train_qwen3_codeforces.py

# /// script
# dependencies = [
#     "trl>=0.12.0",
#     "peft>=0.7.0",
#     "transformers>=4.50.0",
#     "accelerate>=0.24.0",
#     "trackio",
#     "bitsandbytes",
# ]
# ///

"""
Fine-tune Qwen3-0.6B on open-r1/codeforces-cots for instruction following.
Dataset: Competitive programming with chain-of-thought reasoning.
"""

import trackio
from datasets import load_dataset
from peft import LoraConfig
from transformers import AutoTokenizer
from trl import SFTTrainer, SFTConfig

# Load tokenizer first to apply chat template
print("Loading tokenizer...")
tokenizer = AutoTokenizer.from_pretrained("Qwen/Qwen3-0.6B")

# Load dataset with Python solutions (decontaminated)
print("Loading dataset open-r1/codeforces-cots...")
dataset = load_dataset(
    "open-r1/codeforces-cots",
    name="solutions_py_decontaminated",
    split="train"
)
print(f"Dataset loaded: {len(dataset)} examples")

# Preprocess dataset to create 'text' column with chat template applied
def preprocess_function(example):
    """Apply chat template to convert messages to text format."""
    messages = example["messages"]
    text = tokenizer.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=False
    )
    return {"text": text}

print("Preprocessing dataset with chat template...")
dataset = dataset.map(
    preprocess_function,
    remove_columns=dataset.column_names,
    desc="Applying chat template"
)
print(f"Preprocessed dataset: {len(dataset)} examples")

# Create train/eval split
print("Creating train/eval split...")
dataset_split = dataset.train_test_split(test_size=0.05, seed=42)
train_dataset = dataset_split["train"]
eval_dataset = dataset_split["test"]
print(f"   Train: {len(train_dataset)} examples")
print(f"   Eval: {len(eval_dataset)} examples")

# Training configuration
config = SFTConfig(
    # Hub settings - CRITICAL
    output_dir="qwen3-0.6b-codeforces-cots",
    push_to_hub=True,
    hub_model_id="stmasson/qwen3-0.6b-codeforces-cots",
    hub_strategy="every_save",

    # Training parameters
    num_train_epochs=1,
    per_device_train_batch_size=2,
    gradient_accumulation_steps=8,
    learning_rate=2e-4,
    max_length=2048,

    # Logging & checkpointing
    logging_steps=25,
    save_strategy="steps",
    save_steps=500,
    save_total_limit=2,

    # Evaluation
    eval_strategy="steps",
    eval_steps=500,

    # Optimization
    warmup_ratio=0.1,
    lr_scheduler_type="cosine",
    bf16=True,
    gradient_checkpointing=True,

    # Monitoring
    report_to="trackio",
    project="codeforces-finetuning",
    run_name="qwen3-0.6b-codeforces-sft",

    # Dataset field
    dataset_text_field="text",
)

# LoRA configuration for efficient training
peft_config = LoraConfig(
    r=32,
    lora_alpha=64,
    lora_dropout=0.05,
    bias="none",
    task_type="CAUSAL_LM",
    target_modules=["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
)

# Initialize trainer
print("Initializing trainer with Qwen/Qwen3-0.6B...")
trainer = SFTTrainer(
    model="Qwen/Qwen3-0.6B",
    train_dataset=train_dataset,
    eval_dataset=eval_dataset,
    args=config,
    peft_config=peft_config,
)

print("Starting training...")
trainer.train()

print("Pushing final model to Hub...")
trainer.push_to_hub()

print("Training complete! Model at: https://huggingface.co/stmasson/qwen3-0.6b-codeforces-cots")
print("View metrics at: https://huggingface.co/spaces/stmasson/trackio")