch, model_path, revision, trust_remote_code, hf_config
+                )
+                if wrapper_arch is not None:
+                    # Use the wrapper architecture's ModelInfo
+                    model_info = self._try_inspect_model_cls(wrapper_arch)
+                    if model_info is not None:
+                        return model_info
+
         return self._raise_for_unsupported(architectures)
 
     def resolve_model_cls(
         self,
         architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
     ) -> Tuple[Type[nn.Module], str]:
         architectures = self._normalize_archs(architectures)
 
@@ -404,39 +527,91 @@ class _ModelRegistry:
             if model_cls is not None:
                 return (model_cls, arch)
 
+        # Fallback: try to resolve using transformers backend (auto_map)
+        if model_path and trust_remote_code and hf_config:
+            for arch in architectures:
+                wrapper_arch = self._try_resolve_transformers(
+                    arch, model_path, revision, trust_remote_code, hf_config
+                )
+                if wrapper_arch is not None:
+                    model_cls = self._try_load_model_cls(wrapper_arch)
+                    if model_cls is not None:
+                        # Return wrapper class but keep original architecture name
+                        return (model_cls, arch)
+
         return self._raise_for_unsupported(architectures)
 
     def is_text_generation_model(
         self,
         architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
     ) -> bool:
-        return self.inspect_model_cls(architectures).is_text_generation_model
+        return self.inspect_model_cls(
+            architectures, model_path, revision, trust_remote_code, hf_config
+        ).is_text_generation_model
 
     def is_embedding_model(
         self,
         architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
     ) -> bool:
-        return self.inspect_model_cls(architectures).is_embedding_model
+        return self.inspect_model_cls(
+            architectures, model_path, revision, trust_remote_code, hf_config
+        ).is_embedding_model
 
     def is_multimodal_model(
         self,
         architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
     ) -> bool:
-        return self.inspect_model_cls(architectures).supports_multimodal
+        return self.inspect_model_cls(
+            architectures, model_path, revision, trust_remote_code, hf_config
+        ).supports_multimodal
 
     def is_pp_supported_model(
         self,
         architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
     ) -> bool:
-        return self.inspect_model_cls(architectures).supports_pp
+        return self.inspect_model_cls(
+            architectures, model_path, revision, trust_remote_code, hf_config
+        ).supports_pp
 
-    def model_has_inner_state(self, architectures: Union[str,
-                                                         List[str]]) -> bool:
-        return self.inspect_model_cls(architectures).has_inner_state
+    def model_has_inner_state(
+        self,
+        architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
+    ) -> bool:
+        return self.inspect_model_cls(
+            architectures, model_path, revision, trust_remote_code, hf_config
+        ).has_inner_state
 
-    def is_attention_free_model(self, architectures: Union[str,
-                                                           List[str]]) -> bool:
-        return self.inspect_model_cls(architectures).is_attention_free
+    def is_attention_free_model(
+        self,
+        architectures: Union[str, List[str]],
+        model_path: Optional[str] = None,
+        revision: Optional[str] = None,
+        trust_remote_code: bool = False,
+        hf_config: Optional[object] = None,
+    ) -> bool:
+        return self.inspect_model_cls(
+            architectures, model_path, revision, trust_remote_code, hf_config
+        ).is_attention_free
 
 
 ModelRegistry = _ModelRegistry({

vllm-v0.6.2/vllm/model_executor/models/transformers/__init__.py

@@ -0,0 +1,127 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright 2024 The vLLM team.
+"""Wrapper around `transformers` models for vLLM v0.6.2.
+
+This module provides the Transformers modeling backend that wraps
+any HuggingFace model with the vLLM interface, enabling support for custom
+models that define their implementation via `auto_map` in config.json.
+
+Architecture (following latest vLLM patterns):
+- Base: Core functionality (meta init, PP/TP support, module replacement, attention, weight loading)
+- CausalMixin: Causal LM specific (lm_head, compute_logits, sample)
+- EmbeddingMixin: Embedding/pooling specific (pooler, pooling)
+- SequenceClassificationMixin: Classification specific (classifier, pooling)
+
+Composed model classes:
+- TransformersForCausalLM = CausalMixin + Base
+- TransformersForEmbedding = EmbeddingMixin + Base
+- TransformersForSequenceClassification = SequenceClassificationMixin + Base
+
+Key optimizations:
+- Meta device initialization for memory efficiency
+- Pipeline Parallel support (PPMissingLayer)
+- Tensor Parallel support (tp_plan based module replacement)
+- Module replacement (Linear, RMSNorm, Embedding) with vLLM optimized versions
+- vLLM Attention instances for proper KV cache allocation
+- AutoWeightsLoader for efficient weight loading with name mapping
+"""
+
+from vllm.model_executor.models.transformers.base import (
+    Base,
+    set_attention_context,
+    clear_attention_context,
+    get_attention_context,
+    vllm_flash_attention_forward,
+)
+from vllm.model_executor.models.transformers.causal import CausalMixin
+from vllm.model_executor.models.transformers.pooling import (
+    EmbeddingMixin,
+    SequenceClassificationMixin,
+)
+from vllm.model_executor.models.transformers.legacy import LegacyMixin
+from vllm.model_executor.models.transformers.utils import (
+    init_on_device_without_buffers,
+    replace_linear_class,
+    replace_rms_norm_class,
+    log_replacement,
+    maybe_prefix,
+)
+
+
+# ============================================================================
+# Composed Model Classes (Mixin + Base pattern)
+# ============================================================================
+
+class TransformersForCausalLM(CausalMixin, Base):
+    """
+    Transformers backend wrapper for causal language models.
+    
+    Combines CausalMixin (lm_head, compute_logits, sample) with
+    Base (meta init, PP/TP support, module replacement, attention, weight loading).
+    
+    Supports any HuggingFace model with auto_map in config.json.
+    """
+    pass
+
+
+class TransformersForEmbedding(EmbeddingMixin, Base):
+    """
+    Transformers backend wrapper for embedding/sentence similarity models.
+    
+    Combines EmbeddingMixin (pooler, pooling) with
+    Base (meta init, PP/TP support, module replacement, attention, weight loading).
+    
+    Supports embedding models like BERT, sentence-transformers, etc.
+    """
+    pass
+
+
+class TransformersForSequenceClassification(SequenceClassificationMixin, Base):
+    """
+    Transformers backend wrapper for sequence classification models.
+    
+    Combines SequenceClassificationMixin (classifier, pooling) with
+    Base (meta init, PP/TP support, module replacement, attention, weight loading).
+    
+    Supports cross-encoders and classification models.
+    """
+    pass
+
+
+class TransformersForLegacy(LegacyMixin, EmbeddingMixin, Base):
+    """
+    Transformers backend wrapper for legacy/encoder models.
+    
+    Combines LegacyMixin (BERT/RoBERTa weight mapping, position handling) with
+    EmbeddingMixin (pooler) and Base (core functionality).
+    
+    Supports BERT, RoBERTa, and similar encoder models.
+    """
+    pass
+
+
+__all__ = [
+    # Main wrapper classes
+    "TransformersForCausalLM",
+    "TransformersForEmbedding",
+    "TransformersForSequenceClassification",
+    "TransformersForLegacy",
+    # Base class for extension
+    "Base",
+    # Mixin classes for custom combinations
+    "CausalMixin",
+    "EmbeddingMixin",
+    "SequenceClassificationMixin",
+    "LegacyMixin",
+    # Attention context management
+    "set_attention_context",
+    "clear_attention_context",
+    "get_attention_context",
+    "vllm_flash_attention_forward",
+    # Utility functions
+    "init_on_device_without_buffers",
+    "replace_linear_class",
+    "replace_rms_norm_class",
+    "log_replacement",
+    "maybe_prefix",
+]

