Files
FastDeploy/benchmarks/yaml/eb45-vl-32k-wint4-tp4.yaml
2025-06-29 23:29:37 +00:00

10 lines
228 B
YAML

enable_mm: True
max_model_len: 32768
max_num_seqs: 36
gpu_memory_utilization: 0.9
kv_cache_ratio: 0.8
tensor_parallel_size: 4
quantization: wint4
limit_mm_per_prompt: '{"image": 100, "video": 100}'
reasoning_parser: ernie-45-vl