[New Model] DeepSeek-V3.2 (Rebased to Main) (#25896)

Signed-off-by: Chen Zhang <zhangch99@outlook.com>
Signed-off-by: youkaichao <youkaichao@gmail.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: NickLucche <nlucches@redhat.com>
Signed-off-by: Yongye Zhu <zyy1102000@gmail.com>
Signed-off-by: Barry Kang <43644113+Barry-Delaney@users.noreply.github.com>
Signed-off-by: Lucia Fang <fanglu@meta.com>
Co-authored-by: Chen Zhang <zhangch99@outlook.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: Lucas Wilkinson <lwilkins@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
Co-authored-by: Lucia Fang <116399278+luccafong@users.noreply.github.com>
Co-authored-by: Lucia Fang <fanglu@meta.com>
Co-authored-by: NickLucche <nlucches@redhat.com>
Co-authored-by: Siyuan Fu <siyuanf@nvidia.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Xiaozhu Meng <mxz297@gmail.com>
Co-authored-by: Barry Kang <43644113+Barry-Delaney@users.noreply.github.com>
This commit is contained in:
Yongye Zhu
2025-09-30 17:14:41 +08:00
committed by GitHub
co-authored by Chen Zhang youkaichao Lucas Wilkinson Robert Shaw Lucas Wilkinson yewentao256 Wentao Ye mgoin Lucia Fang Lucia Fang NickLucche Siyuan Fu Matthew Bonanni Xiaozhu Meng Barry Kang
parent e23cacda35
commit fa7e254a7f
71 changed files with 3915 additions and 218 deletions
+2
View File
@@ -66,6 +66,8 @@ class LazyConfigDict(dict):
_CONFIG_REGISTRY: dict[str, type[PretrainedConfig]] = LazyConfigDict(
chatglm="ChatGLMConfig",
deepseek_vl_v2="DeepseekVLV2Config",
deepseek_v3="DeepseekV3Config",
deepseek_v32="DeepseekV3Config",
kimi_vl="KimiVLConfig",
Llama_Nemotron_Nano_VL="Nemotron_Nano_VL_Config",
RefinedWeb="RWConfig", # For tiiuae/falcon-40b(-instruct)
@@ -8,6 +8,7 @@ Model configs may be defined in this directory for the following reasons:
"""
from vllm.transformers_utils.configs.chatglm import ChatGLMConfig
from vllm.transformers_utils.configs.deepseek_v3 import DeepseekV3Config
from vllm.transformers_utils.configs.deepseek_vl2 import DeepseekVLV2Config
from vllm.transformers_utils.configs.dotsocr import DotsOCRConfig
from vllm.transformers_utils.configs.eagle import EAGLEConfig
@@ -37,6 +38,7 @@ from vllm.transformers_utils.configs.ultravox import UltravoxConfig
__all__ = [
"ChatGLMConfig",
"DeepseekVLV2Config",
"DeepseekV3Config",
"DotsOCRConfig",
"EAGLEConfig",
"RWConfig",
@@ -0,0 +1,101 @@
# SPDX-License-Identifier: Apache-2.0
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
from transformers.configuration_utils import PretrainedConfig
from transformers.utils import logging
logger = logging.get_logger(__name__)
class DeepseekV3Config(PretrainedConfig):
model_type = "deepseek_v3"
keys_to_ignore_at_inference = ["past_key_values"]
def __init__(
self,
vocab_size=129280,
hidden_size=7168,
intermediate_size=18432,
moe_intermediate_size=2048,
num_hidden_layers=61,
num_nextn_predict_layers=1,
num_attention_heads=128,
num_key_value_heads=128,
n_shared_experts=1,
n_routed_experts=256,
ep_size=1,
routed_scaling_factor=2.5,
kv_lora_rank=512,
q_lora_rank=1536,
qk_rope_head_dim=64,
v_head_dim=128,
qk_nope_head_dim=128,
topk_method='noaux_tc',
n_group=8,
topk_group=4,
num_experts_per_tok=8,
moe_layer_freq=1,
first_k_dense_replace=3,
norm_topk_prob=True,
scoring_func='sigmoid',
hidden_act="silu",
max_position_embeddings=4096,
initializer_range=0.02,
rms_norm_eps=1e-6,
use_cache=True,
pad_token_id=None,
bos_token_id=0,
eos_token_id=1,
tie_word_embeddings=False,
rope_theta=10000.0,
rope_scaling=None,
attention_bias=False,
attention_dropout=0.0,
**kwargs,
):
self.vocab_size = vocab_size
self.max_position_embeddings = max_position_embeddings
self.hidden_size = hidden_size
self.intermediate_size = intermediate_size
self.moe_intermediate_size = moe_intermediate_size
self.num_hidden_layers = num_hidden_layers
self.num_nextn_predict_layers = num_nextn_predict_layers
self.num_attention_heads = num_attention_heads
self.n_shared_experts = n_shared_experts
self.n_routed_experts = n_routed_experts
self.ep_size = ep_size
self.routed_scaling_factor = routed_scaling_factor
self.kv_lora_rank = kv_lora_rank
self.q_lora_rank = q_lora_rank
self.qk_rope_head_dim = qk_rope_head_dim
self.v_head_dim = v_head_dim
self.qk_nope_head_dim = qk_nope_head_dim
self.topk_method = topk_method
self.n_group = n_group
self.topk_group = topk_group
self.num_experts_per_tok = num_experts_per_tok
self.moe_layer_freq = moe_layer_freq
self.first_k_dense_replace = first_k_dense_replace
self.norm_topk_prob = norm_topk_prob
self.scoring_func = scoring_func
# for backward compatibility
if num_key_value_heads is None:
num_key_value_heads = num_attention_heads
self.num_key_value_heads = num_key_value_heads
self.hidden_act = hidden_act
self.initializer_range = initializer_range
self.rms_norm_eps = rms_norm_eps
self.use_cache = use_cache
self.rope_theta = rope_theta
self.rope_scaling = rope_scaling
self.attention_bias = attention_bias
self.attention_dropout = attention_dropout
super().__init__(
pad_token_id=pad_token_id,
bos_token_id=bos_token_id,
eos_token_id=eos_token_id,
tie_word_embeddings=tie_word_embeddings,
**kwargs,
)