sglang 0.5.2rc2__py3-none-any.whl → 0.5.3__py3-none-any.whl

sglang/srt/models/bailing_moe.py

@@ -1,377 +1,907 @@
-# Copyright 2023-2024 SGLang Team
-# Adapted from https://github.com/vllm-project/vllm/blob/main/vllm/model_executor/models/bailing_moe.py
-
-from collections.abc import Iterable
-from typing import Optional, Tuple
+# coding=utf-8
+# Copyright 2023 Antgroup 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.
+""" SGLang BailingMoE model."""
+import logging
+from typing import Any, Dict, Iterable, Optional, Tuple, Union
 
 import torch
 import torch.nn.functional as F
 from torch import nn
-from transformers.configuration_utils import PretrainedConfig
+from transformers import PretrainedConfig
 
 from sglang.srt.distributed import (
+    get_pp_group,
     get_tensor_model_parallel_world_size,
+    parallel_state,
     tensor_model_parallel_all_reduce,
 )
+from sglang.srt.eplb.expert_distribution import get_global_expert_distribution_recorder
+from sglang.srt.eplb.expert_location import ModelConfigForExpertLocation
+from sglang.srt.eplb.expert_location_dispatch import ExpertLocationDispatchInfo
 from sglang.srt.layers.activation import SiluAndMul
+from sglang.srt.layers.communicator import (
+    LayerCommunicator,
+    LayerScatterModes,
+    enable_moe_dense_fully_dp,
+)
+from sglang.srt.layers.dp_attention import (
+    get_attention_dp_size,
+    get_attention_tp_rank,
+    get_attention_tp_size,
+    is_dp_attention_enabled,
+)
 from sglang.srt.layers.layernorm import RMSNorm
 from sglang.srt.layers.linear import (
     MergedColumnParallelLinear,
     QKVParallelLinear,
-    ReplicatedLinear,
     RowParallelLinear,
 )
 from sglang.srt.layers.logits_processor import LogitsProcessor
-from sglang.srt.layers.moe.fused_moe_triton import FusedMoE
+from sglang.srt.layers.moe import get_moe_a2a_backend
+from sglang.srt.layers.moe.ep_moe.layer import get_moe_impl_class
+from sglang.srt.layers.moe.fused_moe_triton.layer import FusedMoE
+from sglang.srt.layers.moe.token_dispatcher import DeepEPDispatcher
 from sglang.srt.layers.moe.topk import TopK
+from sglang.srt.layers.moe.utils import DeepEPMode
 from sglang.srt.layers.quantization.base_config import QuantizationConfig
 from sglang.srt.layers.radix_attention import RadixAttention
 from sglang.srt.layers.rotary_embedding import get_rope
+from sglang.srt.layers.utils import PPMissingLayer
 from sglang.srt.layers.vocab_parallel_embedding import (
     ParallelLMHead,
     VocabParallelEmbedding,
 )
-from sglang.srt.model_executor.forward_batch_info import ForwardBatch
+from sglang.srt.managers.schedule_batch import global_server_args_dict
+from sglang.srt.model_executor.cuda_graph_runner import get_is_capture_mode
+from sglang.srt.model_executor.forward_batch_info import ForwardBatch, PPProxyTensors
 from sglang.srt.model_loader.weight_utils import default_weight_loader
-from sglang.srt.utils import add_prefix, make_layers
+from sglang.srt.models.utils import (
+    create_fused_set_kv_buffer_arg,
+    enable_fused_set_kv_buffer,
+)
+from sglang.srt.utils import add_prefix, is_cuda, is_non_idle_and_non_empty, make_layers
 
+LoraConfig = None
+logger = logging.getLogger(__name__)
+_is_cuda = is_cuda()
 
-class BailingAttention(nn.Module):
 
+class BailingMoEMLP(nn.Module):
     def __init__(
         self,
+        intermediate_size: int,
         config: PretrainedConfig,
-        layer_id: int = 0,
         quant_config: Optional[QuantizationConfig] = None,
+        reduce_results: Optional[bool] = True,
         prefix: str = "",
-    ):
+        tp_rank: Optional[int] = None,
+        tp_size: Optional[int] = None,
+    ) -> None:
         super().__init__()
-        self.hidden_size = config.hidden_size
-        tp_size = get_tensor_model_parallel_world_size()
-
-        self.total_num_heads = config.num_attention_heads
-        self.total_num_kv_heads = config.num_key_value_heads
-
-        assert self.total_num_heads % tp_size == 0
-        assert self.total_num_kv_heads % tp_size == 0
-
-        self.num_heads = self.total_num_heads // tp_size
-        self.head_dim = config.head_dim or (self.hidden_size // self.total_num_heads)
-        self.q_size = self.num_heads * self.head_dim
-
-        self.num_kv_heads = self.total_num_kv_heads // tp_size
-        self.kv_size = self.num_kv_heads * self.head_dim
-        self.scale = self.head_dim**-0.5
-
-        self.query_key_value = QKVParallelLinear(
-            self.hidden_size,
-            self.head_dim,
-            self.total_num_heads,
-            self.total_num_kv_heads,
-            bias=(config.use_bias or config.use_qkv_bias),
-            quant_config=quant_config,
-            prefix=add_prefix("query_key_value", prefix),
-        )
+        self.tp_size = tp_size
 
