Tensorflow Deep Learning

What it does

When your agent uses it

• Writing tf.data input pipelines
• Training a Keras model
• Configuring callbacks for checkpointing and early stopping
• Evaluating a trained model

Example prompts

• “/tensorflow-deep-learning”

Requirements

• Python 3

Workflow steps

8 steps, taken from the first numbered list in SKILL.md.

1. 1Separate concerns into modules — Keep data loading, model definition, training, evaluation, and serving code in distinct files; treat…
2. 2Build the input pipeline first — Use tf.data.Dataset for scalable, prefetching input processing; validate shapes, dtypes, label ranges…
3. 3Split before any fitting — Create train/validation/test splits before fitting normalization statistics or augmentation parameters, to…
4. 4Start with a tiny baseline — Build a small model and run a tiny overfit test (fit on a handful of examples until loss goes to ~0) to prove…
5. 5Train with callbacks — Wire in ModelCheckpoint (save best by validation metric, not final epoch), EarlyStopping, a learning-rate schedule…
6. 6Evaluate with task-appropriate metrics — Use a held-out test set only for final reporting, never for tuning.
7. 7Export with a explicit signature — Save with model.export() or tf.saved_model.save() and define serving input signatures explicitly.
8. 8Smoke test the export — Load the exported model and run inference on sample inputs before deploying.

What it can do on your machine

Read from SKILL.md and the folder at commit 9718410. It shows what the files ask for, not the result of running them.

• Tool permissions

  Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

  From allowed-tools in the SKILL.md frontmatter.

• Runs code

  No scripts in the folder and no shell commands in SKILL.md (its code samples are python).

  From the folder's file list and the shell code blocks in SKILL.md.

• Network

  No URLs in SKILL.md.

  From URLs in SKILL.md, links to its own repository left out.

• Credentials

  Names no API keys, tokens, secrets or passwords.

  From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Tensorflow Deep Learning loads about 2.5k tokens when it runs. Until then it costs about 83 tokens; SKILL.md has 828 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~83

When it runs · the whole SKILL.md, loaded when a task matches

~2.5k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Mindrally/skills at commit 9718410, republished under its Apache-2.0 licence (© Mindrally). 828 words, ~2,476 tokens.

name

tensorflow-deep-learning

description

Best practices for building, training, evaluating, and deploying neural networks with TensorFlow and Keras. Use when writing tf.data input pipelines, defining or training a Keras model, configuring callbacks for checkpointing and early stopping, evaluating a trained model, or exporting a model for serving.

project/
  data.py          # tf.data pipeline construction
  model.py         # Keras model definition
  config.py        # typed hyperparameter config
  train.py         # training loop / Keras fit orchestration
  evaluate.py       # metrics computation on held-out data
  serve.py         # export + inference smoke test
  artifacts/       # checkpoints, logs, saved models (gitignored)

import tensorflow as tf

def build_dataset(file_pattern: str, batch_size: int, shuffle: bool) -> tf.data.Dataset:
    files = tf.data.Dataset.list_files(file_pattern, shuffle=shuffle)
    ds = files.interleave(
        tf.data.TFRecordDataset,
        cycle_length=tf.data.AUTOTUNE,
        num_parallel_calls=tf.data.AUTOTUNE,
    )
    ds = ds.map(_parse_example, num_parallel_calls=tf.data.AUTOTUNE)
    if shuffle:
        ds = ds.shuffle(buffer_size=10_000)
    ds = ds.batch(batch_size, drop_remainder=False)
    return ds.prefetch(tf.data.AUTOTUNE)

def _parse_example(record: tf.Tensor) -> tuple[tf.Tensor, tf.Tensor]:
    feature_spec = {
        "features": tf.io.FixedLenFeature([32], tf.float32),
        "label": tf.io.FixedLenFeature([], tf.int64),
    }
    parsed = tf.io.parse_single_example(record, feature_spec)
    return parsed["features"], parsed["label"]

import tensorflow as tf

def build_model(input_dim: int, num_classes: int) -> tf.keras.Model:
    inputs = tf.keras.Input(shape=(input_dim,), name="features")
    x = tf.keras.layers.Dense(128, activation="relu")(inputs)
    x = tf.keras.layers.Dropout(0.3)(x)
    x = tf.keras.layers.Dense(64, activation="relu")(x)
    outputs = tf.keras.layers.Dense(num_classes, activation="softmax", name="predictions")(x)
    return tf.keras.Model(inputs=inputs, outputs=outputs, name="baseline_classifier")

import tensorflow as tf

model = build_model(input_dim=32, num_classes=10)
model.compile(
    optimizer=tf.keras.optimizers.Adam(learning_rate=1e-3),
    loss="sparse_categorical_crossentropy",
    metrics=["accuracy"],
)

callbacks = [
    tf.keras.callbacks.ModelCheckpoint(
        filepath="artifacts/checkpoints/best.keras",
        monitor="val_accuracy",
        save_best_only=True,
    ),
    tf.keras.callbacks.EarlyStopping(
        monitor="val_loss", patience=5, restore_best_weights=True,
    ),
    tf.keras.callbacks.ReduceLROnPlateau(
        monitor="val_loss", factor=0.5, patience=3, min_lr=1e-6,
    ),
    tf.keras.callbacks.TensorBoard(log_dir="artifacts/logs"),
]

train_ds = build_dataset("data/train-*.tfrecord", batch_size=64, shuffle=True)
val_ds = build_dataset("data/val-*.tfrecord", batch_size=64, shuffle=False)

history = model.fit(
    train_ds,
    validation_data=val_ds,
    epochs=50,
    callbacks=callbacks,
)

test_ds = build_dataset("data/test-*.tfrecord", batch_size=64, shuffle=False)
results = model.evaluate(test_ds, return_dict=True)
print(results)  # {'loss': ..., 'accuracy': ...}

model.export("artifacts/saved_model/v1")

# Or with an explicit serving signature:
@tf.function(input_signature=[tf.TensorSpec(shape=[None, 32], dtype=tf.float32, name="features")])
def serve_fn(features):
    return {"predictions": model(features, training=False)}

tf.saved_model.save(model, "artifacts/saved_model/v1", signatures={"serving_default": serve_fn})

© Mindrally, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in tensorflow-deep-learning of Mindrally/skills.

Open the folder on GitHubat commit 9718410