Support Deepseek MoE Model (#689)

python/sglang/srt/managers/controller/cuda_graph_runner.py

 """Run the model with cuda graph."""
 
 import bisect
+from contextlib import contextmanager
 
 import torch
 from flashinfer import BatchDecodeWithPagedKVCacheWrapper
@@ -15,9 +16,10 @@ from sglang.srt.managers.controller.infer_batch import (
     InputMetadata,
     init_flashinfer_args,
 )
+from sglang.srt.utils import monkey_patch_vllm_all_gather
 
 
-def _to_torch(model: torch.nn.Module, reverse=False):
+def _to_torch(model: torch.nn.Module, reverse: bool = False):
     for sub in model._modules.values():
         if isinstance(sub, CustomOp):
             if reverse:
@@ -28,13 +30,26 @@ def _to_torch(model: torch.nn.Module, reverse=False):
             _to_torch(sub, reverse)
 
 
-def get_forward(model: torch.nn.Module, use_torch: bool):
-    if use_torch:
-        _to_torch(model, reverse=False)
-        return torch.compile(model.forward, mode="max-autotune-no-cudagraphs")
-    else:
-        _to_torch(model, reverse=True)
-        return model.forward
+@contextmanager
+def patch_model(
+    model: torch.nn.Module, use_compile: bool, tp_group: "GroupCoordinator"
+):
+    backup_ca_comm = None
+
+    try:
+        if use_compile:
+            _to_torch(model)
+            monkey_patch_vllm_all_gather()
+            backup_ca_comm = tp_group.ca_comm
+            tp_group.ca_comm = None
+            yield torch.compile(model.forward, mode="max-autotune-no-cudagraphs")
+        else:
+            yield model.forward
+    finally:
+        if use_compile:
+            _to_torch(model, reverse=True)
+            monkey_patch_vllm_all_gather(reverse=True)
+            tp_group.ca_comm = backup_ca_comm
 
 
 class CudaGraphRunner:
@@ -86,17 +101,21 @@ class CudaGraphRunner:
         with graph_capture() as graph_capture_context:
             self.stream = graph_capture_context.stream
             for bs in batch_size_list:
-                forward = get_forward(self.model_runner.model, bs in self.compile_bs)
-                (
-                    graph,
-                    input_buffers,
-                    output_buffers,
-                    flashinfer_handler,
-                ) = self.capture_one_batch_size(bs, forward)
-                self.graphs[bs] = graph
-                self.input_buffers[bs] = input_buffers
-                self.output_buffers[bs] = output_buffers
-                self.flashinfer_handlers[bs] = flashinfer_handler
+                with patch_model(
+                    self.model_runner.model,
+                    bs in self.compile_bs,
+                    self.model_runner.tp_group,
+                ) as forward:
+                    (
+                        graph,
+                        input_buffers,
+                        output_buffers,
+                        flashinfer_handler,
+                    ) = self.capture_one_batch_size(bs, forward)
+                    self.graphs[bs] = graph
+                    self.input_buffers[bs] = input_buffers
+                    self.output_buffers[bs] = output_buffers
+                    self.flashinfer_handlers[bs] = flashinfer_handler
 
     def capture_one_batch_size(self, bs, forward):
         graph = torch.cuda.CUDAGraph()

python/sglang/srt/managers/controller/model_runner.py

@@ -22,7 +22,6 @@ from vllm.distributed import (
     init_distributed_environment,
     initialize_model_parallel,
 )
-from vllm.model_executor.model_loader import get_model
 from vllm.model_executor.models import ModelRegistry
 
 from sglang.global_config import global_config
@@ -241,7 +240,9 @@ class ModelRunner:
             self.cuda_graph_runner = None
             return
 
