A small TensorFlow script in my project has started throwing errors, and I need it back in working order within 48 hours. The bug sits somewhere between the data-loading step and model training; once you jump in you’ll have full access to the Python code, sample data, and the original (working) training logs so you can compare behaviours. I work entirely in Python and TensorFlow, with a handful of standard utilities such as NumPy, pandas, and Matplotlib. No other frameworks are involved, so deep familiarity with TensorFlow’s debugging tools, graph/symbolic APIs, and eager execution will let you move quickly. Here is what I expect at hand-off: • A corrected script that runs start-to-finish on my machine (TensorFlow 2.x, Python 3.9). • A concise changelog or inline comments so I can see exactly what you fixed and why. Please include a short note on your past TensorFlow experience when you respond. I’m ready to share the repo as soon as we agree, and I can release the fixed-price budget immediately once the script trains without errors and reproduces the original accuracy numbers.