Files
sglang/python/sglang/srt/models/glm4_moe.py
2025-09-28 11:21:27 -07:00

1084 lines
41 KiB
Python

# Copyright 2025-2026 SGLang Team
# 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 GLM-4.5, GLM-4.6 model compatible with HuggingFace weights"""
import logging
from typing import Any, Dict, Iterable, Optional, Tuple
import torch
import torch.nn.functional as F
from torch import nn
from transformers import PretrainedConfig
from sglang.srt.distributed import (
get_moe_expert_parallel_world_size,
get_pp_group,
get_tensor_model_parallel_rank,
get_tensor_model_parallel_world_size,
parallel_state,
tensor_model_parallel_all_reduce,
)
from sglang.srt.layers.activation import SiluAndMul
from sglang.srt.layers.amx_utils import PackWeightMethod
from sglang.srt.layers.communicator import (
LayerCommunicator,
LayerScatterModes,
enable_moe_dense_fully_dp,
)
from sglang.srt.layers.dp_attention import (
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 (
ColumnParallelLinear,
MergedColumnParallelLinear,
QKVParallelLinear,
ReplicatedLinear,
RowParallelLinear,
)
from sglang.srt.layers.logits_processor import LogitsProcessor
from sglang.srt.layers.moe import get_deepep_mode, 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.topk import TopK
from sglang.srt.layers.quantization.base_config import QuantizationConfig
from sglang.srt.layers.quantization.fp8_kernel import (
is_fp8_fnuz,
per_tensor_quant_mla_fp8,
per_token_group_quant_mla_deep_gemm_masked_fp8,
)
from sglang.srt.layers.radix_attention import RadixAttention
from sglang.srt.layers.rotary_embedding import get_rope
from sglang.srt.layers.vocab_parallel_embedding import (
ParallelLMHead,
VocabParallelEmbedding,
)
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
from sglang.srt.model_loader.weight_utils import default_weight_loader
from sglang.srt.models.deepseek_v2 import (
DeepseekV2DecoderLayer,
DeepseekV2ForCausalLM,
DeepseekV2Model,
DeepseekV2MoE,
)
from sglang.srt.two_batch_overlap import MaybeTboDeepEPDispatcher
from sglang.srt.utils import (
BumpAllocator,
LazyValue,
add_prefix,
bind_or_assign,
cpu_has_amx_support,
get_bool_env_var,
get_device_sm,
get_int_env_var,
is_cpu,
is_cuda,
is_flashinfer_available,
is_hip,
is_non_idle_and_non_empty,
log_info_on_rank0,
use_intel_amx_backend,
)
_is_hip = is_hip()
_is_cuda = is_cuda()
_is_fp8_fnuz = is_fp8_fnuz()
_use_aiter = get_bool_env_var("SGLANG_USE_AITER") and _is_hip
_is_cpu_amx_available = cpu_has_amx_support()
_is_cpu = is_cpu()
_device_sm = get_device_sm()
if _is_cuda:
from sgl_kernel import dsv3_router_gemm
elif _is_cpu and _is_cpu_amx_available:
pass
logger = logging.getLogger(__name__)
class Glm4MoeMLP(nn.Module):
def __init__(
self,
hidden_size: int,
intermediate_size: int,
hidden_act: str,
quant_config: Optional[QuantizationConfig] = None,
reduce_results: bool = True,
prefix: str = "",
tp_rank: Optional[int] = None,
tp_size: Optional[int] = None,
) -> None:
super().__init__()
self.tp_size = tp_size
self.gate_up_proj = MergedColumnParallelLinear(
hidden_size,
[intermediate_size] * 2,
bias=False,
quant_config=quant_config,
prefix=add_prefix("gate_up_proj", prefix),
tp_rank=tp_rank,
tp_size=tp_size,
)
self.down_proj = RowParallelLinear(
intermediate_size,
hidden_size,
bias=False,
quant_config=quant_config,
reduce_results=reduce_results,
prefix=add_prefix("down_proj", prefix),
tp_rank=tp_rank,
tp_size=tp_size,
)
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,
forward_batch=None,
should_allreduce_fusion=False,
gemm_output_zero_allocator: BumpAllocator = None,
):
if (self.tp_size == 1) and x.shape[0] == 0:
return x
gate_up, _ = self.gate_up_proj(x)
