Skip to content

Vocab Size Mismatch #9

@xuwayyy

Description

@xuwayyy

使用官方requirements.txt里面的 transformers==4.38.2 在测试时可以顺利测试,但发现使用 transformers==4.56 时运行

 model = SegEarthR2.from_pretrained(model_path, mask_decoder_cfg=mask_cfg, **kwargs)

会报错词表大小不一致, 错误信息:

RuntimeError: Error(s) in loading state_dict for Linear:
        size mismatch for weight: copying a param with shape torch.Size([50296, 2560]) from checkpoint, the shape in current model is torch.Size([51200, 2560]).

因为在llava_phi.py中lm_head硬编码为51200 (Mipha词表大小),但是训练权重下config.vocab_size: 50296, 这是否代表测试时 lm_head为随机初始化参数并没有加载权重参数, 只不过在低版本 transformers 版本下容许了这种行为?

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