# GUI-Shift Requirements # Core dependencies for VLM training and evaluation # Deep Learning & Transformers torch>=2.1.0 transformers>=4.40.0 trl>=0.13.0 accelerate>=0.28.0 deepspeed>=0.14.0 # Vision & Multimodal qwen-vl-utils Pillow>=10.0.0 # Data Processing datasets>=2.18.0 numpy>=1.24.0 jsonlines>=4.0.0 # Training Utilities wandb>=0.16.0 trackio>=0.5.0 # VLM-R1 Framework (install from GitHub) # git+https://github.com/om-ai-lab/VLM-R1.git # Evaluation scikit-learn>=1.4.0 # Utilities tqdm>=4.66.0 python-dotenv>=1.0.0 babel>=2.15.0 levenshtein>=0.25.0 json-repair>=0.20.0 math-verify>=0.2.0 # Optional: Flash Attention (use hub kernels instead) # flash-attn>=2.5.0