Spotlight

https://renumics.com/docs
Admin
Spotlight is an open-source data exploration and visualization tool that helps users understand...
Tokens:21,689
Snippets:240
Trust Score:8.4
Update:10 months ago
Show doc for...
Tokens:
Raw
### Install Spotlight via pip

Source: https://renumics.com/docs/getting-started

Installs the renumics-spotlight package using pip. It's recommended to install in a virtual environment.

```bash
pip install renumics-spotlight  
```

--------------------------------

### Initialize Spotlight Development Environment

Source: https://renumics.com/docs/development

Initializes the Spotlight development environment by installing dependencies and setting up pre-commit hooks using the provided Makefile. This command streamlines the setup process.

```bash
make init
```

--------------------------------

### Load CSV and Show with Spotlight (CLI)

Source: https://renumics.com/docs/getting-started

Downloads a CSV file using curl and then displays it using the Spotlight CLI. Custom data types can be specified using --dtype.

```bash
curl https://renumics.com/data/mnist/mnist-tiny.csv -o mnist-tiny.csv
spotlight mnist-tiny.csv --dtype image=Image --dtype embedding=Embedding  
```

--------------------------------

### Install renumics-spotlight and datasets

Source: https://renumics.com/docs/data-centric-ai/playbook/decision-boundary

Installs the required Python packages, renumics-spotlight and datasets, using pip. This is a prerequisite for running the provided code examples.

```python
#@title Install required packages with PIP  

!pip install renumics-spotlight datasets  

```

--------------------------------

### Load Hugging Face Dataset and Show with Spotlight (Python)

Source: https://renumics.com/docs/getting-started

Loads a dataset from the Hugging Face hub and displays it with Spotlight. Spotlight automatically parses data type descriptions and label mappings.

```python
import datasets
from renumics import spotlight

ds = datasets.load_dataset('speech_commands', 'v0.01', split='all')

spotlight.show(ds)  
```

--------------------------------

### Install Dependencies for Towhee and Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/towhee-embedding

Installs the necessary Python packages including renumics-spotlight, towhee, and datasets for image embedding generation and data handling. This is a prerequisite for running the provided code examples.

```python
# Install dependencies
!pip install renumics-spotlight towhee datasets
```

--------------------------------

### Opt-out of Usage Tracking (Bash)

Source: https://renumics.com/docs/getting-started

Sets an environment variable to opt out of crash report and performance collection in Spotlight.

```bash
export SPOTLIGHT_OPT_OUT=true  
```

--------------------------------

### Install Dependencies

Source: https://renumics.com/docs/data-centric-ai/playbook/huggingface-embedding

Installs the necessary Python packages for using Huggingface transformers, torch, datasets, and renumics-spotlight.

```python
!pip install renumics-spotlight transformers torch datasets
```

--------------------------------

### Install and Verify Pre-commit Hooks

Source: https://renumics.com/docs/development

Installs and verifies pre-commit hooks for the Spotlight project using Poetry. These hooks automate code formatting and testing before commits and pushes, ensuring code quality.

```bash
poetry run pre-commit install --hook-type pre-commit
poetry run pre-commit install --hook-type pre-push
```

--------------------------------

### Install Dependencies for Sliceguard and Spotlight

Source: https://renumics.com/docs/use-cases/image-fine-tuning

Installs the necessary Python packages, including renumics-spotlight and sliceguard with all optional dependencies, required for image dataset creation and model fine-tuning.

```bash
pip install renumics-spotlight sliceguard[all]
```

--------------------------------

### Run Spotlight Development Server

Source: https://renumics.com/docs/development

Starts the development server for the Spotlight project using the 'dev' target in the Makefile. This command allows for live development and testing of both the backend and frontend.

```bash
make dev
```

--------------------------------

### Clone and Setup Spotlight Repository

Source: https://renumics.com/docs/development

Clones the Spotlight repository from GitHub and sets up an 'upstream' remote pointing to the original repository. This allows for tracking changes and contributing back to the project.

```bash
git clone https://github.com/YOUR_GIT_USERNAME/spotlight.git
cd spotlight
git remote add upstream https://github.com/renumics/spotlight.git
```

--------------------------------

### Install Node.js and pnpm on Linux

Source: https://renumics.com/docs/development

Installs Node.js and pnpm, a package manager for Node.js, on a Linux system. This is required for developing the React frontend of Spotlight. It also includes a check for the Node.js version.

```bash
sudo apt install nodejs
curl -fsSL https://get.pnpm.io/install.sh | sh -

# check your installed node version
node -v
```