vllm-v0.6.2/vllm/model_executor/models/transformers/base.py

@@ -0,0 +1,704 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright 2024 The vLLM team.
+"""Transformers modeling backend base class for v0.6.2.
+
+This module provides the Base class following latest vLLM architecture:
+- Meta device initialization for memory efficiency
+- Pipeline parallel support (PPMissingLayer)
+- Tensor parallel support (tp_plan based module replacement)
+- Module replacement (Linear, RMSNorm) with vLLM optimized versions
+- VocabParallelEmbedding for input embeddings
+- Attention instances for KV cache allocation
+- Weight loading with AutoWeightsLoader and WeightsMapper
+"""
+
+import re
+from typing import TYPE_CHECKING, Dict, Iterable, List, Optional, Set, Tuple
+
+import torch
+import torch.nn as nn
+
+from vllm.config import VllmConfig
+from vllm.distributed import get_pp_group, get_tp_group
+from vllm.distributed.utils import get_pp_indices
+from vllm.logger import init_logger
+from vllm.model_executor.layers.layernorm import RMSNorm
+from vllm.model_executor.layers.vocab_parallel_embedding import VocabParallelEmbedding
+from vllm.model_executor.models.utils import (
+    AutoWeightsLoader,
+    PPMissingLayer,
+    WeightsMapper,
+    make_empty_intermediate_tensors_factory,
+)
+from vllm.attention.layer import Attention
+from vllm.sequence import IntermediateTensors
+
+from .utils import (
+    init_on_device_without_buffers,
+    replace_linear_class,
+    replace_rms_norm_class,
+    log_replacement,
+    maybe_prefix,
+)
+
+if TYPE_CHECKING:
+    from transformers import PreTrainedModel
+    from vllm.attention import AttentionMetadata
+
+logger = init_logger(__name__)
+
+
+# ============================================================================
+# Attention Context Management (for vLLM attention integration)
+# ============================================================================
+
+_current_attn_metadata = None
+_current_kv_caches = None
+
+
+def set_attention_context(attn_metadata, kv_caches):
+    """Set the current attention context for vLLM attention functions."""
+    global _current_attn_metadata, _current_kv_caches
+    _current_attn_metadata = attn_metadata
+    _current_kv_caches = kv_caches
+
+
+def clear_attention_context():
+    """Clear the current attention context after forward pass."""
+    global _current_attn_metadata, _current_kv_caches
+    _current_attn_metadata = None
+    _current_kv_caches = None
+
+
+def get_attention_context():
+    """Get the current attention context."""
+    return _current_attn_metadata, _current_kv_caches
+
+
+# ============================================================================
+# vLLM Attention Function for Transformers Integration
+# ============================================================================
+
+def vllm_flash_attention_forward(
+    module: torch.nn.Module,
+    query: torch.Tensor,
+    key: torch.Tensor,
+    value: torch.Tensor,
+    attention_mask: torch.Tensor,
+    scaling: float = None,
+    attention_instances: Dict[int, Attention] = None,
+    **kwargs,
+):
+    """
+    vLLM's optimized attention function for transformers integration.
+    
+    In v0.6.2, Attention.forward signature is:
+        (query, key, value, kv_cache, attn_metadata)
+    """
+    layer_idx = getattr(module, 'layer_idx', 0)
+    
+    if attention_instances is None or layer_idx not in attention_instances:
+        return _standard_attention(query, key, value, attention_mask, scaling)
+    
+    self_attn = attention_instances[layer_idx]
+    attn_metadata, kv_caches = get_attention_context()
+    
+    if attn_metadata is None or kv_caches is None:
+        return _standard_attention(query, key, value, attention_mask, scaling)
+    
+    if scaling is not None:
+        self_attn.impl.scale = float(scaling)
+    
+    # Reshape: [batch, heads, seq, head_dim] -> [seq, heads * head_dim]
+    hidden = query.shape[-2]
+    query, key, value = (x.transpose(1, 2) for x in (query, key, value))
+    query, key, value = (x.reshape(hidden, -1) for x in (query, key, value))
+    
+    kv_cache = kv_caches[layer_idx] if layer_idx < len(kv_caches) else None
+    output = self_attn.forward(query, key, value, kv_cache, attn_metadata)
+    
+    return output, None
+
+
+def _standard_attention(query, key, value, attention_mask, scaling):
+    """Standard scaled dot-product attention fallback."""
+    attn_weights = torch.matmul(query, key.transpose(-2, -1))
+    if scaling is not None:
+        attn_weights = attn_weights * scaling
+    if attention_mask is not None:
+        attn_weights = attn_weights + attention_mask
+    attn_weights = torch.nn.functional.softmax(attn_weights, dim=-1)
+    attn_output = torch.matmul(attn_weights, value)
+    return attn_output, None
+
+
+# Register vLLM attention to transformers
+_vllm_attention_registered = False
+try:
+    from transformers.modeling_utils import ALL_ATTENTION_FUNCTIONS
+    ALL_ATTENTION_FUNCTIONS["vllm"] = vllm_flash_attention_forward
+    _vllm_attention_registered = True
+    logger.info("Registered vLLM attention function to transformers")
+except (ImportError, AttributeError) as e:
+    logger.warning("Could not register vLLM attention: %s", e)
+
+
+# ============================================================================
+# Base Class with Pipeline Parallel and Tensor Parallel Support
+# ============================================================================
+
+class Base(nn.Module):
+    """
+    Base class for Transformers backend models with full parallel support.
+    
+    Features:
+    - Pipeline Parallel: PPMissingLayer for distributed layers
+    - Tensor Parallel: tp_plan based module replacement
+    - Meta device initialization
+    - Module replacement (Linear → vLLM Linear, RMSNorm → vLLM RMSNorm)
+    - VocabParallelEmbedding for input embeddings
+    - Attention instances for KV cache allocation
+    """
+    
+    # For vLLM's weight loader
+    embedding_modules = ["embed_tokens"]
+    
+    # Weight name mapping following latest vLLM pattern
+    hf_to_vllm_mapper = WeightsMapper(
+        orig_to_new_prefix={
+            # Add `model.` prefix for base model checkpoints,
+            # handling the case where it is already present
+            "": "model.",
+            "model.model.": "model.",
+            # Heads will be adjacent to `model` (pooling included because of adapters)
+            "model.lm_head.": "lm_head.",
+            "model.score.": "classifier.",
+            "model.classifier.": "classifier.",
+        }
+    )
+    
+    # Note: __init_subclass__ with WeightsMapper merging is not supported in v0.6.2
+    # because WeightsMapper doesn't implement __or__/__ior__ operators.
+    # Each Mixin should define its own hf_to_vllm_mapper if needed.
+    
+    def __init__(self, *, vllm_config: VllmConfig, prefix: str = "") -> None:
+        super().__init__()
+        
+        logger.info("Using Transformers modeling backend.")
+        
+        # Store configuration
+        self.config = vllm_config.model_config.hf_config
+        self.text_config = getattr(self.config, "text_config", self.config)
+        self.model_config = vllm_config.model_config
+        self.cache_config = vllm_config.cache_config
+        self.device_config = vllm_config.device_config
+        self.parallel_config = vllm_config.parallel_config
+        self.quant_config = vllm_config.quant_config
+        self.prefix = prefix
+        
+        # Parallel groups
+        self.pp_group = get_pp_group()
+        self.tp_group = get_tp_group()
+        
+        # Model dimensions
+        self.hidden_size = getattr(self.text_config, "hidden_size", 4096)
+        self.vocab_size = getattr(self.text_config, "vocab_size", 32000)
+        
+        # Weight loading configuration
+        self.skip_prefixes: List[str] = []
+        self.ignore_unexpected_prefixes: List[str] = []
+        
+        # Configure attention backend
+        self._configure_attention_backend()
+        
+        # Create model on meta device
+        self._init_model_on_meta()
+        
+        # Apply pipeline parallel
+        self._apply_pipeline_parallel()
+        
+        # Replace modules (with tensor parallel support)
+        self._replace_modules()
+        
+        # Fix attention head_dim in case config was incorrect
+        self._fix_attention_head_dim()
+        
+        # Add debug hook to first attention module to capture tensor shapes
+        self._add_attention_debug_hook()
+        
+        # Replace input embeddings
+        self._replace_input_embeddings()
+        
+        # Create attention instances
+        self.attention_instances = self._create_attention_instances()
+        
+        # Initialize parameters on target device
+        self._init_parameters()
+        
+        # Pipeline parallel intermediate tensors
+        self.make_empty_intermediate_tensors = make_empty_intermediate_tensors_factory(
+            ["hidden_states"], self.hidden_size
+        )
+    
+    def _configure_attention_backend(self) -> None:
+        """Configure vLLM attention backend."""
+        # Note: attention implementation is set in _init_model_on_meta
+        # This method is kept for potential platform-specific configuration
+        pass
+    
+    def _init_model_on_meta(self) -> None:
+        """Create model structure on meta device."""
+        from transformers import AutoModel
+        
+        logger.info("Creating model structure on meta device...")
+        
+        # Set attention implementation to vLLM's
+        self.text_config._attn_implementation = "vllm"
+        
+        # Ensure head_dim is correctly set in BOTH config and text_config
+        # Transformers models use config.head_dim to compute attention dimensions
