Total Trainable param

[Guardian-in-the-WF]

Why did I initialize the otter model with the results about "Total Trainable param: 0.000000 B" after
accelerator = Accelerator(
gradient_accumulation_steps=args.gradient_accumulation_steps,
mixed_precision="no", # mark origin: bf16
)
? I run that code on the RTX A6000 machine.