from unsloth import FastModel
model, tokenizer = FastModel.from_pretrained(
    model_name = "unsloth/gemma-3-4b-it",
    max_seq_length = 2048, # Choose any for long context!
    load_in_4bit = True,  # 4 bit quantization to reduce memory
    load_in_8bit = False, # [NEW!] A bit more accurate, uses 2x memory
    full_finetuning = False, # [NEW!] We have full finetuning now!
    # token = "hf_...", # use one if using gated models
)

model.save_pretrained("mano-wii/gemma-3-finetune")  # Local saving
tokenizer.save_pretrained("mano-wii/gemma-3-finetune")

model.save_pretrained_gguf(
    "mano-wii/gemma-3-finetune",
    quantization_type = "Q8_0", # For now only Q8_0, BF16, F16 supported
)

from unsloth_zoo.saving_utils import prepare_saving
repo_id = "mano-wii/gemma-3-finetune-gguf"
prepare_saving(
    model,
    repo_id,
    push_to_hub = True,
    max_shard_size = "50GB",
    private = False,
    token = hf_token,
)

from huggingface_hub import HfApi
api = HfApi(token = hf_token)
api.upload_folder(
    folder_path = "mano-wii",
    repo_id = repo_id,
    repo_type = "model",
    allow_patterns = ["*.gguf"],
)