+        # Some models may have incorrect head_dim, so we compute and set it
+        if hasattr(self.text_config, "num_attention_heads") and hasattr(self.text_config, "hidden_size"):
+            correct_head_dim = self.text_config.hidden_size // self.text_config.num_attention_heads
+            
+            # Check and fix head_dim in text_config
+            if hasattr(self.text_config, "head_dim"):
+                if self.text_config.head_dim != correct_head_dim:
+                    logger.warning(
+                        "Correcting head_dim in text_config: %d -> %d",
+                        self.text_config.head_dim, correct_head_dim
+                    )
+                    self.text_config.head_dim = correct_head_dim
+            else:
+                self.text_config.head_dim = correct_head_dim
+            
+            # Also set in self.config (which is passed to AutoModel.from_config)
+            if hasattr(self.config, "head_dim"):
+                if self.config.head_dim != correct_head_dim:
+                    logger.warning(
+                        "Correcting head_dim in config: %d -> %d",
+                        self.config.head_dim, correct_head_dim
+                    )
+                    self.config.head_dim = correct_head_dim
+            else:
+                self.config.head_dim = correct_head_dim
+            
+            # Some models also need _attn_implementation in config
+            self.config._attn_implementation = "vllm"
+        
+        with init_on_device_without_buffers("meta"):
+            self.model: "PreTrainedModel" = AutoModel.from_config(
+                self.config,
+                torch_dtype=self.model_config.dtype,
+                trust_remote_code=self.model_config.trust_remote_code,
+            )
+        
+        self.model.eval()
+        for param in self.model.parameters():
+            param.requires_grad = False
+    
+    def _apply_pipeline_parallel(self) -> None:
+        """
+        Apply pipeline parallelization plan.
+        
+        For models that don't explicitly support pp_plan, we do a best-effort
+        approach by splitting layers based on num_hidden_layers.
+        """
+        if self.pp_group.world_size <= 1:
+            return
+        
+        logger.info("Applying pipeline parallel (world_size=%d, rank=%d)",
+                   self.pp_group.world_size, self.pp_group.rank_in_group)
+        
+        num_layers = getattr(self.text_config, "num_hidden_layers",
+                            getattr(self.text_config, "num_layers", 32))
+        
+        start_layer, end_layer = get_pp_indices(
+            num_layers,
+            self.pp_group.rank_in_group,
+            self.pp_group.world_size,
+        )
+        
+        # Find and process layer modules
+        layers_module = self._find_layers_module()
+        if layers_module is not None:
+            layers = list(layers_module.children())
+            for i, layer in enumerate(layers):
+                if not (start_layer <= i < end_layer):
+                    # Replace layers not on this rank with PPMissingLayer
+                    setattr(layers_module, str(i), PPMissingLayer())
+        
+        # Handle embeddings (only on first rank)
+        if not self.pp_group.is_first_rank:
+            input_embeddings = self.model.get_input_embeddings()
+            if input_embeddings is not None:
+                # Keep a reference but mark as missing for forward
+                self._has_embeddings = False
+        else:
+            self._has_embeddings = True
+        
+        # Handle final norm and lm_head (only on last rank)  
+        if not self.pp_group.is_last_rank:
+            # Mark lm_head as missing
+            if hasattr(self.model, 'lm_head'):
+                self.model.lm_head = PPMissingLayer()
+        
+        logger.info("Pipeline parallel applied: layers %d-%d on this rank",
+                   start_layer, end_layer)
+    
+    def _find_layers_module(self) -> Optional[nn.Module]:
+        """Find the ModuleList containing transformer layers."""
+        # Common layer container names
+        layer_names = ['layers', 'h', 'blocks', 'layer', 'encoder.layer', 'decoder.layers']
+        
+        def _search_layers(module: nn.Module, prefix: str = "") -> Optional[nn.Module]:
+            for name, child in module.named_children():
+                if name in ['layers', 'h', 'blocks', 'layer'] and isinstance(child, nn.ModuleList):
+                    return child
+                # Recursively search in model backbone
+                if name in ['model', 'transformer', 'encoder', 'decoder']:
+                    result = _search_layers(child, f"{prefix}.{name}" if prefix else name)
+                    if result is not None:
+                        return result
+            return None
+        
+        return _search_layers(self.model)
+    
+    def _get_tp_plan(self) -> Dict[str, str]:
+        """
+        Get tensor parallel plan for module replacement.
+        
+        This maps module name patterns to parallelization styles:
+        - "colwise": Column parallel (split output dim)
+        - "rowwise": Row parallel (split input dim)
+        - "replicate": Replicated (no split)
+        
+        Returns a dict mapping regex patterns to styles.
+        """
+        # Check if model has explicit tp_plan
+        if hasattr(self.model, 'tp_plan') and self.model.tp_plan:
+            return {maybe_prefix("model", k): v for k, v in self.model.tp_plan.items()}
+        
+        # Default tp_plan for common LLM architectures
+        # Based on typical transformer structure
+        return {
+            r".*\.q_proj$": "colwise",
+            r".*\.k_proj$": "colwise", 
+            r".*\.v_proj$": "colwise",
+            r".*\.o_proj$": "rowwise",
+            r".*\.gate_proj$": "colwise",
+            r".*\.up_proj$": "colwise",
+            r".*\.down_proj$": "rowwise",
+            r".*\.query$": "colwise",
+            r".*\.key$": "colwise",
+            r".*\.value$": "colwise",
+            r".*\.dense$": "rowwise",
+            r".*\.fc1$": "colwise",
+            r".*\.fc2$": "rowwise",
+        }
+    
+    def _replace_modules(self) -> None:
+        """
+        Replace modules with vLLM optimized versions.
+        
+        Uses tp_plan for tensor parallel style selection.
+        Note: lm_head is NOT replaced here - it's created at wrapper level by CausalMixin.
+        """
+        logger.info("Replacing modules with vLLM optimized versions...")
+        replaced_count = 0
+        
+        # Get tensor parallel plan
+        tp_plan = self._get_tp_plan() if self.tp_group.world_size > 1 else {}
+        
+        # Modules to skip replacement (handled at wrapper level)
+        skip_modules = {"lm_head", "score", "classifier"}
+        
+        def _recursive_replace(module: nn.Module, prefix: str = ""):
+            nonlocal replaced_count
+            
+            for name, child in list(module.named_children()):
+                # Skip PPMissingLayer
+                if isinstance(child, PPMissingLayer):
+                    continue
+                
+                # Skip modules that are handled at wrapper level
+                if name in skip_modules:
+                    logger.debug("Skipping %s (handled at wrapper level)", name)
+                    continue
+                    
+                qual_name = maybe_prefix(prefix, name)
+                new_module = None
+                
+                if isinstance(child, nn.Linear):
+                    # Determine parallelization style from tp_plan
+                    style = "replicate"
+                    for pattern, plan_style in tp_plan.items():
+                        if re.match(pattern, qual_name):
+                            style = plan_style
+                            break
+                    
+                    new_module = replace_linear_class(
+                        child,
+                        style=style,
+                        quant_config=self.quant_config,
+                        prefix=qual_name,
+                    )
+                    replaced_count += 1
+                    
+                elif child.__class__.__name__.endswith("RMSNorm") and \
+                        not isinstance(child, RMSNorm):
+                    new_module = replace_rms_norm_class(child, self.hidden_size)
+                    replaced_count += 1
+                
+                if new_module is not None:
+                    setattr(module, name, new_module)
+                    log_replacement(qual_name, child, new_module)
+                else:
+                    _recursive_replace(child, qual_name)
+        
+        _recursive_replace(self.model, "model")
+        logger.info("Replaced %d modules", replaced_count)
+    
+    def _add_attention_debug_hook(self) -> None:
+        """No-op. Debug hooks removed after root cause identified."""
+        pass
+    
+    def _fix_attention_head_dim(self) -> None:
+        """
+        Fix head_dim in attention modules and rotary embeddings after model creation.
+        
+        Some models may have incorrect head_dim in config, which causes
+        Transformers attention modules and RoPE to use wrong dimensions.
+        This method corrects head_dim in all attention modules and recreates
+        rotary embeddings if needed.
+        """
+        correct_head_dim = self.hidden_size // getattr(
+            self.text_config, "num_attention_heads", 32
+        )
+        
+        fixed_count = 0
+        
+        for name, module in self.model.named_modules():
+            module_name = module.__class__.__name__
+            
+            # Fix head_dim in Attention modules
+            if "Attention" in module_name:
+                if hasattr(module, "head_dim"):
+                    if module.head_dim != correct_head_dim:
+                        logger.warning(
+                            "Fixing head_dim in %s: %d -> %d",
+                            name, module.head_dim, correct_head_dim
+                        )
+                        module.head_dim = correct_head_dim
+                        fixed_count += 1
+            
+            # Fix rotary embeddings - recreate inv_freq buffer if needed
+            if "RotaryEmbedding" in module_name:
+                if hasattr(module, "inv_freq"):
+                    current_dim = module.inv_freq.shape[0] * 2
+                    if current_dim != correct_head_dim:
+                        logger.warning(
+                            "Recreating rotary embedding %s: dim %d -> %d",
+                            name, current_dim, correct_head_dim
+                        )
+                        base = getattr(module.config, 'rope_theta', 10000.0)
+                        if hasattr(module.config, 'rope_parameters'):
+                            base = module.config.rope_parameters.get('rope_theta', base)
+                        device = module.inv_freq.device
+                        inv_freq = 1.0 / (
+                            base ** (
+                                torch.arange(0, correct_head_dim, 2, dtype=torch.int64)
+                                .to(device=device, dtype=torch.float) / correct_head_dim
+                            )
+                        )
+                        module.register_buffer("inv_freq", inv_freq, persistent=False)
+                        if hasattr(module, "original_inv_freq"):
+                            module.register_buffer("original_inv_freq", inv_freq.clone(), persistent=False)
+        
+        if fixed_count > 0:
+            logger.info("Fixed head_dim in %d attention modules", fixed_count)
+    
+    def _replace_input_embeddings(self) -> None:
+        """Replace input embeddings with VocabParallelEmbedding."""
+        input_embeddings = self.model.get_input_embeddings()
+        if input_embeddings is None or isinstance(input_embeddings, PPMissingLayer):
+            return
+        
+        if hasattr(input_embeddings, "embedding_dim"):
+            embedding_dim = input_embeddings.embedding_dim
+        elif hasattr(input_embeddings, "weight"):
+            embedding_dim = input_embeddings.weight.shape[1]
+        else:
+            embedding_dim = self.hidden_size
+        
+        self.embed_scale = getattr(input_embeddings, "embed_scale", None)
+        
+        logger.info("Replacing input embeddings (vocab=%d, dim=%d)",
+                   self.vocab_size, embedding_dim)
+        
+        new_embeddings = VocabParallelEmbedding(
+            self.vocab_size,
+            embedding_dim,
+            org_num_embeddings=self.vocab_size,
+            quant_config=self.quant_config,
+        )
+        self.model.set_input_embeddings(new_embeddings)
+    
+    def _create_attention_instances(self) -> Dict[int, Attention]:
+        """Create Attention instances for KV cache allocation."""
+        num_layers = getattr(self.text_config, "num_hidden_layers",
+                            getattr(self.text_config, "num_layers", 32))
+        num_heads = getattr(self.text_config, "num_attention_heads", 32)
+        head_size = self.hidden_size // num_heads
+        num_kv_heads = getattr(self.text_config, "num_key_value_heads", num_heads)
+        
+        # Get PP layer range
+        pp_rank = self.pp_group.rank_in_group
+        pp_size = self.pp_group.world_size
+        start_layer, end_layer = get_pp_indices(num_layers, pp_rank, pp_size)
+        
+        logger.info("Creating attention instances for layers %d-%d "
+                   "(heads=%d, head_size=%d, kv_heads=%d)",
+                   start_layer, end_layer, num_heads, head_size, num_kv_heads)
+        
+        attention_instances: Dict[int, Attention] = {}
+        for layer_idx in range(start_layer, end_layer):
+            per_layer_sliding_window = None
+            if hasattr(self.config, "layer_types"):
+                layer_types = self.config.layer_types
+                if layer_idx < len(layer_types) and layer_types[layer_idx] == "sliding_attention":
+                    per_layer_sliding_window = getattr(self.config, "sliding_window", None)
+            
+            attention = Attention(
+                num_heads=num_heads,
+                head_size=head_size,
+                scale=1.0 / (head_size ** 0.5),
+                num_kv_heads=num_kv_heads,
+                cache_config=self.cache_config,
+                quant_config=self.quant_config,
+                prefix=f"model.layers.{layer_idx}.self_attn",
+            )
+            attention_instances[layer_idx] = attention
+        
+        return attention_instances
+    
+    def _init_parameters(self) -> None:
+        """Initialize parameters from meta device to target device."""
+        device = self.device_config.device
+        if device is None:
+            device = torch.device("cpu")
+        dtype = self.model_config.dtype
+        
+        def _init_params(module: nn.Module):
+            if isinstance(module, PPMissingLayer):
+                return
+            for name, param in list(module.named_parameters(recurse=False)):
+                if param.device == torch.device("meta"):
+                    new_param = nn.Parameter(
+                        torch.empty_like(param.data, dtype=dtype, device=device),
+                        requires_grad=False,
+                    )
+                    setattr(module, name, new_param)
+            for child in module.children():
+                _init_params(child)
+        
+        _init_params(self.model)
+        logger.info("Parameters initialized on %s", device)
+    
+    def embed_input_ids(self, input_ids: torch.Tensor) -> torch.Tensor:
+        """Get embeddings for input IDs."""
+        inputs_embeds = self.model.get_input_embeddings()(input_ids)
+        if self.embed_scale is not None:
+            inputs_embeds = inputs_embeds * self.embed_scale
+        return inputs_embeds
+    
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        kv_caches: List[torch.Tensor],
+        attn_metadata: "AttentionMetadata",
+        intermediate_tensors: Optional[IntermediateTensors] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        **kwargs,
+    ) -> torch.Tensor:
+        """Forward pass with pipeline parallel support."""
+        # Handle intermediate tensors for PP
+        if not self.pp_group.is_first_rank:
+            assert intermediate_tensors is not None
+            input_ids = None
+            inputs_embeds = intermediate_tensors["hidden_states"]
+        
+        set_attention_context(attn_metadata, kv_caches)
+        
+        try:
+            # Prepare inputs
+            if inputs_embeds is not None:
+                if inputs_embeds.dim() == 2:
+                    inputs_embeds = inputs_embeds.unsqueeze(0)
+                model_inputs = {"inputs_embeds": inputs_embeds}
+            else:
+                if input_ids is not None and input_ids.dim() == 1:
+                    input_ids = input_ids.unsqueeze(0)
+                model_inputs = {"input_ids": input_ids}
+            
+            if positions is not None:
+                if positions.dim() == 1:
+                    positions = positions.unsqueeze(0)
+                model_inputs["position_ids"] = positions
+            
+            # Apply embed_scale if needed
+            if (
+                self.embed_scale is not None
+                and input_ids is not None
+                and inputs_embeds is None
+            ):
+                inputs_embeds = self.embed_input_ids(model_inputs["input_ids"])
+                model_inputs = {"inputs_embeds": inputs_embeds}
+                if positions is not None:
+                    model_inputs["position_ids"] = positions
+            
+            # Forward through model
+            # Note: return_dict=False returns tuple, first element is last hidden state
+            with torch.no_grad():
+                outputs = self.model(
+                    **model_inputs,
+                    use_cache=False,
+                    return_dict=False,
+                    attention_instances=self.attention_instances,
+                )
+            
+            # Get hidden states from model output
+            # For models using return_dict=False, outputs is a tuple
+            # outputs[0] is usually the last hidden state
+            if isinstance(outputs, tuple):
+                hidden_states = outputs[0]
+            else:
+                hidden_states = outputs
+            
+            # Remove batch dimension
+            if hidden_states.dim() == 3 and hidden_states.size(0) == 1:
+                hidden_states = hidden_states.squeeze(0)
+            
+            # Return intermediate tensors for PP
+            if not self.pp_group.is_last_rank:
+                return IntermediateTensors({"hidden_states": hidden_states})
+            
+            return hidden_states
+            
+        finally:
+            clear_attention_context()
+    
+    def load_weights(
+        self,
+        weights: Iterable[Tuple[str, torch.Tensor]],
+    ) -> Set[str]:
+        """Load weights using AutoWeightsLoader with name mapping."""
+        loader = AutoWeightsLoader(
+            self,
+            skip_prefixes=self.skip_prefixes,
+            ignore_unexpected_prefixes=self.ignore_unexpected_prefixes,
+        )
+        loaded = loader.load_weights(weights, mapper=self.hf_to_vllm_mapper)
+        logger.info("Loaded %d weight tensors", len(loaded))
+        return set(loaded)