--------------------------------

### Install Python and Poetry on Linux

Source: https://renumics.com/docs/development

Installs Python 3 and Poetry, a dependency management tool for Python, on a Linux system using apt package manager. This is a prerequisite for developing the Python backend of Spotlight.

```bash
sudo apt update
sudo apt install python3 python3-dev
curl -sSL https://install.python-poetry.org | python3 -
```

--------------------------------

### Install Dependencies for SliceGuard and Spotlight

Source: https://renumics.com/docs/use-cases/audio-classification

Installs the necessary Python packages for using renumics-spotlight and sliceguard, including optional dependencies for full functionality. This command is essential before running the data detection code.

```bash
pip install renumics-spotlight sliceguard[all] scikit-learn
```

--------------------------------

### Install Dependencies for Duplicate Detection

Source: https://renumics.com/docs/data-centric-ai/playbook/duplicates-annoy

Installs the necessary Python packages for using Annoy and Renumics Spotlight. This includes 'renumics-spotlight', 'datasets', and 'annoy'. These libraries are essential for loading data, computing embeddings, and performing nearest neighbor searches.

```python
#@title Install required packages with PIP  

!pip install renumics-spotlight datasets annoy  
```

--------------------------------

### Load Pandas DataFrame and Show with Spotlight (Python)

Source: https://renumics.com/docs/getting-started

Loads a CSV file into a Pandas DataFrame and displays it using Spotlight's show function. Custom data types like Image and Embedding can be specified.

```python
import pandas as pd
from renumics import spotlight

df = pd.read_csv("https://renumics.com/data/mnist/mnist-tiny.csv")
spotlight.show(df, dtype={"image": spotlight.Image, "embedding": spotlight.Embedding})
```

--------------------------------

### Install Required Packages with PIP

Source: https://renumics.com/docs/data-centric-ai/playbook/label-errors-cleanlab

Installs the necessary Python packages: renumics-spotlight, cleanlab, and datasets. These are required for data analysis and visualization.

```python
#@title Install required packages with PIP  

!pip install renumics-spotlight cleanlab datasets  

```

--------------------------------

### Install Sliceguard and related packages

Source: https://renumics.com/docs/data-centric-ai/playbook/data-slices-sliceguard

Installs the necessary Python packages including renumics-spotlight, sliceguard, datasets, and cleanvision using pip. This is a prerequisite for using the provided code snippets.

```python
#@title Install required packages with PIP  

!pip install renumics-spotlight sliceguard datasets cleanvision  

```

--------------------------------

### Install Dependencies for F1 Data Analysis

Source: https://renumics.com/docs/use-cases/motorsports-telemetry

Installs the necessary Python libraries for loading F1 data and for visualization. This includes 'renumics-spotlight' for data visualization and 'fastf1' for accessing Formula 1 telemetry data.

```bash
pip install renumics-spotlight fastf1  

```

--------------------------------

### Explore Enriched Dataset with Custom Layout (Python)

Source: https://renumics.com/docs/getting-started

Displays an enriched Hugging Face dataset using Spotlight with a custom visualization layout for classification debugging. It specifies embedding columns and inspects audio data types.

```python
from renumics import spotlight

layout = spotlight.layouts.debug_classification(embedding='embedding', inspect={'audio': spotlight.dtypes.audio_dtype})
spotlight.show(ds, dtype={'embedding': spotlight.Embedding}, layout=layout )  
```

--------------------------------

### Load CSV File via CLI

Source: https://renumics.com/docs/loading-data

Loads a dataset from a CSV file using the Spotlight Command Line Interface (CLI). This is a straightforward way to start exploring tabular data.

```bash
spotlight mnist-tiny.csv
```

--------------------------------

### Concatenate Hugging Face Datasets (Python)

Source: https://renumics.com/docs/getting-started

Loads two datasets from the Hugging Face hub and concatenates them along axis 1. This is useful for combining raw data with enrichment results.

```python
import datasets

ds = datasets.load_dataset('speech_commands', 'v0.01', split='all')
ds_results = datasets.load_dataset('renumics/speech_commands-ast-finetuned-results', 'v0.01', split='all')
ds = datasets.concatenate_datasets([ds, ds_results], axis=1)
```

--------------------------------

### Load CIFAR-100 dataset and convert to Pandas DataFrame

Source: https://renumics.com/docs/data-centric-ai/playbook/data-slices-sliceguard

Loads the CIFAR-100 dataset from the Huggingface hub and converts it into a pandas DataFrame for further analysis. This is the first step in the example walkthrough.

```python
dataset = datasets.load_dataset("renumics/cifar100-enriched", split="test")
df = dataset.to_pandas()