-        self.dense = RowParallelLinear(
-            self.total_num_heads * self.head_dim,
-            self.hidden_size,
+        self.gate_up_proj = MergedColumnParallelLinear(
+            config.hidden_size,
+            [intermediate_size] * 2,
             bias=config.use_bias,
             quant_config=quant_config,
-            prefix=add_prefix("dense", prefix),
+            prefix=add_prefix("gate_up_proj", prefix),
+            tp_rank=tp_rank,
+            tp_size=tp_size,
         )
-
-        self.attn = RadixAttention(
-            self.num_heads,
-            self.head_dim,
-            self.scale,
-            num_kv_heads=self.num_kv_heads,
-            layer_id=layer_id,
+        self.down_proj = RowParallelLinear(
+            intermediate_size,
+            config.hidden_size,
+            bias=config.use_bias,
+            reduce_results=reduce_results,
             quant_config=quant_config,
-            prefix=add_prefix("attn", prefix),
+            prefix=add_prefix("down_proj", prefix),
+            tp_rank=tp_rank,
+            tp_size=tp_size,
         )
 
-        self.rotary_emb = get_rope(
-            self.head_dim,
-            rotary_dim=self.head_dim,
-            max_position=config.max_position_embeddings,
-            base=config.rope_theta,
-            is_neox_style=True,
-            rope_scaling=config.rope_scaling,
-        )
+        if config.hidden_act != "silu":
+            raise ValueError("Unsupported activation. Only silu is supported for now.")
+        self.act_fn = SiluAndMul()
 
     def forward(
         self,
         hidden_states: torch.Tensor,
-        position_ids: torch.Tensor,
-        forward_batch: ForwardBatch,
+        forward_batch: Optional[ForwardBatch] = None,
+        use_reduce_scatter: bool = False,
     ) -> torch.Tensor:
-        qkv, _ = self.query_key_value(hidden_states)
-        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
+        if (self.tp_size == 1) and hidden_states.shape[0] == 0:
+            return hidden_states
 
-        q, k = self.rotary_emb(position_ids, q, k)
-        context_layer = self.attn(q, k, v, forward_batch)
-        attn_output, _ = self.dense(context_layer)
-        return attn_output
+        gate_up, _ = self.gate_up_proj(hidden_states)
+        hidden_states = self.act_fn(gate_up)
+        hidden_states, _ = self.down_proj(
+            hidden_states, skip_all_reduce=use_reduce_scatter
+        )
+        return hidden_states
 
 
-class BailingMLP(nn.Module):
+class BailingMoEGate(nn.Module):
     def __init__(
         self,
-        intermediate_size: int,
-        config: PretrainedConfig,
-        quant_config: Optional[QuantizationConfig] = None,
-        reduce_results: Optional[bool] = True,
+        config,
+        params_dtype: Optional[torch.dtype] = None,
         prefix: str = "",
-    ) -> None:
+    ):
         super().__init__()
-        self.gate_up_proj = MergedColumnParallelLinear(
-            config.hidden_size,
-            [intermediate_size] * 2,
-            bias=config.use_bias,
-            quant_config=quant_config,
-            prefix=add_prefix("gate_up_proj", prefix),
-        )
-        self.down_proj = RowParallelLinear(
-            intermediate_size,
-            config.hidden_size,
-            bias=config.use_bias,
-            quant_config=quant_config,
-            reduce_results=reduce_results,
-            prefix=add_prefix("down_proj", prefix),
+        if params_dtype is None:
+            params_dtype = torch.get_default_dtype()
+        self.params_dtype = params_dtype
+        self.weight = nn.Parameter(
+            torch.empty(
+                (config.num_experts, config.hidden_size),
+                dtype=self.params_dtype,
+            ),
         )
-        self.act_fn = SiluAndMul()
-
-    def forward(self, x):
-        x, _ = self.gate_up_proj(x)
-        x = self.act_fn(x)
-        x, _ = self.down_proj(x)
-        return x
+        if getattr(config, "moe_router_enable_expert_bias", False):
+            self.expert_bias = nn.Parameter(
+                torch.empty((config.num_experts,), dtype=torch.float32),
+            )
+        else:
+            self.expert_bias = None
 
+    def forward(self, hidden_states):
+        logits = F.linear(hidden_states.to(self.weight.dtype), self.weight, None).to(
+            hidden_states.dtype
+        )
+        return logits
 
-class BailingMoE(nn.Module):
 
+class BailingMoESparseMoeBlock(nn.Module):
     def __init__(
         self,
-        config: PretrainedConfig,
         layer_id: int,
+        config: PretrainedConfig,
         quant_config: Optional[QuantizationConfig] = None,
+        alt_stream: Optional[torch.cuda.Stream] = None,
         prefix: str = "",
     ):
         super().__init__()
+        self.layer_id = layer_id
+        self.alt_stream = alt_stream
         self.tp_size = get_tensor_model_parallel_world_size()
-        self.num_experts = config.num_experts
         self.top_k = config.num_experts_per_tok
+        self.norm_topk_prob = config.norm_topk_prob
         self.hidden_size = config.hidden_size
         self.num_shared_experts = config.num_shared_experts
-        self.norm_expert_prob = config.norm_topk_prob
-        self.moe_intermediate_size = config.moe_intermediate_size
+        self.routed_scaling_factor = getattr(config, "routed_scaling_factor", 1.0)
+        self.score_function = getattr(config, "score_function", None)
 
