The vmamba tiny model used for segmentation cannot be loaded into the classified tiny vmamba ckpt

[Yi-Heng]

This is the config file I use

     checkpoint_file = "/data/yhzhou23/methods/pretrained/VMamba/vssm_tiny_0230_ckpt_epoch_262.pth"
      encoder=dict(
          out_indices=(0, 1, 2, 3),
          pretrained=checkpoint_file,
          dims=96,
          # depths=(2, 2, 5, 2),
          depths=(2, 2, 8, 2),
          ssm_d_state=1,
          ssm_dt_rank="auto",
          # ssm_ratio=2.0,
          ssm_ratio=1.0,
          ssm_conv=3,
          ssm_conv_bias=False,
          forward_type="v05_noz", # v3_noz,
          mlp_ratio=4.0,
          downsample_version="v3",
          patchembed_version="v2",
          drop_path_rate=0.2,
          norm_layer="ln2d",
      ),

The error message is as follows

Failed loading checkpoint form /data/yhzhou23/methods/pretrained/VMamba/vssm_tiny_0230_ckpt_epoch_262.pth: shape '[96, 96]' is invalid for input of size 18432

[MzeroMiko]

I am sorry about this. I changed the config file days ago, but without changing the readme. It is the file with tiny1 that matches the checkpoint.

I've updated the readme now. Thank you for your reminding.