def training_function()

in sagemaker/24_train_bloom_peft_lora/scripts/run_clm.py [0:0]

• 58 lines of code
• 8 McCabe index (conditional complexity)

def training_function(args):
    # set seed
    set_seed(args.seed)

    dataset = load_from_disk(args.dataset_path)
    # load model from the hub
    model = AutoModelForCausalLM.from_pretrained(
        args.model_id,
        use_cache=False if args.gradient_checkpointing else True,  # this is needed for gradient checkpointing
        device_map="auto",
        load_in_8bit=True,
    )
    # create peft config
    model = create_peft_config(model)

    # Define training args
    output_dir = "/tmp"
    training_args = TrainingArguments(
        output_dir=output_dir,
        overwrite_output_dir=True,
        per_device_train_batch_size=args.per_device_train_batch_size,
        bf16=args.bf16,  # Use BF16 if available
        learning_rate=args.lr,
        num_train_epochs=args.epochs,
        gradient_checkpointing=args.gradient_checkpointing,
        gradient_accumulation_steps=2,
        # logging strategies
        logging_dir=f"{output_dir}/logs",
        logging_strategy="steps",
        logging_steps=10,
        save_strategy="no",
        optim="adafactor",
    )

    # Create Trainer instance
    trainer = Trainer(
        model=model,
        args=training_args,
        train_dataset=dataset,
        data_collator=default_data_collator,
    )

    # Start training
    trainer.train()

    # merge adapter weights with base model and save
    # save int 8 model
    trainer.model.save_pretrained(output_dir)
    # clear memory
    del model
    del trainer
    # load PEFT model in fp16
    peft_config = PeftConfig.from_pretrained(output_dir)
    model = AutoModelForCausalLM.from_pretrained(
        peft_config.base_model_name_or_path,
        return_dict=True,
        torch_dtype=torch.float16,
        low_cpu_mem_usage=True,
    )
    model = PeftModel.from_pretrained(model, output_dir)
    model.eval()
    # Merge LoRA and base model and save
    merged_model = model.merge_and_unload()
    merged_model.save_pretrained("/opt/ml/model/")

    # save tokenizer for easy inference
    tokenizer = AutoTokenizer.from_pretrained(args.model_id)
    tokenizer.save_pretrained("/opt/ml/model/")

    # copy inference script
    os.makedirs("/opt/ml/model/code", exist_ok=True)
    shutil.copyfile(
        os.path.join(os.path.dirname(__file__), "inference.py"),
        "/opt/ml/model/code/inference.py",
    )
    shutil.copyfile(
        os.path.join(os.path.dirname(__file__), "requirements.txt"),
        "/opt/ml/model/code/requirements.txt",
    )