-        self.gate = ReplicatedLinear(
-            self.hidden_size, self.num_experts, bias=False, quant_config=None
+        if config.hidden_act != "silu":
+            raise ValueError(
+                f"Unsupported activation: {config.hidden_act}. "
+                "Only silu is supported for now."
+            )
+
+        # Gate always runs at half / full precision for now.
+        router_dtype = getattr(config, "router_dtype", None)
+        if router_dtype is None:
+            self.router_dtype = None
+        elif router_dtype == "fp32":
+            self.router_dtype = torch.float32
+        else:
+            self.router_dtype = torch.bfloat16
+
+        # TODO global_server_args_dict["ep_num_redundant_experts"] is used for eplb, not supported now
+        assert global_server_args_dict["ep_num_redundant_experts"] == 0
+        # check group topk
+        self.num_expert_group = getattr(config, "n_group", 0)
+        self.topk_group = getattr(config, "topk_group", 0)
+        if self.num_expert_group > 0 or self.topk_group > 0:
+            assert (
+                self.num_expert_group > 0
+                and 0 < self.topk_group <= self.num_expert_group
+            )
+            self.use_grouped_topk = True
+        else:
+            self.num_expert_group = self.topk_group = None
+            self.use_grouped_topk = False
+
+        self.num_experts = (
+            config.num_experts + global_server_args_dict["ep_num_redundant_experts"]
+        )
+
+        self.gate = BailingMoEGate(
+            config=config,
+            params_dtype=self.router_dtype,
+            prefix=add_prefix("gate", prefix),
+        )
+        self.correction_bias = (
+            self.gate.expert_bias.data if self.gate.expert_bias is not None else None
         )
 
-        self.topk = TopK(top_k=self.top_k, renormalize=self.norm_expert_prob)
+        if self.score_function is not None:
+            assert (
+                self.score_function == "softmax" and self.correction_bias is None
+            ) or (
+                self.score_function == "sigmoid" and self.correction_bias is not None
+            ), "score_function and correction_bias should be in 2 combination (softmax, None) or (sigmoid, not None)"
 
-        self.experts = FusedMoE(
+        self.topk = TopK(
+            top_k=self.top_k,
+            renormalize=self.norm_topk_prob,
+            use_grouped_topk=self.use_grouped_topk,
+            num_expert_group=self.num_expert_group,
+            # num_fused_shared_experts=self.num_fused_shared_experts,
+            topk_group=self.topk_group,
+            correction_bias=self.correction_bias,
+            routed_scaling_factor=self.routed_scaling_factor,
+        )
+
+        self.experts = get_moe_impl_class(quant_config)(
             num_experts=self.num_experts,
             top_k=self.top_k,
-            layer_id=layer_id,
-            hidden_size=self.hidden_size,
-            intermediate_size=self.moe_intermediate_size,
-            reduce_results=False,
+            layer_id=self.layer_id,
+            hidden_size=config.hidden_size,
+            intermediate_size=config.moe_intermediate_size,
             quant_config=quant_config,
+            routed_scaling_factor=self.routed_scaling_factor,
             prefix=add_prefix("experts", prefix),
         )
-
-        if self.num_shared_experts > 0:
-            shared_intermediate_size = (
-                self.moe_intermediate_size * self.num_shared_experts
-            )
-            self.shared_experts = BailingMLP(
-                intermediate_size=shared_intermediate_size,
+        # shared expert
+        if config.num_shared_experts is not None:
+            if hasattr(config, "moe_shared_expert_intermediate_size"):
+                intermediate_size = config.moe_shared_expert_intermediate_size
+            else:
+                intermediate_size = config.moe_intermediate_size
+            intermediate_size *= config.num_shared_experts
+            # disable tp for shared experts when enable deepep moe
+            self.shared_experts = BailingMoEMLP(
+                intermediate_size=intermediate_size,
                 config=config,
                 quant_config=quant_config,
                 reduce_results=False,
                 prefix=add_prefix("shared_experts", prefix),
+                **(
+                    dict(tp_rank=0, tp_size=1)
+                    if get_moe_a2a_backend().is_deepep()
+                    else {}
+                ),
             )
+        # dispatcher
+        if get_moe_a2a_backend().is_deepep():
+            # TODO: we will support tp < ep in the future
+            self.ep_size = get_tensor_model_parallel_world_size()
+
+            self.deepep_dispatcher = DeepEPDispatcher(
+                group=parallel_state.get_tp_group().device_group,
+                router_topk=self.top_k,
+                permute_fusion=True,
+                num_experts=self.num_experts,
+                num_local_experts=config.num_experts // self.tp_size,
+                hidden_size=config.hidden_size,
+                params_dtype=config.torch_dtype,
+                deepep_mode=DeepEPMode[global_server_args_dict["deepep_mode"]],
+                async_finish=True,  # TODO
+                return_recv_hook=True,
+            )
+
+    def forward(
+        self,
+        hidden_states: torch.Tensor,
+        forward_batch: Optional[ForwardBatch] = None,
+        use_reduce_scatter: bool = False,
+    ) -> torch.Tensor:
+        if not get_moe_a2a_backend().is_deepep():
+            return self.forward_normal(hidden_states, use_reduce_scatter)
         else:
-            self.shared_experts = None
+            return self.forward_deepep(hidden_states, forward_batch)
 
-    def forward(self, hidden_states: torch.Tensor) -> torch.Tensor:
-        orig_shape = hidden_states.shape
-        hidden_states_flat = hidden_states.view(-1, self.hidden_size)
+    def get_moe_weights(self):
+        return [
+            x.data
+            for name, x in self.experts.named_parameters()
+            if name not in ["correction_bias"]
+        ]
 
