Files
FastDeploy/benchmarks/yaml/qwen3moe235b-32k-wint8-h800-tp4.yaml
2025-07-19 23:19:27 +08:00

7 lines
131 B
YAML

max_model_len: 32768
max_num_seqs: 25
gpu_memory_utilization: 0.9
kv_cache_ratio: 0.75
quantization: wint8
tensor_parallel_size: 4