How to use Mamba2

[XylonXu01]

Mamba2 is provided in the classification folder, how to apply to VMamba

[MzeroMiko]

You can follow the examples in vmamba.py to use it. One of the examples is as follows:

def vmamba_tiny_m2():
    return VSSM(
        depths=[2, 2, 4, 2], dims=96, drop_path_rate=0.2, 
        patch_size=4, in_chans=3, num_classes=1000, 
        ssm_d_state=64, ssm_ratio=1.0, ssm_dt_rank="auto", ssm_act_layer="gelu",
        ssm_conv=3, ssm_conv_bias=False, ssm_drop_rate=0.0, 
        ssm_init="v2", forward_type="m0_noz", 
        mlp_ratio=4.0, mlp_act_layer="gelu", mlp_drop_rate=0.0, gmlp=False,
        patch_norm=True, norm_layer="ln",
        downsample_version="v3", patchembed_version="v2", 
        use_checkpoint=False, posembed=False, imgsize=224, 
    )