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

Support tuning moe for llama 4 model (#6042)

Browse Source
This commit is contained in:
fzyzcjy
2025-05-13 06:47:01 +08:00
committed by GitHub
parent f24fc5b86d
commit ef8ec07b2c
1 changed files with 7 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
benchmark/kernels/fused_moe_triton/tuning_fused_moe_triton.py
View File

@@ -408,6 +408,12 @@ 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] == "Llama4ForConditionalGeneration":
n_share_fusion_experts = args.n_share_experts_fusion
E = config.text_config.num_local_experts + n_share_fusion_experts
topk = config.text_config.num_experts_per_tok
intermediate_size = config.text_config.intermediate_size
shard_intermediate_size = 2 * intermediate_size // args.tp_size
elif config.architectures[0] in [
"Grok1ForCausalLM",
"Grok1ImgGen",
@@ -424,7 +430,7 @@ def main(args: argparse.Namespace):
intermediate_size = config.intermediate_size
shard_intermediate_size = 2 * intermediate_size // args.tp_size
hidden_size = config.hidden_size
hidden_size = getattr(config, "hidden_size", None) or config.text_config.hidden_size
dtype = config.torch_dtype
use_fp8_w8a8 = args.dtype == "fp8_w8a8"
use_int8_w8a8 = args.dtype == "int8_w8a8"