-        logger.info(f"[gpu_id={self.gpu_id}] Capture cuda graph begin.")
+        logger.info(
+            f"[gpu_id={self.gpu_id}] Capture cuda graph begin. This can take up to several minutes."
+        )
         batch_size_list = [1, 2, 4] + [i * 8 for i in range(1, 17)]
         self.cuda_graph_runner = CudaGraphRunner(
             self,
@@ -252,7 +253,7 @@ class ModelRunner:
             self.cuda_graph_runner.capture(batch_size_list)
         except RuntimeError as e:
             raise Exception(
-                f"Capture cuda graph failed {e}. Possible solutions:\n"
+                f"Capture cuda graph failed: {e}. Possible solutions:\n"
                 f"1. disable cuda graph by --disable-cuda-graph\n"
                 f"2. set --mem-fraction-static to a smaller value\n"
                 f"Open an issue on GitHub with reproducible scripts if you need help.\n"

python/sglang/srt/models/deepseek.py

+# Adapted from:
+# https://github.com/vllm-project/vllm/blob/14f91fe67c2342f2fe859dc6a5c40810df0e1c61/vllm/model_executor/models/deepseek.py
+"""Inference-only Deepseek model."""
+from typing import Any, Dict, Iterable, Optional, Tuple
+
+import torch
+from torch import nn
+from transformers import PretrainedConfig
+from vllm.config import CacheConfig
+from vllm.distributed import (
+    get_tensor_model_parallel_rank,
+    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 fused_moe
+from vllm.model_executor.layers.layernorm import RMSNorm
+from vllm.model_executor.layers.linear import (
+    MergedColumnParallelLinear,
+    QKVParallelLinear,
+    ReplicatedLinear,
+    RowParallelLinear,
+)
+from vllm.model_executor.layers.quantization.base_config import QuantizationConfig
+from vllm.model_executor.layers.rotary_embedding import get_rope
+from vllm.model_executor.layers.vocab_parallel_embedding import (
+    ParallelLMHead,
+    VocabParallelEmbedding,
+)
+from vllm.model_executor.model_loader.weight_utils import default_weight_loader
+
+from sglang.srt.layers.logits_processor import LogitsProcessor
+from sglang.srt.layers.radix_attention import RadixAttention
+from sglang.srt.managers.controller.infer_batch import InputMetadata
+
+
+class DeepseekMLP(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 DeepseekMoE(nn.Module):
+
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        quant_config: Optional[QuantizationConfig] = None,
+    ):
+        super().__init__()
+        self.config = config
+        self.rank = get_tensor_model_parallel_rank()
+        self.tp_size = get_tensor_model_parallel_world_size()
+        self.n_routed_experts = config.n_routed_experts
+        self.top_k = config.num_experts_per_tok
+        if self.tp_size > self.n_routed_experts:
+            raise ValueError(
+                f"Tensor parallel size {self.tp_size} is greater than "
+                f"the number of experts {self.n_routed_experts}."
+            )
+
+        self.experts = nn.ModuleList(
+            [
+                DeepseekMLP(
+                    hidden_size=config.hidden_size,
+                    intermediate_size=config.moe_intermediate_size,
+                    hidden_act=config.hidden_act,
+                    quant_config=quant_config,
+                    reduce_results=False,
+                )
+                for idx in range(self.n_routed_experts)
+            ]
+        )
+        self.pack_params()
+
+        self.gate = ReplicatedLinear(
+            config.hidden_size, self.n_routed_experts, bias=False, quant_config=None
+        )
+
+        if config.n_shared_experts is not None:
+            intermediate_size = config.moe_intermediate_size * config.n_shared_experts
+            self.shared_experts = DeepseekMLP(
+                hidden_size=config.hidden_size,
+                intermediate_size=intermediate_size,
+                hidden_act=config.hidden_act,
+                quant_config=quant_config,
+                reduce_results=False,
+            )
+
+    def pack_params(self):
+        w1 = []
+        w2 = []
+        for expert in self.experts:
+            w1.append(expert.gate_up_proj.weight)
+            w2.append(expert.down_proj.weight)
+        self.w1 = torch._utils._flatten_dense_tensors(w1)
+        w1s = torch._utils._unflatten_dense_tensors(self.w1, w1)
+        for data, param in zip(w1s, w1):
+            param.data = data
+        self.w1 = self.w1.view(len(w1), *w1s[0].shape)
+
+        self.w2 = torch._utils._flatten_dense_tensors(w2)
+        w2s = torch._utils._unflatten_dense_tensors(self.w2, w2)
+        for data, param in zip(w2s, w2):
+            param.data = data
+
+        self.w2 = self.w2.view(len(w2), *w2s[0].shape)
+
+    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
+        num_tokens, hidden_dim = hidden_states.shape
+        hidden_states = hidden_states.view(-1, hidden_dim)
+        if self.config.n_shared_experts is not None:
+            shared_output = self.shared_experts(hidden_states)
+        # router_logits: (num_tokens, n_experts)
+        router_logits, _ = self.gate(hidden_states)
+        final_hidden_states = fused_moe(
+            hidden_states,
+            self.w1,
+            self.w2,
+            router_logits,
+            self.top_k,
+            renormalize=self.config.norm_topk_prob,
+            inplace=True,
+        )
+
+        if self.config.n_shared_experts is not None:
+            final_hidden_states = final_hidden_states + shared_output
+        final_hidden_states = tensor_model_parallel_all_reduce(final_hidden_states)
+
+        return final_hidden_states.view(num_tokens, hidden_dim)
+
+
+class DeepseekAttention(nn.Module):
+
+    def __init__(
+        self,
+        hidden_size: int,
+        num_heads: int,
+        num_kv_heads: int,
+        layer_id: int = 0,
+        rope_theta: float = 10000,
+        rope_scaling: Optional[Dict[str, Any]] = None,
+        max_position_embeddings: int = 8192,
+        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:
+            # Number of KV heads is greater than TP size, so we partition