+    def _forward_shared_experts(self, hidden_states: torch.Tensor):
         shared_output = None
-        if self.shared_experts is not None:
-            shared_output = self.shared_experts(hidden_states_flat)
+        if self.num_shared_experts > 0:
+            shared_output = self.shared_experts(hidden_states)
+        return shared_output
 
-        router_logits, _ = self.gate(hidden_states_flat)
-        topk_output = self.topk(hidden_states_flat, router_logits)
-        final_hidden_states = self.experts(hidden_states_flat, topk_output)
+    def _forward_router_experts(self, hidden_states: torch.Tensor):
+        # router_logits: (num_tokens, n_experts)
+        router_logits = self.gate(hidden_states)
+        topk_output = self.topk(hidden_states, router_logits)
+        return self.experts(hidden_states, topk_output)
 
-        if shared_output is not None:
+    def forward_normal_dual_stream(
+        self,
+        hidden_states: torch.Tensor,
+    ) -> torch.Tensor:
+        current_stream = torch.cuda.current_stream()
+        self.alt_stream.wait_stream(current_stream)
+        shared_output = self._forward_shared_experts(hidden_states.clone())
+
+        with torch.cuda.stream(self.alt_stream):
+            router_output = self._forward_router_experts(hidden_states)
+        current_stream.wait_stream(self.alt_stream)
+
+        return router_output, shared_output
+
+    def forward_normal(
+        self,
+        hidden_states: torch.Tensor,
+        use_reduce_scatter: bool = False,
+    ) -> torch.Tensor:
+        num_tokens, hidden_size = hidden_states.shape
+        hidden_states = hidden_states.view(-1, hidden_size)
+
+        DUAL_STREAM_TOKEN_THRESHOLD = 1024
+        if (
+            self.alt_stream is not None
+            and hidden_states.shape[0] > 0
+            and hidden_states.shape[0] <= DUAL_STREAM_TOKEN_THRESHOLD
+            and get_is_capture_mode()
+        ):
+            final_hidden_states, shared_output = self.forward_normal_dual_stream(
+                hidden_states
+            )
+        else:
+            shared_output = self._forward_shared_experts(hidden_states)
+            final_hidden_states = self._forward_router_experts(hidden_states)
+
+        if self.num_shared_experts > 0:
             final_hidden_states = final_hidden_states + shared_output
 
-        if self.tp_size > 1:
+        if self.tp_size > 1 and not use_reduce_scatter:
             final_hidden_states = tensor_model_parallel_all_reduce(final_hidden_states)
+        return final_hidden_states.view(num_tokens, hidden_size)
+
+    def forward_deepep(
+        self, hidden_states: torch.Tensor, forward_batch: ForwardBatch
+    ) -> torch.Tensor:
+        shared_output = None
+        forward_mode = forward_batch.forward_mode
+        if is_non_idle_and_non_empty(forward_mode, hidden_states):
+            router_logits = self.gate(hidden_states)
+            if self.num_shared_experts > 0:
+                shared_output = self.shared_experts(hidden_states)
+
+            topk_weights, topk_idx, _ = self.topk(
+                hidden_states,
+                router_logits,
+                num_token_non_padded=forward_batch.num_token_non_padded,
+                expert_location_dispatch_info=ExpertLocationDispatchInfo.init_new(
+                    layer_id=self.layer_id,
+                ),
+            )
+        else:
+            topk_idx = torch.full(
+                (0, self.top_k), -1, dtype=torch.int, device=hidden_states.device
+            )
+            topk_weights = torch.empty(
+                (0, self.top_k), dtype=torch.float32, device=hidden_states.device
+            )
+
+        if self.ep_size > 1:
+            (
+                hidden_states,
+                topk_idx,
+                topk_weights,
+                reorder_topk_ids,
+                num_recv_tokens_per_expert,
+                seg_indptr,
+                masked_m,
+                expected_m,
+            ) = self.deepep_dispatcher.dispatch(
+                hidden_states,
+                topk_idx,
+                topk_weights,
+                forward_batch=forward_batch,
+            )
+
+        final_hidden_states = self.experts(
+            hidden_states=hidden_states,
+            topk_idx=topk_idx,
+            topk_weights=topk_weights,
+            reorder_topk_ids=reorder_topk_ids,
+            seg_indptr=seg_indptr,
+            masked_m=masked_m,
+            expected_m=expected_m,
+            num_recv_tokens_per_expert=num_recv_tokens_per_expert,
+            forward_batch=forward_batch,
+        )
+        if self.ep_size > 1:
+            final_hidden_states = self.deepep_dispatcher.combine(
+                final_hidden_states,
+                topk_idx,
+                topk_weights,
+                forward_batch=forward_batch,
+            )
 
-        return final_hidden_states.view(orig_shape)
+        final_hidden_states *= self.routed_scaling_factor
 
+        if shared_output is not None:
+            final_hidden_states = final_hidden_states + shared_output
+        return final_hidden_states
 
-class BailingMoeBlock(nn.Module):
 