vllm-v0.6.2/vllm/model_executor/models/transformers/causal.py

@@ -0,0 +1,142 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright 2024 The vLLM team.
+"""Transformers modeling backend mixin for causal language models.
+
+This module provides CausalMixin that adds causal language model specific
+functionality (lm_head, compute_logits, sample) to the Base class.
+
+Following latest vLLM architecture:
+- TransformersForCausalLM = CausalMixin + Base
+- lm_head is created at the wrapper level (not inside self.model)
+"""
+
+from typing import TYPE_CHECKING, Optional
+
+import torch
+
+from vllm.logger import init_logger
+from vllm.model_executor.layers.logits_processor import LogitsProcessor
+from vllm.model_executor.layers.sampler import Sampler, SamplerOutput
+from vllm.model_executor.layers.vocab_parallel_embedding import ParallelLMHead
+from vllm.model_executor.models.utils import PPMissingLayer, maybe_prefix
+from vllm.model_executor.sampling_metadata import SamplingMetadata
+
+if TYPE_CHECKING:
+    from vllm.config import VllmConfig
+
+logger = init_logger(__name__)
+
+
+class CausalMixin:
+    """
+    Mixin class that adds causal language model functionality.
+    
+    This mixin provides:
+    - ParallelLMHead for language model head (created at wrapper level)
+    - LogitsProcessor for logits computation
+    - Sampler for token sampling
+    - compute_logits method for VllmModelForTextGeneration protocol
+    - sample method for VllmModelForTextGeneration protocol
+    
+    Following latest vLLM architecture:
+    - lm_head is a direct attribute of TransformersForCausalLM (not inside self.model)
+    - hf_to_vllm_mapper maps "model.lm_head." -> "lm_head." to handle this
+    - For tied embeddings, lm_head weight loading is skipped and weights are tied
+    
+    Should be used with Base class:
+        class TransformersForCausalLM(CausalMixin, Base): ...
+    """
+    
+    def __init__(self, *, vllm_config: "VllmConfig", prefix: str = "") -> None:
+        # Call next class in MRO (should be Base)
+        super().__init__(vllm_config=vllm_config, prefix=prefix)
+        
+        # Handle tied word embeddings - skip loading lm_head weights
+        tie_word_embeddings = getattr(self.text_config, "tie_word_embeddings", False)
+        if tie_word_embeddings:
+            self.skip_prefixes.append("lm_head.")
+            logger.info("Model has tied word embeddings, will tie lm_head weights")
+        
+        # Create lm_head at wrapper level (following latest vLLM architecture)
+        # This is outside self.model, so weights map "model.lm_head." -> "lm_head."
+        if self.pp_group.is_last_rank:
+            self.lm_head = ParallelLMHead(
+                self.vocab_size,
+                self.hidden_size,
+                quant_config=self.quant_config,
+                prefix=maybe_prefix(prefix, "lm_head"),
+            )
+            
+            # Tie weights if needed
+            if tie_word_embeddings:
+                input_embeddings = self.model.get_input_embeddings()
+                if input_embeddings is not None:
+                    self.lm_head = self.lm_head.tie_weights(input_embeddings)
+                    logger.info("Tied lm_head weights with input embeddings")
+            
+            # Setup logits processor
+            logit_scale = getattr(self.text_config, "logit_scale", 1.0)
+            self.logits_processor = LogitsProcessor(
+                self.vocab_size,
+                logits_as_input=False,
+                scale=logit_scale,
+            )
+            
+            logger.info("CausalMixin initialized (vocab_size=%d, hidden_size=%d, logit_scale=%s)",
+                       self.vocab_size, self.hidden_size, logit_scale)
+        else:
+            # For non-last PP ranks, use PPMissingLayer
+            self.lm_head = PPMissingLayer()
+            self.logits_processor = None
+            logger.info("CausalMixin initialized (PP non-last rank, using PPMissingLayer)")
+        
+        # Setup sampler
+        self.sampler = Sampler()
+    
+    def compute_logits(
+        self,
+        hidden_states: torch.Tensor,
+        sampling_metadata: SamplingMetadata,
+    ) -> Optional[torch.Tensor]:
+        """
+        Compute logits from hidden states.
+        
+        This method conforms to the VllmModelForTextGeneration protocol.
+        
+        Args:
+            hidden_states: Hidden states from the model [seq_len, hidden_size]
+            sampling_metadata: Sampling metadata
+        
+        Returns:
+            Logits tensor or None
+        """
+        if self.logits_processor is None:
+            # Non-last PP rank
+            return None
+        
+        # In v0.6.2, LogitsProcessor handles the lm_head projection internally
+        # via lm_head.linear_method.apply(). Pass lm_head as the first arg.
+        logits = self.logits_processor(self.lm_head, hidden_states,
+                                       sampling_metadata)
+        
+        return logits
+    
+    def sample(
+        self,
+        logits: torch.Tensor,
+        sampling_metadata: SamplingMetadata,
+    ) -> Optional[SamplerOutput]:
+        """
+        Sample tokens from logits.
+        
+        This method conforms to the VllmModelForTextGeneration protocol.
+        
+        Args:
+            logits: Logits tensor
+            sampling_metadata: Sampling metadata
+        
+        Returns:
+            SamplerOutput with sampled tokens
+        """
+        next_tokens = self.sampler(logits, sampling_metadata)
+        return next_tokens

vllm-v0.6.2/vllm/model_executor/models/transformers/legacy.py

@@ -0,0 +1,118 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright 2024 The vLLM team.
+"""Transformers modeling backend mixin for legacy models.
+
+This module provides LegacyMixin for BERT-like encoder models that have
+different weight naming conventions and special position handling.
+
+Following latest vLLM architecture patterns adapted for v0.6.2.
+"""
+
+from typing import TYPE_CHECKING, List, Optional
+
+import torch
+
+from vllm.logger import init_logger
+from vllm.model_executor.models.utils import WeightsMapper
+from vllm.sequence import IntermediateTensors
+
+if TYPE_CHECKING:
+    from vllm.config import VllmConfig
+
+logger = init_logger(__name__)
+
+
+class LegacyMixin:
+    """
+    Mixin class for legacy/encoder models like BERT, RoBERTa.
+    
+    This mixin provides:
+    - Weight name mapping for legacy suffix conventions (.gamma/.beta)
+    - Prefix mapping for BERT-like model structures
+    - RoBERTa-specific position handling
+    - Skip prefixes for unsupported output layers
+    
+    Should be used with Base class:
+        class TransformersForLegacy(LegacyMixin, Base): ...
+    """
+    
+    # Weight name mapping for legacy models
+    hf_to_vllm_mapper = WeightsMapper(
+        # These are applied in order, so the order matters!
+        orig_to_new_prefix={
+            # Handle BERT-like models
+            "roberta": "model",
+            "bert": "model",
+        },
+        orig_to_new_suffix={
+            # Replace legacy suffixes used for norms
+            ".gamma": ".weight",
+            ".beta": ".bias",
+        },
+    )
+    
+    def __init__(self, *, vllm_config: "VllmConfig", prefix: str = "") -> None:
+        # Call next class in MRO (should be Base)
+        super().__init__(vllm_config=vllm_config, prefix=prefix)
+        
+        # Skip unsupported/unwanted output embeddings layers
+        self.skip_prefixes.extend([
+            "model.lm_head.",
+            "model.predictions.",
+            "model.qa_outputs.",
+            "model.embeddings_project.",
+            "model.discriminator_predictions.",
+        ])
+        
+        # v0.6.2 doesn't have skip_substrs, so we handle it differently
+        # Store patterns to skip during weight loading
+        self._legacy_skip_patterns: List[str] = [
+            "position_ids",  # Some encoder models have position_ids buffer
+            "score.bias",    # Final classifier bias not used by vLLM
+        ]
+        
+        # RoBERTa-like models have extra padding in positions
+        model_type = getattr(self.text_config, "model_type", "").lower()
+        self.is_roberta = "roberta" in model_type
+        self.padding_idx = getattr(self.text_config, "pad_token_id", 1)
+        
+        if self.is_roberta:
+            logger.info("LegacyMixin detected RoBERTa model, enabling position padding")
+        
+        logger.info("LegacyMixin initialized for legacy/encoder model")
+    
+    def _should_skip_weight(self, name: str) -> bool:
+        """Check if a weight should be skipped during loading."""
+        for pattern in self._legacy_skip_patterns:
+            if pattern in name:
+                return True
+        return False
+    
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor],
+        positions: torch.Tensor,
+        kv_caches: List[torch.Tensor],
+        attn_metadata,
+        intermediate_tensors: Optional[IntermediateTensors] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        **kwargs,
+    ) -> torch.Tensor:
+        """
+        Forward pass with RoBERTa position handling.
+        
+        RoBERTa models require positions to be offset by padding_idx + 1.
+        """
+        if self.is_roberta and positions is not None:
+            # RoBERTa-specific positions padding
+            positions = positions + self.padding_idx + 1
+        
+        return super().forward(
+            input_ids=input_ids,
+            positions=positions,
+            kv_caches=kv_caches,
+            attn_metadata=attn_metadata,
+            intermediate_tensors=intermediate_tensors,
+            inputs_embeds=inputs_embeds,
+            **kwargs,
+        )

