gemma-3n-E4B-it-FineTome-100k

fourbit_models = [
    # 4bit dynamic quants for superior accuracy and low memory use
    "unsloth/gemma-3n-E4B-it-unsloth-bnb-4bit",
    "unsloth/gemma-3n-E2B-it-unsloth-bnb-4bit",
    # Pretrained models
    "unsloth/gemma-3n-E4B-unsloth-bnb-4bit",
    "unsloth/gemma-3n-E2B-unsloth-bnb-4bit",

    # Other Gemma 3 quants
    "unsloth/gemma-3-1b-it-unsloth-bnb-4bit",
    "unsloth/gemma-3-4b-it-unsloth-bnb-4bit",
    "unsloth/gemma-3-12b-it-unsloth-bnb-4bit",
    "unsloth/gemma-3-27b-it-unsloth-bnb-4bit",
] # More models at https://huggingface.co/unsloth

model, tokenizer = FastModel.from_pretrained(
    model_name = "unsloth/gemma-3n-E4B-it",
    dtype = None, # None for auto detection
    max_seq_length = 1024, # Choose any for long context!
    load_in_4bit = True,  # 4 bit quantization to reduce memory
    full_finetuning = False, # [NEW!] We have full finetuning now!
    # token = "hf_...", # use one if using gated models
)

model = FastModel.get_peft_model(
    model,
    finetune_vision_layers     = False, # Turn off for just text!
    finetune_language_layers   = True,  # Should leave on!
    finetune_attention_modules = True,  # Attention good for GRPO
    finetune_mlp_modules       = True,  # Should leave on always!

    r = 8,           # Larger = higher accuracy, but might overfit
    lora_alpha = 8,  # Recommended alpha == r at least
    lora_dropout = 0,
    bias = "none",
    random_state = 3407,
)

tokenizer = get_chat_template(
    tokenizer,
    chat_template = "gemma-3",
)

from trl import SFTTrainer, SFTConfig
trainer = SFTTrainer(
    model = model,
    tokenizer = tokenizer,
    train_dataset = dataset,
    eval_dataset = None, # Can set up evaluation!
    args = SFTConfig(
        dataset_text_field = "text",
        per_device_train_batch_size = 1,
        gradient_accumulation_steps = 4, # Use GA to mimic batch size!
        warmup_steps = 5,
        # num_train_epochs = 1, # Set this for 1 full training run.
        max_steps = 60,
        learning_rate = 2e-4, # Reduce to 2e-5 for long training runs
        logging_steps = 1,
        optim = "adamw_8bit",
        weight_decay = 0.001,
        lr_scheduler_type = "linear",
        seed = 3407,
        report_to = "none", # Use TrackIO/WandB etc
    ),
)

trainer = train_on_responses_only(
    trainer,
    instruction_part = "<start_of_turn>user\n",
    response_part = "<start_of_turn>model\n",
)