+            # the KV heads across multiple tensor parallel GPUs.
+            assert self.total_num_kv_heads % tp_size == 0
+        else:
+            # Number of KV heads is less than TP size, so we replicate
+            # the KV heads across multiple tensor parallel GPUs.
+            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=False,
+            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 = RadixAttention(
+            self.num_heads,
+            self.head_dim,
+            self.scaling,
+            num_kv_heads=self.num_kv_heads,
+            layer_id=layer_id,
+        )
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        input_metadata: InputMetadata,
+    ) -> torch.Tensor:
+        qkv, _ = self.qkv_proj(hidden_states)
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
+        q, k = self.rotary_emb(positions, q, k)
+        attn_output = self.attn(q, k, v, input_metadata)
+        output, _ = self.o_proj(attn_output)
+        return output
+
+
+class DeepseekDecoderLayer(nn.Module):
+
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        layer_id: 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 = DeepseekAttention(
+            hidden_size=self.hidden_size,
+            num_heads=config.num_attention_heads,
+            num_kv_heads=config.num_key_value_heads,
+            layer_id=layer_id,
+            rope_theta=rope_theta,
+            rope_scaling=rope_scaling,
+            max_position_embeddings=max_position_embeddings,
+            cache_config=cache_config,
+            quant_config=quant_config,
+        )
+        if (
+            config.n_routed_experts is not None
+            and layer_id >= config.first_k_dense_replace
+            and layer_id % config.moe_layer_freq == 0
+        ):
+            self.mlp = DeepseekMoE(config=config, quant_config=quant_config)
+        else:
+            self.mlp = DeepseekMLP(
+                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,
+        input_metadata: InputMetadata,
+        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,
+            input_metadata=input_metadata,
+        )
+
+        # Fully Connected
+        hidden_states, residual = self.post_attention_layernorm(hidden_states, residual)
+        hidden_states = self.mlp(hidden_states)
+        return hidden_states, residual
+
+
+class DeepseekModel(nn.Module):
+
+    fall_back_to_pt_during_load = False
+
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        cache_config: Optional[CacheConfig] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.padding_idx = config.pad_token_id
+        self.vocab_size = config.vocab_size
+
+        self.embed_tokens = VocabParallelEmbedding(
+            config.vocab_size,
+            config.hidden_size,
+        )
+        self.layers = nn.ModuleList(
+            [
+                DeepseekDecoderLayer(
+                    config, layer_id, cache_config, quant_config=quant_config
+                )
+                for layer_id in range(config.num_hidden_layers)
+            ]
+        )
+        self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
+
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        input_metadata: InputMetadata,
+    ) -> torch.Tensor:
+        hidden_states = self.embed_tokens(input_ids)
+        residual = None
+        for i in range(len(self.layers)):
+            layer = self.layers[i]
+            hidden_states, residual = layer(
+                positions, hidden_states, input_metadata, residual
+            )
+        hidden_states, _ = self.norm(hidden_states, residual)
+        return hidden_states
+
+
+class DeepseekForCausalLM(nn.Module):
+
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        cache_config: Optional[CacheConfig] = None,
+        quant_config: Optional[QuantizationConfig] = None,
+    ) -> None:
+        super().__init__()
+        self.config = config
+        self.quant_config = quant_config
+        self.model = DeepseekModel(config, cache_config, quant_config)
+        self.lm_head = ParallelLMHead(
+            config.vocab_size, config.hidden_size, quant_config=quant_config
+        )
+        self.logits_processor = LogitsProcessor(config)
+
+    @torch.no_grad()
+    def forward(
+        self,
+        input_ids: torch.Tensor,
+        positions: torch.Tensor,
+        input_metadata: InputMetadata,
+    ) -> torch.Tensor:
+        hidden_states = self.model(input_ids, positions, input_metadata)
+        return self.logits_processor(
+            input_ids, hidden_states, self.lm_head.weight, input_metadata
+        )
+
+    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())
+        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:
+                if weight_name not in name:
+                    continue
+                name = name.replace(weight_name, param_name)
+                # Skip loading extra bias for GPTQ models.
+                if name.endswith(".bias") and name not in params_dict:
+                    continue
+                # Skip experts that are not assigned to this worker.
+                if (
+                    "mlp.experts." in name or "mlp.shared_experts." in name
+                ) and name not in params_dict:
+                    continue
+                param = params_dict[name]
+                weight_loader = param.weight_loader
+                weight_loader(param, loaded_weight, shard_id)
+                break
+            else:
+                # Skip loading extra bias for GPTQ models.
+                if name.endswith(".bias") and name not in params_dict:
+                    continue
+                # Skip experts that are not assigned to this worker.
+                if (
+                    "mlp.experts." in name or "mlp.shared_experts." in name
+                ) and name not in params_dict:
+                    continue
+                param = params_dict[name]
+                weight_loader = getattr(param, "weight_loader", default_weight_loader)
+                weight_loader(param, loaded_weight)
+
+
+EntryClass = DeepseekForCausalLM

