Error while trying to merge two loras: mismatch tensor

[zBilalz]

The error I received:

Error occurred when executing LoraMerger|cgem156:

The size of tensor a (64) must match the size of tensor b (32) at non-singleton dimension 1

File "C:\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\ComfyUI\custom_nodes\cgem156-ComfyUI\scripts\lora_merger\merge.py", line 46, in lora_merge
lora = self.merge(lora_1, lora_2, mode, rank, threshold, device, dtype)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\Users\name\AppData\Local\Programs\Python\Python312\Lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\ComfyUI\custom_nodes\cgem156-ComfyUI\scripts\lora_merger\merge.py", line 98, in merge
up = up_1 + up_2

[laksjdjf]

add mode can only be used between LoRAs of the same rank; try svd mode.

[zBilalz]

add mode can only be used between LoRAs of the same rank; try svd mode.

Working now, thanks

[ScrapWare]

Not needed SVD it's very slow... I developed fastest lite merge methods on logically. Anyone could every updown value stacking using torch.add with doubled alpha. These not related almost quality. Simply split already stacked dims If lots dims. If less dims, clone enough dims on torch.cat. LoHa is can convert LoRA updown hada_w1_a * hada_w1_b(up) and similar hada_w2_ab(down) it implements through few line codes.

However, unnecessary split(same famous SVD) code can change to any quantization in lots dims.

Not enough research LoKr... now use SVD, haven't testing LoKr then taraining factor=1to8 Lokr and ia3 lora-fa glora full from on now.