+class BailingMoEAttention(nn.Module):
     def __init__(
         self,
         config: PretrainedConfig,
-        layer_id: int,
+        layer_id: int = 0,
         quant_config: Optional[QuantizationConfig] = None,
+        reduce_results: bool = True,
         prefix: str = "",
+        alt_stream: Optional[torch.cuda.Stream] = None,
     ):
         super().__init__()
-        self.input_layernorm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
-        self.attention = BailingAttention(
-            config, layer_id, quant_config, prefix=add_prefix("attention", prefix)
+        self.hidden_size = config.hidden_size
+        self.total_num_heads = config.num_attention_heads
+        self.total_kv_heads = config.num_key_value_heads
+        self.dp_size = get_attention_dp_size()
+        attn_tp_rank = get_attention_tp_rank()
+        attn_tp_size = get_attention_tp_size()
+
+        assert self.total_num_heads % attn_tp_size == 0
+        assert self.total_kv_heads % attn_tp_size == 0
+        assert self.total_num_heads >= self.total_kv_heads
+
+        self.num_heads = self.total_num_heads // attn_tp_size
+        self.head_dim = config.head_dim or (self.hidden_size // self.total_num_heads)
+        self.q_size = self.head_dim * self.num_heads
+
+        self.num_kv_heads = self.total_kv_heads // attn_tp_size
+        self.kv_size = max(1, self.num_kv_heads * self.head_dim)
+
+        self.scale = self.head_dim**-0.5
+
+        self.use_qk_norm = getattr(config, "use_qk_norm", False)
+
+        self.query_key_value = QKVParallelLinear(
+            self.hidden_size,
+            self.head_dim,
+            self.total_num_heads,
+            self.total_kv_heads,
+            bias=(config.use_bias or config.use_qkv_bias),
+            quant_config=quant_config,
+            prefix=add_prefix("query_key_value", prefix),
+            tp_rank=attn_tp_rank,
+            tp_size=attn_tp_size,
+        )
+
+        if self.use_qk_norm:
+            self.query_layernorm = RMSNorm(self.head_dim, eps=config.rms_norm_eps)
+            self.key_layernorm = RMSNorm(self.head_dim, eps=config.rms_norm_eps)
+
+        self.dense = RowParallelLinear(
+            self.total_num_heads * self.head_dim,
+            self.hidden_size,
+            bias=config.use_bias,
+            quant_config=quant_config,
+            reduce_results=reduce_results,
+            prefix=add_prefix("dense", prefix),
+            tp_rank=attn_tp_rank,
+            tp_size=attn_tp_size,
         )
-        self.post_attention_layernorm = RMSNorm(
-            config.hidden_size, eps=config.rms_norm_eps
+
+        if hasattr(config, "partial_rotary_factor"):
+            self.rotary_dim = int(self.head_dim * config.partial_rotary_factor)
+        elif hasattr(config, "rotary_dim"):
+            self.rotary_dim = config.rotary_dim
+        else:
+            self.rotary_dim = self.head_dim
+        self.rotary_emb = get_rope(
+            self.head_dim,
+            rotary_dim=self.rotary_dim,
+            max_position=config.max_position_embeddings,
+            base=config.rope_theta,
+            rope_scaling=config.rope_scaling,
         )
-        self.mlp = BailingMoE(
-            config=config,
+
+        self.attn = RadixAttention(
+            self.num_heads,
+            self.head_dim,
+            self.scale,
+            num_kv_heads=self.num_kv_heads,
             layer_id=layer_id,
-            quant_config=quant_config,
-            prefix=add_prefix("mlp", prefix),
+            prefix=add_prefix("attn", prefix),
         )
 
+        self.alt_stream = alt_stream
+
+    def _apply_qk_norm(
+        self, q: torch.Tensor, k: torch.Tensor
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        # overlap qk norm
+        if self.alt_stream is not None and get_is_capture_mode():
+            current_stream = torch.cuda.current_stream()
+            self.alt_stream.wait_stream(current_stream)
+            q_by_head = q.reshape(-1, self.head_dim)
+            q_by_head = self.query_layernorm(q_by_head)
+            with torch.cuda.stream(self.alt_stream):
+                k_by_head = k.reshape(-1, self.head_dim)
+                k_by_head = self.key_layernorm(k_by_head)
+            current_stream.wait_stream(self.alt_stream)
+        else:
+            q_by_head = q.reshape(-1, self.head_dim)
+            q_by_head = self.query_layernorm(q_by_head)
+            k_by_head = k.reshape(-1, self.head_dim)
+            k_by_head = self.key_layernorm(k_by_head)
+        q = q_by_head.view(q.shape)
+        k = k_by_head.view(k.shape)
+        return q, k
+
     def forward(
         self,
+        positions: torch.Tensor,
         hidden_states: torch.Tensor,
-        position_ids: torch.Tensor,
-        residual: Optional[torch.Tensor],
         forward_batch: ForwardBatch,
-    ) -> Tuple[torch.Tensor, torch.Tensor]:
-        # Pre-normalization and residual connection for the attention block
-        if residual is None:
-            residual = hidden_states
-            normed_hidden_states = self.input_layernorm(hidden_states)
+    ) -> torch.Tensor:
+        if hidden_states.shape[0] == 0:
+            return hidden_states
+        qkv, _ = self.query_key_value(hidden_states)
+        q, k, v = qkv.split([self.q_size, self.kv_size, self.kv_size], dim=-1)
+        if self.use_qk_norm:
+            q, k = self._apply_qk_norm(q, k)
+        q, k = self.rotary_emb(
+            positions,
+            q,
+            k,
+            fused_set_kv_buffer_arg=(
+                create_fused_set_kv_buffer_arg(
+                    value=v,
+                    layer=self.attn,
+                    forward_batch=forward_batch,
+                )
+                if enable_fused_set_kv_buffer(forward_batch)
+                else None
+            ),
+        )
+        context_layer = self.attn(
+            q,
+            k,
+            v,
+            forward_batch,
+            save_kv_cache=not enable_fused_set_kv_buffer(forward_batch),
+        )
+        attn_output, _ = self.dense(context_layer)
+        return attn_output
+
+
+class BailingMoEBlock(nn.Module):
+    def __init__(
+        self,
+        config: PretrainedConfig,
+        layer_id: int = 0,
+        quant_config: Optional[QuantizationConfig] = None,
+        prefix: str = "",
+        alt_stream: Optional[torch.cuda.Stream] = None,
+    ):
+        super().__init__()
+        hidden_size = config.hidden_size
+
+        self.input_layernorm = RMSNorm(hidden_size, eps=config.rms_norm_eps)
+        self.dp_size = get_attention_dp_size()
+        self.attention = BailingMoEAttention(
+            config,
+            layer_id,
+            quant_config,
+            reduce_results=False,
+            prefix=add_prefix("attention", prefix),
+            alt_stream=alt_stream,
+        )
+        self.layer_id = layer_id
+        self.attn_tp_size = get_attention_tp_size()
+        self.attn_tp_rank = get_attention_tp_rank()
+
+        self.is_layer_sparse = self._is_layer_sparse(
+            config, layer_id=layer_id, is_nextn=False
+        )
+        is_previous_layer_sparse = self._is_layer_sparse(
+            config, layer_id=layer_id - 1, is_nextn=False
+        )
+
+        self.layer_scatter_modes = LayerScatterModes.init_new(
+            layer_id=layer_id,
+            num_layers=config.num_hidden_layers,
+            is_layer_sparse=self.is_layer_sparse,
+            is_previous_layer_sparse=is_previous_layer_sparse,
+        )
+
+        self.is_last_layer = self.layer_id == config.num_hidden_layers - 1
+
+        if self.is_layer_sparse:
+            self.mlp = BailingMoESparseMoeBlock(
+                layer_id=layer_id,
+                config=config,
+                quant_config=quant_config,
+                alt_stream=alt_stream,
+                prefix=add_prefix("mlp", prefix),
+            )
         else:
-            normed_hidden_states, residual = self.input_layernorm(
-                hidden_states, residual
+            if enable_moe_dense_fully_dp():
+                mlp_tp_rank, mlp_tp_size = 0, 1
+            else:
+                mlp_tp_rank, mlp_tp_size = None, None
+            self.mlp = BailingMoEMLP(
+                intermediate_size=config.intermediate_size,
+                config=config,
+                quant_config=quant_config,
+                prefix=add_prefix("mlp", prefix),
+                tp_rank=mlp_tp_rank,
+                tp_size=mlp_tp_size,
             )
 
