import torch
from datasets import load_dataset
from transformers import (
    AutoTokenizer,
    AutoModelForCausalLM,
    Trainer,
    TrainingArguments,
    DataCollatorForLanguageModeling,
)
from peft import LoraConfig, get_peft_model

MODEL_ID = "ibm-granite/granite-4.0-micro"
DATA_PATH = "sebelsn/style-adjustment-dataset_de/2026-01-17_style-adjustment-dataset_de.jsonl"

# Tokenizer
tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)

# Modell
model = AutoModelForCausalLM.from_pretrained(
    MODEL_ID,
    dtype=torch.float16,
    device_map="cuda"
)

# LoRA-Konfiguration (bewusst konservativ)
lora_config = LoraConfig(
    r=2,
    lora_alpha=4,
    target_modules=["q_proj", "v_proj"],
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
)

model = get_peft_model(model, lora_config)
model.print_trainable_parameters()

# Datensatz laden
dataset = load_dataset("json", data_files=DATA_PATH)["train"]

def format_example(example):
    text = (
        "Frage:\n"
        f"{example['instruction']}\n\n"
        "Antwort:\n"
        f"{example['response']}"
    )
    return {"text": text}

dataset = dataset.map(format_example, remove_columns=dataset.column_names)

def tokenize(example):
    return tokenizer(
        example["text"],
        truncation=True,
        max_length=512,
    )

dataset = dataset.map(tokenize, batched=True)

data_collator = DataCollatorForLanguageModeling(
    tokenizer=tokenizer,
    mlm=False,
)

# TrainingArguments – erster Lauf bewusst sanft
training_args = TrainingArguments(
    output_dir="./lora-out",
    per_device_train_batch_size=2,
    gradient_accumulation_steps=4,
    learning_rate=5e-5,
    warmup_ratio=0.05,
    num_train_epochs=4,
    bf16=True,
    logging_steps=10,
    save_strategy="steps",
    save_steps=30,
    report_to="none",
)

trainer = Trainer(
    model=model,
    args=training_args,
    train_dataset=dataset,
    data_collator=data_collator,
)

trainer.train()