```

--------------------------------

### Load and Show Hugging Face 'speech_commands' Dataset with Spotlight

Source: https://renumics.com/docs/loading-data/huggingface

Loads the 'speech_commands' dataset from Hugging Face and visualizes it using Renumics Spotlight. This requires the 'datasets' and 'renumics' libraries. It takes the dataset split as input and outputs an interactive Spotlight visualization.

```python
import datasets
from renumics import spotlight

ds = datasets.load_dataset('speech_commands', 'v0.01', split='validation')

spotlight.show(ds)
```

--------------------------------

### Load and View FSD50K Dataset with Spotlight

Source: https://renumics.com/docs/configure-visualizations/ui-components

This snippet shows how to download the 'fsd50k-tiny.csv' dataset using curl and then load it into Spotlight. It defines the data types for 'window', 'embedding', and 'audio' columns, allowing Spotlight to correctly interpret and visualize audio-related data.

```shell
curl https://spotlight.renumics.com/data/fsd50k/fsd50k-tiny.csv -o fsd50k-tiny.csv
spotlight fsd50k-tiny.csv --dtype window=Window --dtype embedding=Embedding --dtype audio=Audio
```

--------------------------------

### Append Data to HDF5 Dataset with Spotlight

Source: https://renumics.com/docs/loading-data/hdf5

Demonstrates appending data to an existing HDF5 dataset in append mode ('a'). It shows how to add columns with default values and optional columns, and how to set specific rows to None.

```python
from datetime import datetime

with spotlight.Dataset("example.h5", "w") as dataset:
    dataset.append_bool_column("boolean", [True, True, True, False], default=False)
    dataset.append_int_column("integer", range(4), default=-1)
    dataset.append_float_column("float", 1.0, optional=True)
    dataset.append_string_column("string", "foo", optional=True)
    dataset.append_categorical_column(
        "categorical", ["test", "train", "test", "train"], optional=True
    )
    dataset.append_datetime_column("datetime", datetime.now(), optional=True)

with spotlight.Dataset("example.h5", "a") as dataset:
    dataset[1] = {key: None for key in dataset.keys()}

spotlight.show("example.h5")
```

--------------------------------

### Load and Show Hugging Face 'cifar100' Dataset with Spotlight

Source: https://renumics.com/docs/loading-data/huggingface

Loads the 'cifar100' dataset from Hugging Face and displays it using Renumics Spotlight. This demonstrates a straightforward integration for image datasets. It requires the 'datasets' library and outputs an interactive visualization.

```python
ds = datasets.load_dataset('cifar100', split='test')
spotlight.show(ds)
```

--------------------------------

### Load and View MNIST Dataset with Spotlight

Source: https://renumics.com/docs/configure-visualizations/ui-components

This snippet demonstrates how to download the 'mnist-tiny.csv' dataset using curl and then load it into Spotlight for analysis. It specifies the data types for 'image', 'embedding', and 'label' columns, enabling specialized visualizations and interactions within Spotlight.

```shell
curl https://renumics.com/docs/data/mnist/mnist-tiny.csv -o mnist-tiny.csv
spotlight mnist-tiny.csv --dtype image=Image --dtype embedding=Embedding --dtype label=Category
```

--------------------------------

### Inspect Data Drift Candidates with RENUMICS Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/drift-kcore

Prepares a DataFrame for visualization by dropping unnecessary columns and then uses RENUMICS Spotlight to inspect potential data drift candidates. It loads a layout configuration from a URL and displays the data with specified data types and layout.

```python
import requests
import json
df_show = df.drop(columns=['embedding', 'probabilities'])  
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/drift_kcore.json"
response = requests.get(layout_url)
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)

```

--------------------------------

### Visualize Hugging Face Dataset with Spotlight

Source: https://renumics.com/docs/use-cases/motorsports-telemetry

Visualizes a Hugging Face Dataset using renumics-spotlight. This method loads data lazily, allowing for the processing and visualization of very large time series datasets without consuming significant memory.

```python
from renumics import spotlight  

# Assuming ds is a loaded Hugging Face Dataset
spotlight.show(ds)  

```

--------------------------------

### Build Layout with Python API - RENUMICS Spotlight

Source: https://renumics.com/docs/custom-visualizations/ui-components/layout

This Python code snippet demonstrates how to programmatically build a RENUMICS Spotlight layout using its Python API. It involves loading a pandas DataFrame, defining data types, and configuring various widgets like Table, SimilarityMap, and Histogram, along with an Inspector widget, to create a custom layout. Dependencies include pandas and renumics.spotlight.

```python
import pandas as pd
from renumics import spotlight
from renumics.spotlight import layout

df = pd.read_csv("https://spotlight.renumics.com/data/fsd50k/fsd50k-tiny.csv")
spotlight.show(
    df,
    dtype={"audio": spotlight.Audio, "embedding": spotlight.Embedding},
    layout=layout.layout(
        [
            [layout.table()],
            [
                layout.similaritymap(
                    columns=["embedding"],
                    color_by_column="annotation",
                    size_by_column="entropy",
                )
            ],
            [
                layout.histogram(
                    column="annotation", stack_by_column="prediction_incorrect"
                )
            ],
        ],
        layout.widgets.Inspector(),
    ),
)

