Files
FastDeploy/benchmarks/yaml/GLM45-air-32k-bf16-mtp.yaml
T

8 lines
173 B
YAML

max_model_len: 32768
max_num_seqs: 128
tensor_parallel_size: 4
graph_optimization_config:
use_cudagraph: True
draft_model_use_cudagraph: True
load_choices: "default_v1"