EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

Support LingV2 model (#10359)

Browse Source 
Co-authored-by: 羽癫 <yudian.zy@antgroup.com>
Co-authored-by: guoyuhong <yuhong.gyh@antgroup.com>
This commit is contained in:
strgrb
2025-09-12 14:53:52 +08:00
committed by GitHub
parent b3839a7f99
commit fac07c9b08
7 changed files with 1189 additions and 245 deletions
Download Patch File Download Diff File Expand all files Collapse all files

13
benchmark/kernels/fused_moe_triton/tuning_fused_moe_triton.py
View File

@@ -13,8 +13,8 @@ from ray.experimental.tqdm_ray import tqdm
from transformers import AutoConfig
from sglang.srt.layers.moe.fused_moe_triton import override_config
from sglang.srt.layers.moe.fused_moe_triton.fused_moe import (
fused_moe,
from sglang.srt.layers.moe.fused_moe_triton.fused_moe import fused_moe
from sglang.srt.layers.moe.fused_moe_triton.fused_moe_triton_config import (
get_config_dtype_str,
get_config_file_name,
get_default_config,
@@ -441,6 +441,15 @@ def main(args: argparse.Namespace):
topk = config.num_experts_per_tok
intermediate_size = config.moe_intermediate_size
shard_intermediate_size = 2 * intermediate_size // args.tp_size
elif config.architectures[0] in [
"BailingMoEForCausalLM",
"BailingMoeForCausalLM",
"BailingMoeV2ForCausalLM",
]:
E = config.num_experts
topk = config.num_experts_per_tok
intermediate_size = config.moe_intermediate_size
shard_intermediate_size = 2 * intermediate_size // args.tp_size
elif config.architectures[0] in ["Glm4MoeForCausalLM"]:
E = config.n_routed_experts
topk = config.num_experts_per_tok