vllm-v0.6.2/vllm/model_executor/models/transformers/pooling.py

@@ -0,0 +1,170 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright 2024 The vLLM team.
+"""Transformers modeling backend mixins for pooling/embedding models.
+
+This module provides mixins for embedding and sequence classification models:
+- EmbeddingMixin: For embedding/sentence similarity models
+- SequenceClassificationMixin: For sequence classification/cross-encoding
+
+Following latest vLLM architecture patterns adapted for v0.6.2.
+"""
+
+from typing import TYPE_CHECKING, List, Optional
+
+import torch
+import torch.nn as nn
+
+from vllm.logger import init_logger
+from vllm.model_executor.layers.pooler import Pooler, PoolingType
+from vllm.model_executor.pooling_metadata import PoolingMetadata
+from vllm.sequence import PoolerOutput
+
+if TYPE_CHECKING:
+    from vllm.config import VllmConfig
+
+logger = init_logger(__name__)
+
+
+class EmbeddingMixin:
+    """
+    Mixin class that adds embedding/pooling functionality.
+    
+    This mixin provides:
+    - Pooler layer for extracting embeddings
+    - pooling method for VllmModelForPooling protocol
+    
+    Should be used with Base class:
+        class TransformersForEmbedding(EmbeddingMixin, Base): ...
+    """
+    
+    # Default pooling configuration
+    default_pooling_type: PoolingType = PoolingType.CLS
+    default_normalize: bool = True
+    default_softmax: bool = False
+    
+    def __init__(self, *, vllm_config: "VllmConfig", prefix: str = "") -> None:
+        # Call next class in MRO (should be Base)
+        super().__init__(vllm_config=vllm_config, prefix=prefix)
+        
+        # Get pooler config from model config
+        pooler_config = vllm_config.model_config.pooler_config
+        
+        # Setup pooler
+        self.pooler = Pooler.from_config_with_defaults(
+            pooler_config=pooler_config,
+            pooling_type=self.default_pooling_type,
+            normalize=self.default_normalize,
+            softmax=self.default_softmax,
+        )
+        
+        if self.pooler is None:
+            # Create default pooler if config doesn't specify
+            self.pooler = Pooler(
+                pooling_type=self.default_pooling_type,
+                normalize=self.default_normalize,
+                softmax=self.default_softmax,
+            )
+        
+        logger.info("EmbeddingMixin initialized (pooling_type=%s, normalize=%s)",
+                   self.pooler.pooling_type.name, self.pooler.normalize)
+    
+    def pooling(
+        self,
+        hidden_states: torch.Tensor,
+        pooling_metadata: PoolingMetadata,
+    ) -> Optional[PoolerOutput]:
+        """
+        Apply pooling to hidden states.
+        
+        Args:
+            hidden_states: Hidden states from the model [seq_len, hidden_size]
+            pooling_metadata: Pooling metadata
+        
+        Returns:
+            PoolerOutput with pooled embeddings
+        """
+        return self.pooler(hidden_states, pooling_metadata)
+
+
+class SequenceClassificationMixin(EmbeddingMixin):
+    """
+    Mixin class that adds sequence classification functionality.
+    
+    This mixin provides:
+    - Classifier layer for sequence classification
+    - pooling method with classification logits
+    
+    Should be used with Base class:
+        class TransformersForSequenceClassification(SequenceClassificationMixin, Base): ...
+    """
+    
+    default_pooling_type: PoolingType = PoolingType.CLS
+    default_normalize: bool = False
+    default_softmax: bool = True
+    
+    def __init__(self, *, vllm_config: "VllmConfig", prefix: str = "") -> None:
+        # Call EmbeddingMixin.__init__ -> Base.__init__
+        super().__init__(vllm_config=vllm_config, prefix=prefix)
+        
+        # Find and setup classifier layer
+        self.classifier = self._find_classifier()
+        
+        if self.classifier is not None:
+            # Initialize classifier parameters on device
+            self._init_classifier_params()
+            logger.info("SequenceClassificationMixin initialized with classifier")
+        else:
+            logger.warning("Could not find classifier layer")
+    
+    def _find_classifier(self) -> Optional[nn.Module]:
+        """Find the classifier layer in the model."""
+        # Common classifier layer names
+        classifier_names = ['classifier', 'score', 'fc', 'head']
+        
+        for name in classifier_names:
+            if hasattr(self.model, name):
+                return getattr(self.model, name)
+        
+        return None
+    
+    def _init_classifier_params(self) -> None:
+        """Initialize classifier parameters on target device."""
+        device = self.device_config.device
+        if device is None:
+            device = torch.device("cpu")
+        dtype = self.model_config.dtype
+        
+        for name, param in list(self.classifier.named_parameters()):
+            if param.device == torch.device("meta"):
+                new_param = nn.Parameter(
+                    torch.empty_like(param.data, dtype=dtype, device=device),
+                    requires_grad=False,
+                )
+                setattr(self.classifier, name.split('.')[-1], new_param)
+    
+    def pooling(
+        self,
+        hidden_states: torch.Tensor,
+        pooling_metadata: PoolingMetadata,
+    ) -> Optional[PoolerOutput]:
+        """
+        Apply pooling and classification to hidden states.
+        
+        Args:
+            hidden_states: Hidden states from the model [seq_len, hidden_size]
+            pooling_metadata: Pooling metadata
+        
+        Returns:
+            PoolerOutput with classification logits
+        """
+        # First apply base pooling
+        pooled = self.pooler(hidden_states, pooling_metadata)
+        
+        # Apply classifier if available
+        if self.classifier is not None and pooled is not None:
+            # Apply classifier to each pooled output
+            for i, output in enumerate(pooled.outputs):
+                if hasattr(output, 'data'):
+                    output.data = self.classifier(output.data)
+        
+        return pooled

