Dreambooth Lora to save CKPT suddenly fails

[allroundHim]

I reinstalled the venv, and one day generate checkpoint from dreambooth suddenly fail to work and gives me this
Exception compiling checkpoint: The size of tensor a (4) must match the size of tensor b (320) at non-singleton dimension 1 Traceback (most recent call last): File "F:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\diff_to_sd.py", line 440, in compile_checkpoint lora_rev = apply_lora(config, unet_model, lora_file_name, "cpu", False) File "F:\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\diff_to_sd.py", line 577, in apply_lora merge_lora_to_model(model, load_model(lora_file_name, device), is_tenc, config.use_lora_extended, File "F:\stable-diffusion-webui\extensions\sd_dreambooth_extension\lora_diffusion\lora.py", line 1156, in merge_lora_to_model collapse_lora(model, weight) File "F:\stable-diffusion-webui\extensions\sd_dreambooth_extension\lora_diffusion\lora.py", line 600, in collapse_lora _child_module.lora_up.weight.data RuntimeError: The size of tensor a (4) must match the size of tensor b (320) at non-singleton dimension 1

Any ideas of what might be the cause?

Way to recreate problel:

1. Go to dreambooth tab
2. pick a model and a lora
3. hit generate button

Current extensions:
openpose-editor
sd-webui-controlnet
sd_dreambooth_extension

I went to where the problem is

        if isinstance(_child_module, LoraInjectedLinear):
            _child_module.linear.weight = nn.Parameter(
                _child_module.linear.weight.data
                + alpha
                * (
                        _child_module.lora_up.weight.data
                        @ _child_module.lora_down.weight.data
                )
                .type(_child_module.linear.weight.dtype)
                .to(_child_module.linear.weight.device)
            )

What does this mean?