Script to convert the .pt slider to an usable .safetensors slider

peft.py

import os.path
from collections import OrderedDict
from safetensors.torch import save_file

import torch

state_dict=torch.load("path to your slider.pt")

alpha_keys = [
    'lora_unet_single_transformer_blocks_0_attn_to_q.alpha'
]
rank_idx0_keys = [
    'lora_unet_single_transformer_blocks_0_attn_to_q.lora_down.weight'
]

alpha = None
rank = None

for key in rank_idx0_keys:
    if key in state_dict:
        rank = int(state_dict[key].shape[0])
        break

if rank is None:
    raise ValueError(f'Could not find rank in state dict')


for key in alpha_keys:
    if key in state_dict:
        alpha = int(state_dict[key])
        break

if alpha is None:
    # set to rank if not found
    alpha = rank


#up_multiplier = alpha / rank
up_multiplier=0.125 # manual multiplier setting

new_state_dict = {}

for key, value in state_dict.items():
    if key.endswith('.alpha'):
        continue
    orig_dtype = value.dtype

    new_val = value.float() * up_multiplier

    new_key = key
    new_key = new_key.replace('lora_unet_', 'transformer.')
    for i in range(100):
        new_key = new_key.replace(f'transformer_blocks_{i}_', f'transformer_blocks.{i}.')
    new_key = new_key.replace('lora_down', 'lora_A')
    new_key = new_key.replace('lora_up', 'lora_B')
    new_key = new_key.replace('_lora', '.lora')
    new_key = new_key.replace('attn_', 'attn.')
    new_key = new_key.replace('ff_', 'ff.')
    new_key = new_key.replace('context_net_', 'context.net.')
    new_key = new_key.replace('0_proj', '0.proj')
    new_key = new_key.replace('norm_linear', 'norm.linear')
    new_key = new_key.replace('norm_out_linear', 'norm_out.linear')
    new_key = new_key.replace('to_out_', 'to_out.')

    new_state_dict[new_key] = new_val.to(orig_dtype)

meta = OrderedDict()
meta['format'] = 'pt'

save_file(new_state_dict,"path to your output safentensors file") # must be .safetensors