2. Python, notebooks, and tensors

Install a current Python release and verify it with `python --version`. Distinguish the Python interpreter from scripts and select the intended interpreter when multiple installations exist.

Create and activate an isolated environment with `venv`, then install and inspect packages with `pip`. Understand how isolation prevents one generative AI project’s dependencies from disrupting another’s.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Record project dependencies with `requirements.txt` or `pyproject.toml` and recreate an environment from them. Compare loose version constraints with pinned or locked versions for reproducible work.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Create code and Markdown cells, run them, inspect outputs, and use notebook-aware commands such as `%pip`. Connect a notebook to the correct Jupyter kernel and virtual environment.

Reason about how out-of-order execution can leave hidden or stale variables in memory. Restart the kernel and run all cells from top to bottom to verify that a notebook is reproducible.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Work with Python numbers, strings, booleans, `None`, lists, tuples, and dictionaries. Choose a suitable container and inspect values with `type`, `len`, and readable output.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Select values with positive and negative indexes, extract ranges with slices, and retrieve dictionary entries by key. Trace how nested indexing reaches data inside lists and dictionaries.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Use `if` statements, `for` loops, and `while` loops to control execution. Build simple transformed or filtered lists with comprehensions when they make the code clearer.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Define a function with parameters, call it with arguments, and return a result. Use default and keyword arguments to make small data-processing operations reusable and readable.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Import standard-library and third-party modules, use qualified names, and recognize a package’s public API. Organize reusable code in a `.py` file and protect script-only behavior with `if __name__ == "__main__"`.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Read a traceback from the final error upward, identify the exception type and failing line, and inspect intermediate values. Catch expected failures narrowly without hiding programming errors.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Initialize a repository, inspect changes, stage selected files, and create meaningful commits. Use `git status`, `git diff`, and `git log` to trace how a project changed.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Create and switch branches, merge completed work, and resolve the role of a remote repository. Clone, fetch, pull, and push while distinguishing local history from remote-tracking branches.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Use `.gitignore` to exclude environments, caches, generated outputs, datasets, and secrets. Recognize when large model files need artifact storage or Git LFS instead of ordinary Git history.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Create NumPy arrays from Python data and inspect their `shape`, number of dimensions, size, and `dtype`. Interpret axes consistently without turning the lesson into matrix mathematics.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Select elements, rows, columns, and ranges from NumPy arrays and PyTorch tensors. Compare basic slicing with boolean masks and integer index selection.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Change data layout with `reshape`, `squeeze`, `unsqueeze`, and axis transposition or permutation. Check element counts and dimension order to avoid silent shape mistakes.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Replace element-by-element Python loops with vectorized array or tensor operations. Predict how broadcasting aligns dimensions and recognize incompatible shapes before an operation runs.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Create PyTorch tensors and convert between tensors and NumPy arrays. Distinguish copying from shared-memory conversion so that an unexpected mutation does not alter both objects.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Choose integer and floating-point tensor dtypes and move tensors between CPU and an available accelerator. Keep operands and models on compatible devices while avoiding unnecessary transfers.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Construct portable paths with `pathlib`, open text and binary files with context managers, and choose an explicit text encoding. Read or write data without leaking file handles or relying on a fragile working directory.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Parse and produce JSON, process JSON Lines one record at a time, and read simple CSV tables. Choose among these formats based on nesting, streaming, interoperability, and data size.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Save and load numerical arrays, tensors, and model weights with appropriate formats such as NumPy files and `safetensors`. Recognize that pickle-based files can execute code and should not be loaded from untrusted sources.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Send an HTTP request with a library such as `requests`, provide query parameters or a JSON body, and decode a JSON response. Interpret methods, URLs, headers, status codes, and response bodies as parts of one API exchange.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Pass API keys through environment variables or secret-management features rather than source code and notebooks. Prevent credentials from appearing in Git history, logs, screenshots, and shared outputs.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Add timeouts, check status codes, and handle malformed responses, rate limits, and temporary server failures. Use bounded retries with backoff only when repeating the request is safe.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Read a Hugging Face model card to check the task, architecture, size, license, limitations, and expected inputs. Identify checkpoint, configuration, tokenizer, and weight files and understand how local caching avoids repeated downloads.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Trace how a tokenizer turns text into token IDs and decodes IDs back into text. Inspect special tokens, padding, truncation, and attention masks that prepare variable-length text for a model.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Load a small pretrained text model with a Transformers pipeline, pass it a prompt, and inspect the generated result. Confirm that the selected task and model are compatible and observe the first-run download and cache behavior.

Control text generation with `max_new_tokens`, sampling, temperature, top-k or top-p filtering, and a random seed. Compare deterministic decoding with sampled output while treating fluent text as model output that still needs human evaluation.

Check what you understood with a short quiz.

Use what you learned in the previous lesson to solve real-world problems.

Review this chapter with practice based on your mistakes.