-        attn_output = self.attention(
-            hidden_states=normed_hidden_states,
-            position_ids=position_ids,
+        self.post_attention_layernorm = RMSNorm(hidden_size, eps=config.rms_norm_eps)
+
+        self.layer_communicator = LayerCommunicator(
+            layer_scatter_modes=self.layer_scatter_modes,
+            input_layernorm=self.input_layernorm,
+            post_attention_layernorm=self.post_attention_layernorm,
+            allow_reduce_scatter=True,
+        )
+
+    def _is_layer_sparse(
+        self, config: PretrainedConfig, layer_id: int, is_nextn: bool
+    ) -> bool:
+        return is_nextn or (
+            config.num_experts is not None and layer_id >= config.first_k_dense_replace
+        )
+
+    def forward(
+        self,
+        positions: torch.Tensor,
+        hidden_states: torch.Tensor,
+        forward_batch: ForwardBatch,
+        residual: Optional[torch.Tensor],
+    ) -> torch.Tensor:
+        hidden_states, residual = self.layer_communicator.prepare_attn(
+            hidden_states=hidden_states,
+            residual=residual,
+            forward_batch=forward_batch,
+        )
+
+        hidden_states = self.attention(
+            positions=positions,
+            hidden_states=hidden_states,
             forward_batch=forward_batch,
         )
 
-        # Pre-normalization and residual connection for the MLP block
-        normed_hidden_states, residual = self.post_attention_layernorm(
-            attn_output, residual
+        hidden_states, residual = self.layer_communicator.prepare_mlp(
+            hidden_states=hidden_states,
+            residual=residual,
+            forward_batch=forward_batch,
+        )
+
+        # For DP with padding, reduce scatter can be used instead of all-reduce.
+        use_reduce_scatter = self.layer_communicator.should_use_reduce_scatter(
+            forward_batch
+        )
+
+        hidden_states = self.mlp(hidden_states, forward_batch, use_reduce_scatter)
+
+        hidden_states, residual = self.layer_communicator.postprocess_layer(
+            hidden_states=hidden_states,
+            residual=residual,
+            forward_batch=forward_batch,
         )
-        mlp_output = self.mlp(normed_hidden_states)
 
-        return mlp_output, residual
+        return hidden_states, residual
 
 
-class BailingMoeModel(nn.Module):
+class BailingMoEModel(nn.Module):
 
     def __init__(
         self,
         config: PretrainedConfig,
         quant_config: Optional[QuantizationConfig] = None,
+        alt_stream: Optional[torch.cuda.Stream] = None,
         prefix: str = "",
     ):
         super().__init__()
+        self.pp_group = get_pp_group()
         self.config = config
-        self.padding_idx = config.pad_token_id
         self.vocab_size = config.vocab_size
         self.embed_dim = config.hidden_size
+        if self.pp_group.is_first_rank:
+            self.word_embeddings = VocabParallelEmbedding(
+                self.vocab_size,
+                self.embed_dim,
+                quant_config=quant_config,
+                prefix=add_prefix("word_embeddings", prefix),
+                enable_tp=not is_dp_attention_enabled(),
+            )
+        else:
+            self.word_embeddings = PPMissingLayer()
 
