Files
FastDeploy/benchmarks/yaml/GLM45-air-32k-wfp8afp8.yaml
2025-09-26 14:23:29 +08:00

7 lines
133 B
YAML

max_model_len: 32768
max_num_seqs: 128
tensor_parallel_size: 4
use_cudagraph: True
load_choices: "default_v1"
quantization: wfp8afp8