x = self.act_fn(gate_up)
x, _ = self.down_proj(x, skip_all_reduce=should_allreduce_fusion)
return x
class Glm4MoeAttention(nn.Module):
def __init__(
self,
hidden_size: int,
num_heads: int,
num_kv_heads: int,
layer_id: int = 0,
rope_theta: float = 10000,
partial_rotary_factor: float = 0.5,
rope_scaling: Optional[Dict[str, Any]] = None,
max_position_embeddings: int = 8192,
head_dim: Optional[int] = None,
rms_norm_eps: float = 1e-05,
attention_bias: bool = True,
quant_config: Optional[QuantizationConfig] = None,
use_qk_norm: bool = False,
prefix: str = "",
alt_stream: Optional[torch.cuda.Stream] = None,
) -> None:
super().__init__()
self.hidden_size = hidden_size
attn_tp_rank = get_attention_tp_rank()
attn_tp_size = get_attention_tp_size()
self.total_num_heads = num_heads
assert self.total_num_heads % attn_tp_size == 0
self.num_heads = self.total_num_heads // attn_tp_size
self.total_num_kv_heads = num_kv_heads
if self.total_num_kv_heads >= attn_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 % attn_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 attn_tp_size % self.total_num_kv_heads == 0
self.num_kv_heads = max(1, self.total_num_kv_heads // attn_tp_size)
self.head_dim = head_dim or 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.use_qk_norm = use_qk_norm
self.max_position_embeddings = max_position_embeddings
self.tp_rank = get_tensor_model_parallel_rank()
self.qkv_proj = QKVParallelLinear(
hidden_size,
self.head_dim,
self.total_num_heads,
self.total_num_kv_heads,
bias=attention_bias,
quant_config=quant_config,
tp_rank=attn_tp_rank,
tp_size=attn_tp_size,
prefix=add_prefix("qkv_proj", prefix),
)
self.o_proj = RowParallelLinear(
self.total_num_heads * self.head_dim,
hidden_size,
bias=False,
quant_config=quant_config,
tp_rank=attn_tp_rank,
tp_size=attn_tp_size,
reduce_results=False,
prefix=add_prefix("o_proj", prefix),
)
self.rotary_emb = get_rope(
self.head_dim,
rotary_dim=self.head_dim,
max_position=max_position_embeddings,
partial_rotary_factor=partial_rotary_factor,
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,
prefix=add_prefix("attn", prefix),
)
if self.use_qk_norm:
self.q_norm = RMSNorm(self.head_dim, eps=rms_norm_eps)
self.k_norm = RMSNorm(self.head_dim, eps=rms_norm_eps)
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.q_norm(q_by_head)
with torch.cuda.stream(self.alt_stream):
k_by_head = k.reshape(-1, self.head_dim)
k_by_head = self.k_norm(k_by_head)
current_stream.wait_stream(self.alt_stream)
else:
q_by_head = q.reshape(-1, self.head_dim)
q_by_head = self.q_norm(q_by_head)
k_by_head = k.reshape(-1, self.head_dim)
k_by_head = self.k_norm(k_by_head)
q = q_by_head.view(q.shape)
k = k_by_head.view(k.shape)
return q, k
def op_prepare(self, state):
state.attn_intermediate_state = self.forward_prepare(
positions=state.positions,
hidden_states=state.pop("hidden_states_after_comm_pre_attn"),
forward_batch=state.forward_batch,
)
def op_core(self, state):
state.hidden_states_after_attn = self.forward_core(
state.pop("attn_intermediate_state")
)
def forward_prepare(
self,
positions: torch.Tensor,
hidden_states: torch.Tensor,
forward_batch: ForwardBatch,
):
if hidden_states.shape[0] == 0:
return hidden_states, forward_batch, None
qkv, _ = self.qkv_proj(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)
inner_state = q, k, v, forward_batch
return None, forward_batch, inner_state
def forward_core(self, intermediate_state):
hidden_states, forward_batch, inner_state = intermediate_state
if inner_state is None:
return hidden_states
attn_output = self.attn(*inner_state)
output, _ = self.o_proj(attn_output)
return output
