Skip to content

can not reproduce the result of paibench-G #183

@xyl3902596

Description

@xyl3902596

ti2v_demo.py 默认推理参数

DEFAULT_NUM_FRAMES = 189
DEFAULT_HEIGHT = 720
DEFAULT_WIDTH = 1280
DEFAULT_FPS = 24.0
DEFAULT_GUIDANCE_SCALE = 6.0
DEFAULT_NUM_INFERENCE_STEPS = 35

this is my reproduce parameters, and I use the https://github.com/SHI-Labs/physical-ai-bench as repo with qwen-VL-72B, I only get domain score about 83 when testing cosmos3-nano

installation method: diffusers
reference inference code: https://huggingface.co/docs/diffusers/main/en/api/pipelines/cosmos3

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions