Accelerating PyTorch distributed fine-tuning with Intel technologies