python/sglang/srt/server.py

@@ -167,7 +167,7 @@ def _set_torch_compile_config():
     torch._inductor.config.fx_graph_cache = True  # Experimental feature to reduce compilation times, will be on by default in future
 
     # FIXME: tmp workaround
-    torch._dynamo.config.accumulated_cache_size_limit = 128
+    torch._dynamo.config.accumulated_cache_size_limit = 256
 
 
 def launch_server(

python/sglang/srt/utils.py

@@ -411,6 +411,52 @@ def monkey_patch_vllm_dummy_weight_loader():
     setattr(DummyModelLoader, "load_model", load_model)
 
 
+vllm_all_gather_backup = None
+
+
+def monkey_patch_vllm_all_gather(reverse: bool = False):
+    """Monkey patch all-gather to remove in-place operations."""
+    from torch.distributed import _functional_collectives as funcol
+    from vllm.distributed.parallel_state import GroupCoordinator
+
+    global vllm_all_gather_backup
+    if vllm_all_gather_backup is None:
+        vllm_all_gather_backup = GroupCoordinator.all_gather
+
+    def all_gather(self, input_: torch.Tensor, dim: int = -1) -> torch.Tensor:
+        world_size = self.world_size
+        # Bypass the function if we are using only 1 GPU.
+        if world_size == 1:
+            return input_
+        assert (
+            -input_.dim() <= dim < input_.dim()
+        ), f"Invalid dim ({dim}) for input tensor with shape {input_.size()}"
+        if dim < 0:
+            # Convert negative dim to positive.
+            dim += input_.dim()
+        input_size = input_.size()
+        # Allocate output tensor.
+        output_tensor = torch.empty(
+            (world_size,) + input_size, dtype=input_.dtype, device=input_.device
+        )
+
+        output_tensor = funcol.all_gather_tensor(
+            input_, gather_dim=0, group=self.device_group
+        ).view((world_size,) + input_size)
+
+        # Reshape
+        output_tensor = output_tensor.movedim(0, dim)
+        output_tensor = output_tensor.reshape(
+            input_size[:dim] + (world_size * input_size[dim],) + input_size[dim + 1 :]
+        )
+        return output_tensor
+
+    if reverse:
+        setattr(GroupCoordinator, "all_gather", vllm_all_gather_backup)
+    else:
+        setattr(GroupCoordinator, "all_gather", all_gather)
+
+
 API_KEY_HEADER_NAME = "X-API-Key"