Fine tuning kimi k2 thinking.

[savitha-suresh]

Hello
Can someone help me with finetuning kimik2?

from unsloth import FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = model_path,
    max_seq_length = 2048,
    load_in_4bit = False, 
    trust_remote_code = True,
    device_map = "auto",
    local_files_only = True,
)

I have downloaded the model from here https://huggingface.co/moonshotai/Kimi-K2-Thinking/tree/main/ and put in path
But its stuck in a log line Compressing model: 2302it [02:34, 14.96it/s] which i am assuming is coming from https://github.com/vllm-project/compressed-tensors/blob/4e82c7c01380f93ffd6e5a3bd54c9bc7a1efad9f/src/compressed_tensors/utils/match.py#L55.
Its stuck in the same line for 1hour+
How to surpass this and make this faster so i can start with fine-tuning?