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

GLM-4.5 Model Support (#8224)

Browse Source 
Co-authored-by: Lifu Huang <lifu.hlf@gmail.com>
Co-authored-by: Binyao Jiang <byjiang1996@gmail.com>
Co-authored-by: Stefan He <hebiaobuaa@gmail.com>
This commit is contained in:
Yuxuan Zhang
2025-07-28 13:54:07 +08:00
committed by GitHub
parent 2fd5c7049f
commit 6d6a8bc278
14 changed files with 1673 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
benchmark/kernels/fused_moe_triton/benchmark_sglang_fused_moe_triton.py
View File

@@ -33,7 +33,11 @@ def get_model_config(model_name: str, tp_size: int):
topk = config.num_experts_per_tok
intermediate_size = config.moe_intermediate_size
shard_intermediate_size = 2 * intermediate_size // tp_size
elif config.architectures[0] in ["DeepseekV2ForCausalLM", "DeepseekV3ForCausalLM"]:
elif config.architectures[0] in [
"DeepseekV2ForCausalLM",
"DeepseekV3ForCausalLM",
"Glm4MoeForCausalLM",
]:
E = (
config.n_routed_experts + 1
if config.architectures[0] in ["DeepseekV3ForCausalLM"]

6
benchmark/kernels/fused_moe_triton/benchmark_vllm_vs_sglang_fused_moe_triton.py
View File

@@ -42,7 +42,11 @@ def get_model_config(model_name: str, tp_size: int):
topk = config.num_experts_per_tok
intermediate_size = config.moe_intermediate_size
shard_intermediate_size = 2 * intermediate_size // tp_size
elif config.architectures[0] in ["DeepseekV2ForCausalLM", "DeepseekV3ForCausalLM"]:
elif config.architectures[0] in [
"DeepseekV2ForCausalLM",
"DeepseekV3ForCausalLM",
"Glm4MoeForCausalLM",
]:
E = (
config.n_routed_experts + 1
if config.architectures[0] in ["DeepseekV3ForCausalLM"]