I cannot load GPT-J to 12 GB VRAM Titan XP

I read here that it is possible to load GPT-J to 12 GB card - Memory use of GPT-J-6B

However I tried but I could not even with recommended optimisation:

model =  GPTJForCausalLM.from_pretrained("EleutherAI/gpt-j-6B", revision="float16", torch_dtype=torch.float16, low_cpu_mem_usage=True, cache_dir = "/root/Desktop/models_cache/")

I get out of memory error when I try to load.