def forward(
self,
positions: torch.Tensor,
hidden_states: torch.Tensor,
forward_batch: ForwardBatch,
) -> torch.Tensor:
s = self.forward_prepare(
positions=positions,
hidden_states=hidden_states,
forward_batch=forward_batch,
)
return self.forward_core(s)
class Glm4MoeGate(nn.Module):
def __init__(
self,
config,
prefix: str = "",
is_nextn: bool = False,
):
super().__init__()
self.is_nextn = is_nextn
self.weight = nn.Parameter(
torch.empty((config.n_routed_experts, config.hidden_size))
)
self.e_score_correction_bias = nn.Parameter(
torch.empty((config.n_routed_experts), dtype=torch.float32)
)
if _is_cpu and _is_cpu_amx_available:
self.quant_method = PackWeightMethod(weight_names=["weight"])
def forward(self, hidden_states):
if use_intel_amx_backend(self):
return torch.ops.sgl_kernel.weight_packed_linear(
hidden_states,
self.weight,
None, # bias
True, # is_vnni
)
# NOTE: For some unknown reason, router_gemm seems degrade accept length.
if (
_is_cuda
and not self.is_nextn
and hidden_states.shape[0] < 4
and hidden_states.shape[1] == 7168
and self.weight.shape[0] == 256
and _device_sm >= 90
):
logits = dsv3_router_gemm(hidden_states, self.weight).to(
hidden_states.dtype
)
else:
logits = F.linear(hidden_states, self.weight, None)
return logits
class Glm4MoeSparseMoeBlock(DeepseekV2MoE):
def __init__(
self,
config: PretrainedConfig,
layer_id: int,
quant_config: Optional[QuantizationConfig] = None,
prefix: str = "",
alt_stream: Optional[torch.cuda.Stream] = None,
is_nextn: bool = False,
):
nn.Module.__init__(self)
self.tp_size = get_tensor_model_parallel_world_size()
self.ep_size = get_moe_expert_parallel_world_size()
self.routed_scaling_factor = config.routed_scaling_factor
self.n_shared_experts = config.n_shared_experts
self.num_fused_shared_experts = (
0
if global_server_args_dict["disable_shared_experts_fusion"]
else config.n_shared_experts
)
self.config = config
self.layer_id = layer_id
self.alt_stream = alt_stream
if self.tp_size > config.n_routed_experts:
raise ValueError(
f"Tensor parallel size {self.tp_size} is greater than "
f"the number of experts {config.n_routed_experts}."
)
if config.hidden_act != "silu":
raise ValueError(
f"Unsupported activation: {config.hidden_act}. "
"Only silu is supported for now."
)
self.gate = Glm4MoeGate(
config=config, prefix=add_prefix("gate", prefix), is_nextn=is_nextn
)
self.topk = TopK(
top_k=config.num_experts_per_tok + self.num_fused_shared_experts,
renormalize=config.norm_topk_prob,
use_grouped_topk=True,
num_expert_group=config.n_group,
num_fused_shared_experts=self.num_fused_shared_experts,
topk_group=config.topk_group,
correction_bias=self.gate.e_score_correction_bias,
routed_scaling_factor=self.routed_scaling_factor,
)
self.experts = get_moe_impl_class(quant_config)(
num_experts=config.n_routed_experts
+ self.num_fused_shared_experts
+ global_server_args_dict["ep_num_redundant_experts"],
num_fused_shared_experts=self.num_fused_shared_experts,
top_k=config.num_experts_per_tok + self.num_fused_shared_experts,
hidden_size=config.hidden_size,
intermediate_size=config.moe_intermediate_size,
layer_id=self.layer_id,
quant_config=quant_config,
routed_scaling_factor=self.routed_scaling_factor,
prefix=add_prefix("experts", prefix),
)
self.shared_experts_is_int8 = False
self.shared_experts_is_fp8 = False
# self.shared_experts_weight_block_size = None
if config.n_shared_experts is not None and self.num_fused_shared_experts == 0:
intermediate_size = config.moe_intermediate_size * config.n_shared_experts
self.shared_experts = Glm4MoeMLP(
hidden_size=config.hidden_size,
intermediate_size=intermediate_size,
hidden_act=config.hidden_act,
