pytorch-optimizer v2.6.1

Change Log

Fix

• variables are not located on the same device with the gradients, #132 (related to #131) (thanks to @Bing-su)
• fix approximate_sq_grad() in Adafactor optimizer, #132