Tried to load this for fine-tuning using following code and says is not supported --so is this an actual model yet? I don't want a distilled Llama model fine-tuned by DeepSeek-R1 generated data.

from unsloth import FastLanguageModel, FastGraniteModel

Setting up the model and tokenizer

max_seq_length = 2048
dtype = None
load_in_4bit = True

Use FastGraniteModel instead of FastLanguageModel

model, tokenizer = FastLanguageModel.from_pretrained(
model_name="unsloth/DeepSeek-R1",
max_seq_length=max_seq_length,
dtype=dtype,
load_in_4bit=load_in_4bit,
token=hf_token,
trust_remote_code=True
)