vllm-v0.6.2/vllm/model_executor/models/transformers/utils.py

@@ -0,0 +1,247 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright 2024 The vLLM team.
+"""Transformers modeling backend utilities for v0.6.2.
+
+This module provides utility functions for the Transformers backend,
+including context managers for meta device initialization and
+module replacement functions.
+"""
+
+from contextlib import contextmanager
+from typing import TYPE_CHECKING, Literal, Optional, Tuple, Union
+
+import torch
+import torch.nn as nn
+
+from vllm.logger import init_logger
+from vllm.model_executor.layers.layernorm import RMSNorm
+from vllm.model_executor.layers.linear import (
+    ColumnParallelLinear,
+    ReplicatedLinear,
+    RowParallelLinear,
+)
+
+if TYPE_CHECKING:
+    from vllm.model_executor.layers.quantization.base_config import (
+        QuantizationConfig,
+    )
+
+logger = init_logger(__name__)
+
+
+@contextmanager
+def init_on_device_without_buffers(device: Union[str, torch.device]):
+    """
+    A context manager under which models are initialized with all
+    parameters on the specified device. However buffers are not
+    initialized on specified device.
+
+    This is useful for creating model structure without allocating
+    GPU memory, which is essential for memory efficiency.
+
+    Args:
+        device: Device to initialize all parameters on (e.g., "meta").
+    
+    Example:
+        with init_on_device_without_buffers("meta"):
+            model = AutoModel.from_config(config)
+        # Now model is on meta device, no GPU memory allocated
+    """
+    if isinstance(device, str):
+        device = torch.device(device)
+
+    old_register_parameter = nn.Module.register_parameter
+
+    def register_empty_parameter(module, name, param):
+        old_register_parameter(module, name, param)
+        if param is not None:
+            param_cls = type(module._parameters[name])
+            kwargs = module._parameters[name].__dict__
+            kwargs["requires_grad"] = param.requires_grad
+            module._parameters[name] = param_cls(
+                module._parameters[name].to(device), **kwargs
+            )
+
+    try:
+        nn.Module.register_parameter = register_empty_parameter
+        yield
+    finally:
+        nn.Module.register_parameter = old_register_parameter
+
+
+# Linear replacement styles
+Style = Literal["colwise", "colwise_rep", "rowwise", "rowwise_rep", "replicate"]
+
+
+def replace_linear_class(
+    linear: nn.Linear,
+    style: Style = "replicate",
+    quant_config: Optional["QuantizationConfig"] = None,
+    prefix: str = "",
+) -> Union[ColumnParallelLinear, RowParallelLinear, ReplicatedLinear]:
+    """
+    Replace nn.Linear with one of vLLM's tensor parallel linear classes.
+
+    This replacement provides:
+    - Memory efficiency through proper tensor allocation
+    - Support for quantization
+    - Tensor parallel support (when using ColumnParallel/RowParallel)
+
+    Args:
+        linear: `nn.Linear` to be replaced.
+        style: Tensor parallel style of the new linear:
+            - "colwise": Column parallel (split output dim)
+            - "colwise_rep": Column parallel with gather output
+            - "rowwise": Row parallel (split input dim)
+            - "rowwise_rep": Row parallel without parallel input
+            - "replicate": Replicated (no parallelism)
+        quant_config: Quantization config for the new linear.
+        prefix: The name of the layer for weight loading.
+    
+    Returns:
+        The new vLLM linear layer.
+    """
+    if not isinstance(style, str):
+        raise ValueError(f"Unsupported parallel style type {type(style)}, expected str")
+
+    vllm_linear_cls, vllm_linear_kwargs = {
+        "colwise": (ColumnParallelLinear, {}),
+        "colwise_rep": (ColumnParallelLinear, {"gather_output": True}),
+        "rowwise": (RowParallelLinear, {}),
+        "rowwise_rep": (RowParallelLinear, {"input_is_parallel": False}),
+        "replicate": (ReplicatedLinear, {}),
+    }.get(style, (ReplicatedLinear, {}))
+
+    return vllm_linear_cls(
+        input_size=linear.in_features,
+        output_size=linear.out_features,
+        bias=linear.bias is not None,
+        quant_config=quant_config,
+        prefix=prefix,
+        return_bias=False,  # Return tensor only, not (tensor, bias) tuple
+        **vllm_linear_kwargs,
+    )
+
+
+class TransformersRMSNorm(RMSNorm):
+    """
+    vLLM RMSNorm subclass that preserves tensor dimensions.
+    
+    vLLM's RMSNorm (especially the MLU backend) flattens input to 2D
+    (e.g., [batch, seq, hidden] -> [batch*seq, hidden]), but transformers
+    expects the batch dimension to be preserved. This subclass wraps
+    the parent forward methods to save and restore the original tensor shape.
+    
+    Since this inherits from RMSNorm directly, weight loading via
+    named_parameters() works correctly (weight path stays the same).
+    """
+    
+    def forward_native(
+        self,
+        x: torch.Tensor,
+        residual: Optional[torch.Tensor] = None,
+    ):
+        orig_shape = x.shape
+        result = super().forward_native(x, residual)
+        return self._restore_shape(result, orig_shape)
+    
+    def forward_cuda(
+        self,
+        x: torch.Tensor,
+        residual: Optional[torch.Tensor] = None,
+    ):
+        orig_shape = x.shape
+        result = super().forward_cuda(x, residual)
+        return self._restore_shape(result, orig_shape)
+    
+    def forward_mlu(
+        self,
+        x: torch.Tensor,
+        residual: Optional[torch.Tensor] = None,
+    ):
+        orig_shape = x.shape
+        result = super().forward_mlu(x, residual)
+        return self._restore_shape(result, orig_shape)
+    
+    def forward_xpu(
+        self,
+        x: torch.Tensor,
+        residual: Optional[torch.Tensor] = None,
+    ):
+        orig_shape = x.shape
+        result = super().forward_xpu(x, residual)
+        return self._restore_shape(result, orig_shape)
+    
+    def forward_hpu(
+        self,
+        x: torch.Tensor,
+        residual: Optional[torch.Tensor] = None,
+    ):
+        orig_shape = x.shape
+        result = super().forward_hpu(x, residual)
+        return self._restore_shape(result, orig_shape)
+    
+    @staticmethod
+    def _restore_shape(result, orig_shape: Tuple):
+        """Restore original tensor shape if it was changed."""
+        if isinstance(result, tuple):
+            restored = []
+            for t in result:
+                if t is not None and t.shape != orig_shape:
+                    t = t.view(orig_shape)
+                restored.append(t)
+            return tuple(restored)
+        else:
+            if result.shape != orig_shape:
+                result = result.view(orig_shape)
+            return result
+
+
+def replace_rms_norm_class(
+    rms_norm: nn.Module,
+    hidden_size: int,
+) -> nn.Module:
+    """
+    Replace a Transformers RMSNorm with vLLM's optimized RMSNorm,
+    wrapped to preserve tensor dimensions.
+
+    vLLM's RMSNorm provides:
+    - Fused CUDA kernels for better performance
+    - Support for fused add + norm operations
+
+    The wrapper ensures that the original tensor shape (including batch
+    dimension) is preserved, which is required by transformers' model
+    forward methods.
+
+    Args:
+        rms_norm: The RMSNorm module to replace.
+        hidden_size: The hidden size of the model.
+    
+    Returns:
+        The new vLLM RMSNorm layer wrapped for shape preservation.
+    """
+    # Try to get epsilon from various attribute names
+    eps = getattr(rms_norm, "eps", None)
+    if eps is None:
+        eps = getattr(rms_norm, "variance_epsilon", None)
+    if eps is None:
+        eps = 1e-6
+
+    # Check if weight exists and get its size
+    weight = getattr(rms_norm, "weight", None)
+    if weight is not None:
+        hidden_size = weight.size(0)
+
+    return TransformersRMSNorm(hidden_size=hidden_size, eps=eps)
+
+
+def log_replacement(name: str, old_module: nn.Module, new_module: nn.Module):
+    """Log module replacement for debugging."""
+    logger.debug("Replaced %s: %s -> %s", name, type(old_module).__name__, type(new_module).__name__)
+
+
+def maybe_prefix(prefix: str, name: str) -> str:
+    """Combine prefix and name with a dot separator."""
+    if prefix:
+        return f"{prefix}.{name}"
+    return name

vllm-v0.6.2/vllm/model_executor/models/utils.py

@@ -492,6 +492,29 @@ def maybe_offload_to_cpu(module: torch.nn.Module) -> torch.nn.Module:
     return module
 
 
+def _get_device_memory_info() -> Tuple[Optional[float], Optional[float], Optional[float]]:
+    """Get device memory info in GiB. Returns (allocated, reserved, total) or Nones."""
+    try:
+        import torch.mlu
+        allocated = torch.mlu.memory_allocated() / (1024 ** 3)
+        reserved = torch.mlu.memory_reserved() / (1024 ** 3)
+        free, total = torch.mlu.mem_get_info()
+        total = total / (1024 ** 3)
+        return allocated, reserved, total
+    except Exception:
+        pass
+    try:
+        if torch.cuda.is_available():
+            allocated = torch.cuda.memory_allocated() / (1024 ** 3)
+            reserved = torch.cuda.memory_reserved() / (1024 ** 3)
+            free, total = torch.cuda.mem_get_info()
+            total = total / (1024 ** 3)
+            return allocated, reserved, total
+    except Exception:
+        pass
+    return None, None, None
+
+
 def make_layers(
     num_hidden_layers: int,
     layer_fn: LayerFn,
@@ -505,11 +528,31 @@ def make_layers(
     start_layer, end_layer = get_pp_indices(num_hidden_layers,
                                             get_pp_group().rank_in_group,
                                             get_pp_group().world_size)
+
+    alloc_before, _, total = _get_device_memory_info()
+    if alloc_before is not None:
+        logger.info(
+            "[DEBUG-MEM] make_layers start: allocated=%.2f GiB, "
+            "total=%.2f GiB, layers to create: %d-%d / %d",
+            alloc_before, total, start_layer, end_layer, num_hidden_layers)
+
+    created_layers = []
+    for idx in range(start_layer, end_layer):
+        layer = maybe_offload_to_cpu(layer_fn(prefix=f"{prefix}.{idx}"))
+        alloc_after, reserved, _ = _get_device_memory_info()
+        if alloc_after is not None:
+            delta = alloc_after - alloc_before
+            logger.info(
+                "[DEBUG-MEM] Layer %s.%d created: "
+                "allocated=%.2f GiB (+%.4f GiB), reserved=%.2f GiB",
+                prefix, idx, alloc_after, delta, reserved)
+            alloc_before = alloc_after
+        created_layers.append(layer)
+
     modules = torch.nn.ModuleList(
-        [PPMissingLayer() for _ in range(start_layer)] + [
-            maybe_offload_to_cpu(layer_fn(prefix=f"{prefix}.{idx}"))
-            for idx in range(start_layer, end_layer)
-        ] + [PPMissingLayer() for _ in range(end_layer, num_hidden_layers)])
+        [PPMissingLayer() for _ in range(start_layer)]
+        + created_layers
+        + [PPMissingLayer() for _ in range(end_layer, num_hidden_layers)])
     return start_layer, end_layer, modules
 
 

vllm-v0.6.2/vllm/transformers_utils/config.py

@@ -89,9 +89,10 @@ def file_or_path_exists(model: Union[str, Path], config_name, revision,
     # hf_hub. This will fail in offline mode.
     try:
         return file_exists(model, config_name, revision=revision, token=token)
-    except huggingface_hub.errors.OfflineModeIsEnabled:
-        # Don't raise in offline mode, all we know is that we don't have this
-        # file cached.
+    except (huggingface_hub.errors.OfflineModeIsEnabled,
+            huggingface_hub.errors.HFValidationError):
+        # Don't raise in offline mode or when model path fails HF validation
+        # (e.g., local paths that don't match HF repo id format)
         return False
 