quant_config=quant_config,
reduce_results=False,
prefix=add_prefix("shared_experts", prefix),
**(dict(tp_rank=0, tp_size=1) if self.ep_size > 1 else {}),
)
is_packed_weight = hasattr(
self.shared_experts.gate_up_proj.quant_method, "quant_config"
)
self.shared_experts_is_int8 = (
not is_packed_weight
and self.shared_experts.gate_up_proj.weight.dtype == torch.int8
)
self.shared_experts_is_fp8 = (
not is_packed_weight
and self.shared_experts.gate_up_proj.weight.dtype == torch.float8_e4m3fn
)
self.top_k = config.num_experts_per_tok
if get_moe_a2a_backend().is_deepep():
# TODO: we will support tp < ep in the future
self.ep_size = get_moe_expert_parallel_world_size()
self.num_experts = (
config.n_routed_experts
+ global_server_args_dict["ep_num_redundant_experts"]
)
self.renormalize = config.norm_topk_prob
self.topk_group = config.topk_group
self.num_expert_group = config.n_group
self.correction_bias = (
self.gate.e_score_correction_bias.data
if self.gate.e_score_correction_bias is not None
else None
)
self.deepep_dispatcher = MaybeTboDeepEPDispatcher(
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.n_routed_experts // self.tp_size,
hidden_size=config.hidden_size,
params_dtype=config.torch_dtype,
deepep_mode=get_deepep_mode(),
async_finish=True,
return_recv_hook=True,
)
self._enable_deepep_moe = get_moe_a2a_backend().is_deepep()
def forward_normal_dual_stream(
self,
hidden_states: torch.Tensor,
should_allreduce_fusion: bool = False,
use_reduce_scatter: bool = False,
gemm_output_zero_allocator: BumpAllocator = None,
) -> torch.Tensor:
current_stream = torch.cuda.current_stream()
self.alt_stream.wait_stream(current_stream)
shared_output = self._forward_shared_experts(hidden_states)
with torch.cuda.stream(self.alt_stream):
# router_logits: (num_tokens, n_experts)
router_logits = self.gate(hidden_states)
topk_output = self.topk(hidden_states, router_logits)
final_hidden_states = self.experts(hidden_states, topk_output)
if not _is_cuda:
final_hidden_states *= self.routed_scaling_factor
current_stream.wait_stream(self.alt_stream)
if self.ep_size > 1:
if (
self.tp_size > 1
and not should_allreduce_fusion
and not use_reduce_scatter
):
final_hidden_states = tensor_model_parallel_all_reduce(
final_hidden_states
)
final_hidden_states += shared_output
else:
final_hidden_states += shared_output
if (
self.tp_size > 1
and not should_allreduce_fusion
and not use_reduce_scatter
):
final_hidden_states = tensor_model_parallel_all_reduce(
final_hidden_states
)
return final_hidden_states
def forward_normal(
self,
hidden_states: torch.Tensor,
should_allreduce_fusion: bool = False,
use_reduce_scatter: bool = False,
gemm_output_zero_allocator: BumpAllocator = None,
) -> torch.Tensor:
if hasattr(self, "shared_experts") and use_intel_amx_backend(
self.shared_experts.gate_up_proj
):
return self.forward_cpu(hidden_states, should_allreduce_fusion)
shared_output = self._forward_shared_experts(hidden_states)
# router_logits: (num_tokens, n_experts)
router_logits = self.gate(hidden_states)
topk_output = self.topk(hidden_states, router_logits)
final_hidden_states = self.experts(hidden_states, topk_output)
if not _is_cuda and not _use_aiter:
# fused in biased_grouped_topk so we can skip here
final_hidden_states *= self.routed_scaling_factor
if self.ep_size > 1:
if self.tp_size > 1 and not should_allreduce_fusion:
final_hidden_states = tensor_model_parallel_all_reduce(
final_hidden_states
)
if shared_output is not None:
final_hidden_states += shared_output
else:
if shared_output is not None:
final_hidden_states += shared_output
if self.tp_size > 1 and not should_allreduce_fusion:
