CS230 Code Examples

This repo is a small workflow collection rather than an installable package. It contains four user-facing example families:

• PyTorch vision: SIGNS image classification.
• PyTorch NLP: named-entity recognition.
• TensorFlow vision: SIGNS image classification.
• TensorFlow NLP: named-entity recognition.

Use this root skill to choose the right framework sub-skill, confirm the shared environment, and find the repo-wide troubleshooting notes.

Read first

• references/repo-provenance.md when you need to check whether this skill is current for the repository checkout.
• references/troubleshooting.md for cross-cutting setup, import, and data layout issues.
• scripts/check_env.py for a safe shared import/version check.

Route map

• sub-skills/pytorch-examples/ for all PyTorch vision and NLP workflows.
• sub-skills/tensorflow-examples/ for all TensorFlow vision and NLP workflows.

Choose the sub-skill by framework first, then read that sub-skill's workflow reference for the specific domain command.

What the root skill covers

Use the root skill when you need one of these:

• a high-level overview of the repository topology;
• a pointer to the right framework and workflow family;
• shared installation or import guidance;
• a repo-wide troubleshooting hint before you drill into a sub-skill;
• provenance/staleness checking for this generated skill.

Do not use the root skill for command-level details. The sub-skills own the actual commands, data layouts, and workflow notes.

Shared setup guidance

• There is no top-level installable Python distribution.
• Install the runtime dependencies from the selected framework requirements files under pytorch/ or tensorflow/.
• For a mixed inspection environment, install both framework stacks plus the shared helpers used by the examples: numpy, Pillow, tabulate, and tqdm.
• A fresh isolated environment is preferred because TensorFlow 1.15 is sensitive to protobuf and legacy CUDA runtime mismatches.
• PyTorch will use CUDA automatically when the host and wheel support it, but the repo's workflows are still valid on CPU-only hosts.

Example install commands:

python -m pip install -r pytorch/vision/requirements.txt
python -m pip install -r pytorch/nlp/requirements.txt
python -m pip install -r tensorflow/vision/requirements.txt
python -m pip install -r tensorflow/nlp/requirements.txt

Pick the requirement files that match the framework workflows you plan to use.

Minimal shared check

Run the bundled diagnostic before a workflow-specific command:

python scripts/check_env.py --frameworks pytorch tensorflow

Add --repo-root <repo-path> when you also want the helper to probe the local workflow modules from the current checkout.

Repository layout at a glance

• pytorch/vision/ and tensorflow/vision/ both work on the SIGNS dataset.
• pytorch/nlp/ and tensorflow/nlp/ both work on the NER text datasets.
• Each family has its own build_*, train.py, evaluate.py, search_hyperparams.py, and synthesize_results.py scripts.
• The starter experiment directories under experiments/ contain the default params.json files used by the example commands.

When to hand off to a sub-skill

• If the user mentions images, hand signs, SIGNS, resizing, or 64x64 image preprocessing, switch to the relevant vision sub-skill.
• If the user mentions sentences, tags, NER, vocab building, or Kaggle CSV splitting, switch to the relevant NLP sub-skill.
• If the user wants a command that should run in the repo checkout, use the framework sub-skill's bundled workflow helper rather than the source script path directly.