 
@@ -112,7 +113,9 @@ def patch_rope_scaling_dict(rope_scaling: Dict[str, Any]) -> None:
         logger.info("Replacing legacy 'type' key with 'rope_type'")
 
     if "rope_type" not in rope_scaling:
-        raise ValueError("rope_scaling should have a 'rope_type' key")
+        rope_scaling["rope_type"] = "default"
+        logger.warning("rope_scaling missing 'rope_type' key, "
+                       "defaulting to 'default'")
 
     if rope_scaling["rope_type"] == "su":
         rope_scaling["rope_type"] = "longrope"
@@ -167,12 +170,6 @@ def get_config(
                                  token=token):
             config_format = ConfigFormat.MISTRAL
         else:
-            # If we're in offline mode and found no valid config format, then
-            # raise an offline mode error to indicate to the user that they
-            # don't have files cached and may need to go online.
-            # This is conveniently triggered by calling file_exists().
-            file_exists(model, HF_CONFIG_NAME, revision=revision, token=token)
-
             raise ValueError(f"No supported config format found in {model}")
 
     if config_format == ConfigFormat.HF:

vllm-v0.6.2/vllm/transformers_utils/dynamic_module.py

@@ -0,0 +1,76 @@
+"""
+Dynamic module loading utilities for custom HuggingFace models.
+Ported from latest vLLM to support auto_map in model config.
+"""
+import os
+from typing import Dict, Optional, Type, Union
+
+from transformers.dynamic_module_utils import (
+    get_class_from_dynamic_module,
+    resolve_trust_remote_code,
+)
+
+import vllm.envs as envs
+from vllm.logger import init_logger
+
+logger = init_logger(__name__)
+
+
+def try_get_class_from_dynamic_module(
+    class_reference: str,
+    pretrained_model_name_or_path: str,
+    trust_remote_code: bool,
+    cache_dir: Optional[Union[str, os.PathLike]] = None,
+    force_download: bool = False,
+    resume_download: Optional[bool] = None,
+    proxies: Optional[Dict[str, str]] = None,
+    token: Optional[Union[bool, str]] = None,
+    revision: Optional[str] = None,
+    local_files_only: bool = False,
+    repo_type: Optional[str] = None,
+    code_revision: Optional[str] = None,
+    warn_on_fail: bool = True,
+    **kwargs,
+) -> Optional[Type]:
+    """
+    As `transformers.dynamic_module_utils.get_class_from_dynamic_module`,
+    but ignoring any errors.
+    
+    This allows vLLM to load custom models that define their own
+    model classes via the `auto_map` field in config.json.
+    """
+    try:
+        resolve_trust_remote_code(
+            trust_remote_code,
+            pretrained_model_name_or_path,
+            has_local_code=False,
+            has_remote_code=True,
+        )
+
+        return get_class_from_dynamic_module(
+            class_reference,
+            pretrained_model_name_or_path,
+            cache_dir=cache_dir,
+            force_download=force_download,
+            resume_download=resume_download,
+            proxies=proxies,
+            token=token,
+            revision=revision,
+            local_files_only=local_files_only,
+            repo_type=repo_type,
+            code_revision=code_revision,
+            **kwargs,
+        )
+    except Exception:
+        location = "ModelScope" if envs.VLLM_USE_MODELSCOPE else "HF Hub"
+
+        if warn_on_fail:
+            logger.warning(
+                "Unable to load %s from %s on %s.",
+                class_reference,
+                pretrained_model_name_or_path,
+                location,
+                exc_info=True,
+            )
+
+        return None

vllm-v0.6.2/vllm_mlu/vllm_mlu/model_executor/models/llama.py

@@ -74,18 +74,22 @@ def vllm__module_executor__models__llama__LlamaAttention__forward(
     smooth_quant_scale: Optional[torch.Tensor] = None,
 ) -> torch.Tensor:
     qkv, _ = self.qkv_proj(hidden_states, smooth_quant_scale)
-    q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
     '''
     =============================
     Modify by vllm_mlu
     =============================
     @brief: pack q & k to fit tmo.apply_rotary
+    @optimization: avoid redundant split operation
     '''
     if self.rope_scaling is not None and self.rope_scaling["rope_type"] == "longrope":
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
         q, k = self.rotary_emb(positions, q, k)
     else:
-        qk, _ = qkv.split([self.q_size + self.kv_size, self.kv_size], dim=-1)
+        # Optimized: split qkv into [qk, v] directly, avoiding redundant split
+        qk, v = qkv.split([self.q_size + self.kv_size, self.kv_size], dim=-1)
         self.rotary_emb(positions, qk.view(-1, self.num_heads + self.num_kv_heads, self.head_dim))
+        # Split qk into q and k after rotary embedding
+        q, k = qk.split([self.q_size, self.kv_size], dim=-1)
     '''
     ==================
     End of MLU Hijack

vllm-v0.6.2/vllm_mlu/vllm_mlu/worker/cache_engine.py

@@ -24,8 +24,29 @@ def vllm__worker__cache_engine__CacheEngine___allocate_kv_cache(
     =============================
     Modify by vllm_mlu
     =============================
-    @brief: add kv_cache_scale for int8 support
-    ''' 
+    @brief: add kv_cache_scale for int8 support;
+            cap num_blocks to avoid exceeding CNNL int32 element limit
+    '''
+    # CNNL operators have a max supported tensor element count of INT32_MAX.
+    # num_blocks should already be capped by determine_num_available_blocks,
+    # this is a defensive check to catch any edge cases.
+    CNNL_MAX_TENSOR_ELEMENTS = 2**31 - 1
+    total_elements = 1
+    for dim in kv_cache_shape:
+        total_elements *= dim
+    if total_elements > CNNL_MAX_TENSOR_ELEMENTS:
+        elements_per_block = total_elements // num_blocks
+        max_num_blocks = CNNL_MAX_TENSOR_ELEMENTS // elements_per_block
+        logger.warning(
+            "KV cache tensor elements (%d) exceed CNNL max (%d). "
+            "Reducing num_blocks from %d to %d. This indicates "
+            "determine_num_available_blocks did not cap correctly.",
+            total_elements, CNNL_MAX_TENSOR_ELEMENTS,
+            num_blocks, max_num_blocks)
+        num_blocks = max_num_blocks
+        kv_cache_shape = self.attn_backend.get_kv_cache_shape(
+            num_blocks, self.block_size, self.num_kv_heads, self.head_size)
+
     kv_cache_scales_shape = self.attn_backend.get_kv_cache_scale_shape(
         num_blocks, self.block_size, self.num_kv_heads)
     pin_memory = is_pin_memory_available() if device == "cpu" else False

vllm-v0.6.2/vllm_mlu/vllm_mlu/worker/mlu_worker.py

@@ -95,6 +95,30 @@ class MLUWorker_V2(MLUWorker):
         num_gpu_blocks = max(num_gpu_blocks, 0)
         num_cpu_blocks = max(num_cpu_blocks, 0)
 
+        # Cap num_gpu_blocks to avoid exceeding CNNL's int32 tensor element
+        # limit. CNNL operators do not support tensors with more than
+        # 2^31 - 1 elements. The KV cache shape is typically
+        # (2, num_blocks, num_kv_heads, block_size, head_size), and when
+        # num_blocks is very large (e.g. for tiny models with huge free
+        # memory), the total element count can overflow.
+        CNNL_MAX_TENSOR_ELEMENTS = 2**31 - 1
+        block_size = self.cache_config.block_size
+        num_kv_heads = self.model_config.get_num_kv_heads(
+            self.parallel_config)
+        head_size = self.model_config.get_head_size()
+        # kv_cache_shape = (2, num_blocks, num_kv_heads, block_size, head_size)
+        elements_per_block = 2 * num_kv_heads * block_size * head_size
+        if elements_per_block > 0:
+            max_blocks_by_cnnl = CNNL_MAX_TENSOR_ELEMENTS // elements_per_block
+            if num_gpu_blocks > max_blocks_by_cnnl:
+                logger.warning(
+                    "Reducing num_gpu_blocks from %d to %d to stay within "
+                    "CNNL max tensor element limit (%d). "
+                    "elements_per_block=%d",
+                    num_gpu_blocks, max_blocks_by_cnnl,
+                    CNNL_MAX_TENSOR_ELEMENTS, elements_per_block)
+                num_gpu_blocks = max_blocks_by_cnnl
+
         logger.info(
             "Memory profiling results: total_gpu_memory=%.2fGiB"
             " initial_memory_usage=%.2fGiB peak_torch_memory=%.2fGiB"