final_hidden_states = tensor_model_parallel_all_reduce(
final_hidden_states
)
return final_hidden_states
class Glm4MoeDecoderLayer(DeepseekV2DecoderLayer):
def __init__(
self,
config: PretrainedConfig,
layer_id: int,
quant_config: Optional[QuantizationConfig] = None,
is_nextn: bool = False,
prefix: str = "",
alt_stream: Optional[torch.cuda.Stream] = None,
) -> None:
nn.Module.__init__(self)
self.hidden_size = config.hidden_size
self.config = config
rope_theta = getattr(config, "rope_theta", 10000)
rope_scaling = getattr(config, "rope_scaling", None)
partial_rotary_factor = getattr(config, "partial_rotary_factor", 0.5)
max_position_embeddings = getattr(config, "max_position_embeddings", 8192)
head_dim = getattr(
config, "head_dim", config.hidden_size // config.num_attention_heads
)
rms_norm_eps = config.rms_norm_eps
attention_bias = config.attention_bias
self.layer_id = layer_id
self.self_attn = Glm4MoeAttention(
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,
partial_rotary_factor=partial_rotary_factor,
max_position_embeddings=max_position_embeddings,
head_dim=head_dim,
rms_norm_eps=rms_norm_eps,
attention_bias=attention_bias,
quant_config=quant_config,
prefix=add_prefix("self_attn", prefix),
use_qk_norm=config.use_qk_norm,
)
self.is_layer_sparse = self._is_layer_sparse(layer_id, is_nextn=is_nextn)
is_previous_layer_sparse = self._is_layer_sparse(layer_id - 1, is_nextn=False)
num_layers = 1 if is_nextn else config.num_hidden_layers
self.layer_scatter_modes = LayerScatterModes.init_new(
layer_id=layer_id,
num_layers=num_layers,
is_layer_sparse=self.is_layer_sparse,
is_previous_layer_sparse=is_previous_layer_sparse,
)
if self.is_layer_sparse:
self.mlp = Glm4MoeSparseMoeBlock(
config=config,
quant_config=quant_config,
prefix=add_prefix("mlp", prefix),
layer_id=self.layer_id,
)
else:
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 = Glm4MoeMLP(
hidden_size=config.hidden_size,
intermediate_size=config.intermediate_size,
hidden_act=config.hidden_act,
quant_config=quant_config,
prefix=add_prefix("mlp", prefix),
tp_rank=mlp_tp_rank,
tp_size=mlp_tp_size,
)
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
)
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 forward(
self,
positions: torch.Tensor,
hidden_states: torch.Tensor,
forward_batch: ForwardBatch,
residual: Optional[torch.Tensor],
zero_allocator: BumpAllocator,
gemm_output_zero_allocator: BumpAllocator = None,
) -> torch.Tensor:
hidden_states, residual = self.layer_communicator.prepare_attn(
hidden_states, residual, forward_batch
)
hidden_states = self.self_attn(
positions=positions,
hidden_states=hidden_states,
forward_batch=forward_batch,
)
hidden_states, residual = self.layer_communicator.prepare_mlp(
hidden_states, residual, forward_batch
)
hidden_states = self.mlp(hidden_states, forward_batch)
hidden_states, residual = self.layer_communicator.postprocess_layer(
hidden_states, residual, forward_batch
)
return hidden_states, residual
class Glm4MoeModel(DeepseekV2Model):
def __init__(
self,
config: PretrainedConfig,
quant_config: Optional[QuantizationConfig] = None,
prefix: str = "",
) -> None:
nn.Module.__init__(self)
self.padding_id = config.pad_token_id
self.vocab_size = config.vocab_size
self.first_k_dense_replace = config.first_k_dense_replace
self.embed_tokens = VocabParallelEmbedding(
config.vocab_size,
config.hidden_size,
enable_tp=not is_dp_attention_enabled(),
)
self.alt_stream = torch.cuda.Stream() if _is_cuda else None
self.layers = nn.ModuleList(
[
Glm4MoeDecoderLayer(
config,
layer_id,
quant_config=quant_config,
prefix=add_prefix(f"layers.{layer_id}", prefix),