-        self.embed_tokens = VocabParallelEmbedding(
-            config.vocab_size,
-            config.hidden_size,
-            prefix=add_prefix("embed_tokens", prefix),
-        )
         self.embedding_dropout = torch.nn.Dropout(config.embedding_dropout)
 
-        self.layers = make_layers(
+        self.layers, self.start_layer, self.end_layer = make_layers(
             config.num_hidden_layers,
-            lambda idx, prefix: BailingMoeBlock(
-                config=config,
+            lambda idx, prefix: BailingMoEBlock(
                 layer_id=idx,
+                config=config,
                 quant_config=quant_config,
                 prefix=prefix,
+                alt_stream=alt_stream,
             ),
+            pp_rank=self.pp_group.rank_in_group,
+            pp_size=self.pp_group.world_size,
             prefix=add_prefix("layers", prefix),
         )
-
-        self.norm = RMSNorm(self.embed_dim, eps=config.rms_norm_eps)
+        if self.pp_group.is_last_rank:
+            self.norm = RMSNorm(self.embed_dim, eps=config.rms_norm_eps)
+        else:
+            self.norm = PPMissingLayer(return_tuple=True)
 
     def forward(
         self,
         input_ids: torch.Tensor,
-        position_ids: torch.Tensor,
+        positions: torch.Tensor,
         forward_batch: ForwardBatch,
-        input_embeds: Optional[torch.Tensor] = None,
-    ) -> torch.Tensor:
-        if input_embeds is None:
-            hidden_states = self.embed_tokens(input_ids)
+        input_embeds: torch.Tensor = None,
+        pp_proxy_tensors: Optional[PPProxyTensors] = None,
+    ) -> Union[torch.Tensor, PPProxyTensors]:
+        if self.pp_group.is_first_rank:
+            if input_embeds is None:
+                hidden_states = self.word_embeddings(input_ids)
+            else:
+                hidden_states = input_embeds
+            residual = None
         else:
-            hidden_states = input_embeds
+            assert pp_proxy_tensors is not None
+            hidden_states = pp_proxy_tensors["hidden_states"]
+            residual = pp_proxy_tensors["residual"]
 
-        residual = None
-        for layer in self.layers:
-            hidden_states, residual = layer(
-                hidden_states,
-                position_ids,
-                residual,
-                forward_batch,
+        for i in range(self.start_layer, self.end_layer):
+            with get_global_expert_distribution_recorder().with_current_layer(i):
+                layer = self.layers[i]
+                hidden_states, residual = layer(
+                    positions,
+                    hidden_states,
+                    forward_batch,
+                    residual,
+                )
+        if not self.pp_group.is_last_rank:
+            return PPProxyTensors(
+                {
+                    "hidden_states": hidden_states,
+                    "residual": residual,
+                }
             )
+        else:
+            if not forward_batch.forward_mode.is_idle():
+                if residual is None:
+                    hidden_states = self.norm(hidden_states)
+                else:
+                    hidden_states, _ = self.norm(hidden_states, residual)
+            return hidden_states
 
-        hidden_states, _ = self.norm(hidden_states, residual)
-        return hidden_states
-
-
-class BailingMoeForCausalLM(nn.Module):
 
+class BailingMoEForCausalLM(nn.Module):
     def __init__(
         self,
         config: PretrainedConfig,
         quant_config: Optional[QuantizationConfig] = None,
-    ) -> None:
+        prefix: str = "",
+    ):
         super().__init__()
+        self.pp_group = get_pp_group()
         self.config = config
-        self.model = BailingMoeModel(config=config, quant_config=quant_config)
-        self.lm_head = ParallelLMHead(
-            num_embeddings=config.vocab_size,
-            embedding_dim=config.hidden_size,
-            quant_config=quant_config,
+        self.quant_config = quant_config
+        alt_stream = torch.cuda.Stream() if _is_cuda else None
+
+        self.model = BailingMoEModel(
+            config,
+            quant_config,
+            alt_stream=alt_stream,
+            prefix=add_prefix("model", ""),
         )
-        if config.tie_word_embeddings:
-            self.lm_head.weight = self.model.embed_tokens.weight
 
+        # tie_word_embeddings为true，复用tie_word_embeddings，反之是独立的
+        if config.tie_word_embeddings:
+            self.lm_head = self.model.word_embeddings
+        else:
+            # TODO something wrong with ParallelLMHead with DP attention enabled
+            self.lm_head = ParallelLMHead(
+                config.vocab_size,
+                config.hidden_size,
+                quant_config=quant_config,
+                prefix=add_prefix("lm_head", prefix),
+                use_attn_tp_group=global_server_args_dict["enable_dp_lm_head"],
+            )
         self.logits_processor = LogitsProcessor(config)
 