alt_stream=self.alt_stream,
)
for layer_id in range(config.num_hidden_layers)
]
)
self.pp_group = get_pp_group()
self.start_layer = 0
self.end_layer = config.num_hidden_layers
self.norm = RMSNorm(config.hidden_size, eps=config.rms_norm_eps)
class Glm4MoeForCausalLM(DeepseekV2ForCausalLM):
def __init__(
self,
config: PretrainedConfig,
quant_config: Optional[QuantizationConfig] = None,
prefix: str = "",
) -> None:
nn.Module.__init__(self)
config.moe_layer_freq = 1
self.config = config
self.tp_size = get_tensor_model_parallel_world_size()
self.quant_config = quant_config
self.pp_group = get_pp_group()
self.determine_num_fused_shared_experts("Glm4MoeForCausalLM")
self.model = Glm4MoeModel(
config, quant_config, prefix=add_prefix("model", prefix)
)
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)
self._routed_experts_weights_of_layer = LazyValue(
lambda: {
layer_id: layer.mlp.get_moe_weights()
for layer_id, layer in enumerate(self.model.layers)
if isinstance(layer.mlp, DeepseekV2MoE)
}
)
def determine_num_fused_shared_experts(
self, architecture: str = "Glm4MoeForCausalLM"
):
self.num_fused_shared_experts = 0
if global_server_args_dict["disable_shared_experts_fusion"]:
return
# Only Deepseek V3/R1 can use shared experts fusion optimization now.
disable_reason = None
if (
not _is_cuda
or torch.cuda.get_device_capability("cuda") < (8, 0)
or self.config.architectures[0] != architecture
or self.config.n_shared_experts != 1
):
disable_reason = "Only GLM-4.5 or GLM-4.6 on NV-platform with capability >= 80 can use shared experts fusion optimization."
elif get_moe_expert_parallel_world_size() > 1:
disable_reason = "Deepseek and GLM-4.5 or GLM-4.6 can not use shared experts fusion optimization under expert parallelism."
if disable_reason is not None:
global_server_args_dict["disable_shared_experts_fusion"] = True
self.num_fused_shared_experts = 0
log_info_on_rank0(
logger,
f"{disable_reason} Shared experts fusion optimization is disabled.",
)
return
self.num_fused_shared_experts = self.config.n_shared_experts
def get_input_embeddings(self) -> nn.Embedding:
return self.model.embed_tokens
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)
("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),
]
if self.num_fused_shared_experts > 0:
assert self.num_fused_shared_experts == 1
weights_list = list(weights)
weights_dict = dict(weights_list)
if self.quant_config is not None:
if self.quant_config.get_name() == "w8a8_int8":
suffix_list = [
"down_proj.weight",
"down_proj.weight_scale",
"gate_proj.weight",
"gate_proj.weight_scale",
"up_proj.weight",
"up_proj.weight_scale",
]
elif (
self.quant_config.get_name() == "fp8"
or self.quant_config.get_name() == "blockwise_int8"
or self.quant_config.get_name() == "compressed_tensors"
):
suffix_list = [
"down_proj.weight",
"down_proj.weight_scale",
"gate_proj.weight",
"gate_proj.weight_scale",
"up_proj.weight",
"up_proj.weight_scale",
]
elif self.quant_config.get_name() == "awq":
suffix_list = [
"down_proj.qweight",
"down_proj.qzeros",
"down_proj.scales",
"gate_proj.qweight",
"gate_proj.qzeros",
"gate_proj.scales",
"up_proj.qweight",
"up_proj.qzeros",
"up_proj.scales",
]
elif self.quant_config.get_name() == "modelopt_fp4":
suffix_list = [
"down_proj.weight",
"down_proj.weight_scale",
"down_proj.weight_scale_2",
"down_proj.input_scale",
"gate_proj.weight",
"gate_proj.weight_scale",
"gate_proj.weight_scale_2",
"gate_proj.input_scale",
"up_proj.weight",
"up_proj.weight_scale",
"up_proj.weight_scale_2",
"up_proj.input_scale",
]
else:
raise ValueError(
f"Unsupported shared expert fusion for quantization: {self.quant_config.get_name()}."
)
else:
suffix_list = [
"down_proj.weight",
"gate_proj.weight",
"up_proj.weight",
]
names_to_remove = []
moe_layers = (
range(
self.config.first_k_dense_replace,
self.config.num_hidden_layers,
self.config.moe_layer_freq,
)
if not is_nextn
else [nextn_layer_id]
)
for moe_layer in moe_layers:
for suffix in suffix_list:
shared_expert_weight_name = (
f"model.layers.{moe_layer}.mlp.shared_experts.{suffix}"
)
# online fp8 quantization does not load weight_scale
if shared_expert_weight_name not in weights_dict:
continue
weights_list.append(
(
f"model.layers.{moe_layer}."
f"mlp.experts."
f"{self.config.n_routed_experts + 0}"
f".{suffix}",
weights_dict[shared_expert_weight_name],
)
)
names_to_remove += [shared_expert_weight_name]
weights = [w for w in weights_list if w[0] not in names_to_remove]
# 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.n_routed_experts + self.num_fused_shared_experts,
)
# Fuse q_a_proj and kv_a_proj_with_mqa along output dimension when q_lora_rank is not None
fuse_qkv_a_proj = hasattr(self.config, "q_lora_rank") and (
self.config.q_lora_rank is not None
)
cached_a_proj = {} if fuse_qkv_a_proj else None
if is_nextn:
nextn_layer_prefix = f"model.layers.{nextn_layer_id}"
nextn_spec_weight_names = [
"shared_head.norm",
"eh_proj",
"enorm",
"hnorm",
]
params_dict = dict(self.named_parameters())
weight_names = []
for name, loaded_weight in weights:
weight_names.append(name)
if not is_nextn:
if hasattr(self.config, "num_nextn_predict_layers"):
num_nextn_layers = self.config.num_nextn_predict_layers
if num_nextn_layers > 0 and name.startswith("model.layers"):
name_list = name.split(".")
if (
len(name_list) >= 3
and int(name_list[2]) >= self.config.num_hidden_layers
):
continue
else:
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")
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) and name not in params_dict:
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
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)
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 fuse_qkv_a_proj and (
"q_a_proj" in name or "kv_a_proj_with_mqa" in name
):
cached_a_proj[name] = loaded_weight
q_a_proj_name = (
name
if "q_a_proj" in name
else name.replace("kv_a_proj_with_mqa", "q_a_proj")
)
kv_a_proj_name = (
name
if "kv_a_proj_with_mqa" in name
else name.replace("q_a_proj", "kv_a_proj_with_mqa")
)
# When both q_a_proj and kv_a_proj_with_mqa has been cached, load the fused weight to parameter
if (
q_a_proj_name in cached_a_proj
and kv_a_proj_name in cached_a_proj
):
q_a_proj_weight = cached_a_proj[q_a_proj_name]
kv_a_proj_weight = cached_a_proj[kv_a_proj_name]
fused_weight = torch.cat(
[q_a_proj_weight, kv_a_proj_weight], dim=0
)
param_name = (
name.replace("q_a_proj", "fused_qkv_a_proj_with_mqa")
if "q_a_proj" in name
else name.replace(
"kv_a_proj_with_mqa", "fused_qkv_a_proj_with_mqa"
)
)
param = params_dict[param_name]
weight_loader = getattr(
param, "weight_loader", default_weight_loader
)
weight_loader(param, fused_weight)
cached_a_proj.pop(q_a_proj_name)
cached_a_proj.pop(kv_a_proj_name)
else:
if (
"k_scale" in name or "v_scale" in name
) and name not in params_dict:
# modelopt attn kv scale is named differently
if any(scale in name for scale in ["k_scale", "v_scale"]):
name = name.replace("_proj", "attn_mqa")
else:
logger.warning(
f"Unknown scale found in checkpoint: {name}"
)
param = params_dict[name]
weight_loader = getattr(
param, "weight_loader", default_weight_loader
)
weight_loader(param, loaded_weight)
EntryClass = [Glm4MoeForCausalLM]