+    @property
+    def start_layer(self):
+        return self.model.start_layer
+
+    @property
+    def end_layer(self):
+        return self.model.end_layer
+
+    def get_embed_and_head(self):
+        """Used by the eagle_worker."""
+        return self.model.word_embeddings.weight, self.lm_head.weight
+
+    def set_embed_and_head(self, embed, head):
+        """Used by the eagle_worker."""
+        del self.model.word_embeddings.weight
+        del self.lm_head.weight
+        self.model.word_embeddings.weight = embed
+        self.lm_head.weight = head
+        torch.cuda.empty_cache()
+        torch.cuda.synchronize()
+
+    @torch.no_grad()
     def forward(
         self,
         input_ids: torch.Tensor,
         positions: torch.Tensor,
         forward_batch: ForwardBatch,
-        inputs_embeds: Optional[torch.Tensor] = None,
+        input_embeds: torch.Tensor = None,
+        pp_proxy_tensors: Optional[PPProxyTensors] = None,
     ) -> torch.Tensor:
-        hidden_states = self.model(input_ids, positions, forward_batch, inputs_embeds)
-        return self.logits_processor(
-            input_ids, hidden_states, self.lm_head, forward_batch
+        hidden_states = self.model(
+            input_ids,
+            positions,
+            forward_batch,
+            input_embeds,
+            pp_proxy_tensors=pp_proxy_tensors,
         )
+        if self.pp_group.is_last_rank:
+            return self.logits_processor(
+                input_ids, hidden_states, self.lm_head, forward_batch
+            )
+        else:
+            return hidden_states
 
-    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]]):
+    def load_weights(self, weights: Iterable[Tuple[str, torch.Tensor]], is_nextn=False):
+        if is_nextn:
+            if hasattr(self.config, "num_nextn_predict_layers"):
+                num_nextn_layers = self.config.num_nextn_predict_layers
+                assert num_nextn_layers == 1, "Only 1 nextn layer is supported"
+                # compatible with old design
+                nextn_layer_id = (
+                    0
+                    if self.config.num_hidden_layers == 1
+                    else self.config.num_hidden_layers
+                )
+            else:
+                raise ValueError("num_nextn_predict_layers is not in the config")
 
         stacked_params_mapping = [
+            # (param_name, shard_name, shard_id)
             ("gate_up_proj", "gate_proj", 0),
             ("gate_up_proj", "up_proj", 1),
         ]
 
+        if is_nextn:
+            nextn_layer_prefix = f"model.layers.{nextn_layer_id}"
+            nextn_spec_weight_names = [
+                "final_layernorm",
+                "eh_proj",
+                "enorm",
+                "hnorm",
+            ]
+        # 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",
@@ -381,39 +911,87 @@ class BailingMoeForCausalLM(nn.Module):
 
         params_dict = dict(self.named_parameters())
         for name, loaded_weight in weights:
+            if (
+                ("v_head" in name)
+                or ("inv_freq" in name)
+                or (self.config.tie_word_embeddings and "lm_head" in name)
+            ):
+                continue
 
             if (
                 hasattr(self.config, "norm_head")
                 and self.config.norm_head
                 and "lm_head.weight" in name
             ):
+                import torch.nn.functional as F
+
                 loaded_weight = F.normalize(loaded_weight, dim=0, p=2, eps=1e-7)
 
-            if "model.word_embeddings.weight" == name:
-                name = "model.embed_tokens.weight"
+            if is_nextn:
+                if not name.startswith(nextn_layer_prefix):
+                    continue
+
+                # Use shared head and embed weights from target model
+                if "shared_head.head" in name or "embed_tokens" in name:
+                    continue
+
+                is_decoder = True
+                # For nextn specific weights
+                for weight_name in nextn_spec_weight_names:
+                    if weight_name in name:
+                        name = name.replace(nextn_layer_prefix, "model")
+                        is_decoder = False
+                        break
+                # For decoder layer weights
+                if is_decoder:
+                    name = name.replace(nextn_layer_prefix, "model.decoder")
 
             for param_name, weight_name, shard_id in stacked_params_mapping:
-                if weight_name in name and "mlp.experts" not in name:
-                    full_param_name = name.replace(weight_name, param_name)
-                    param = params_dict[full_param_name]
-                    param.weight_loader(param, loaded_weight, shard_id)
-                    break
+                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") and name not in params_dict:
+                    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 p_name, w_name, e_id, s_id in expert_params_mapping:
-                    if w_name in name and "mlp.experts" in name:
-                        full_param_name = name.replace(w_name, p_name)
-                        param = params_dict[full_param_name]
-                        param.weight_loader(
-                            param,
-                            loaded_weight,
-                            full_param_name,
-                            shard_id=s_id,
-                            expert_id=e_id,
-                        )
-                        break
+                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)
+                    if 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") and name not in params_dict:
                         continue
+                    if name not in params_dict:
+                        continue
 
                     param = params_dict[name]
                     weight_loader = getattr(
@@ -421,5 +999,30 @@ class BailingMoeForCausalLM(nn.Module):
                     )
                     weight_loader(param, loaded_weight)
 
+        if not is_nextn:
+            self.routed_experts_weights_of_layer = {
+                layer_id: layer.mlp.get_moe_weights()
+                for layer_id, layer in enumerate(self.model.layers)
+                if not isinstance(layer, PPMissingLayer)
+                and isinstance(layer.mlp, BailingMoESparseMoeBlock)
+            }
+
+    @classmethod
+    def get_model_config_for_expert_location(cls, config):
+        num_groups = getattr(config, "n_group", 0)
+        return ModelConfigForExpertLocation(
+            num_layers=config.num_hidden_layers,
+            num_logical_experts=config.num_experts,
+            num_groups=None if num_groups == 0 else num_groups,
+        )
+
+
+class BailingMoeForCausalLM(BailingMoEForCausalLM):
+    pass
+
+
+class BailingMoeV2ForCausalLM(BailingMoEForCausalLM):
+    pass
+
 
-EntryClass = BailingMoeForCausalLM
+EntryClass = [BailingMoEForCausalLM, BailingMoeForCausalLM, BailingMoeV2ForCausalLM]