```

--------------------------------

### Register TextLengthLens in index.ts

Source: https://renumics.com/docs/development/lenses

Shows how to register the newly created 'TextLengthLens' within the 'ALL_LENSES' array in the 'src/lenses/index.ts' file. This step is crucial for the RENUMICS platform to recognize and utilize the custom Lens.

```typescript
...  
import TextLengthLens from './TextLengthLens';  
  
export const ALL_LENSES = [  
    ...
    TextLengthLens,  
];  

```

--------------------------------

### Inspect Label Errors with Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/label-errors-cleanlab

Prepares data for inspection by dropping unnecessary columns and then uses Renumics Spotlight to visualize the dataset, highlighting potential label errors. It loads a layout configuration from a URL for the Spotlight display.

```python
import requests
import json
df_show = df.drop(columns=['embedding', 'probabilities'])
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/label_errors_cleanlab.json"
response = requests.get(layout_url)
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)

```

--------------------------------

### Inspect and Visualize Duplicates with Renumics Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/duplicates-annoy

Prepares the DataFrame for visualization by dropping embedding and probability columns, fetches a layout configuration from a URL, and then displays the data using Renumics Spotlight. This step allows for interactive inspection and identification of near-duplicate images.

```python
df_show = df.drop(columns=['embedding', 'probabilities'])  
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/duplicates_annoy.json"  
response = requests.get(layout_url)  
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))  
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)  
```

--------------------------------

### Write Image Dataset to HDF5 with Spotlight

Source: https://renumics.com/docs/loading-data/hdf5

Saves an image dataset to an HDF5 file using Spotlight. It includes columns for index, label, and image data. The image data is preprocessed for display.

```python
from sklearn import datasets
import numpy as np
from renumics import spotlight

OUTPUT_DATASET = "image_example_dataset.h5"

digits = datasets.load_digits()
with spotlight.Dataset(OUTPUT_DATASET, "w") as dataset:
    dataset.append_int_column("index", order=1)
    dataset.append_int_column("label", order=0)
    dataset.append_image_column("image")
    for i, (image, label) in enumerate(zip(digits.images, digits.target)):
        # in the sample dataset the value 0 means white and 16 means black
        # in order to display it correctly in the browser, we need to switch that
        # to have 0 as black and 255 as white
        image = (255 * (1 - image / 16)).round().astype("uint8")
        # scale image by 32 along each dimension in order to display it in the browser
        image = np.repeat(image, 32, axis=1)
        image = np.repeat(image, 32, axis=0)

        dataset.append_row(index=i, label=label, image=image)
```

--------------------------------

### Load CIFAR-100 Dataset and Convert to Pandas

Source: https://renumics.com/docs/data-centric-ai/playbook/label-errors-cleanlab

Loads the CIFAR-100 dataset from the Huggingface hub and converts it into a Pandas DataFrame. This is a preparatory step for further analysis.

```python
dataset = datasets.load_dataset("renumics/cifar100-enriched", split="train")  
df = dataset.to_pandas()  

```

--------------------------------

### Manually Specify Data Types via Python API

Source: https://renumics.com/docs/loading-data

Demonstrates how to manually specify data types for columns when loading a Pandas DataFrame with Spotlight. This allows for explicit control over how unstructured data like images and embeddings are interpreted.

```python
from renumics import spotlight

dtype = {"image": spotlight.Image, "embedding":spotlight.Embedding}
spotlight.show(df, dtype=dtype)
```

--------------------------------

### Perform EDA with Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/huggingface-embedding

Uses the renumics-spotlight library to visualize the processed data, including the reduced embeddings. It specifies the data types for 'image' and 'embedding_reduced' to enable interactive exploration.

```python
df_show = df.drop(columns=['embedding', 'probabilities'])
spotlight.show(df_show, port=port, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding})
```

--------------------------------

### Create Bing Image Dataset and Fine-tune ViT Model

Source: https://renumics.com/docs/use-cases/image-fine-tuning

This Python script generates an image dataset from Bing search results for specified class names, splits it into training and validation sets, fine-tunes a ViT model for image classification, and then enriches the dataset with model predictions and embeddings. Finally, it launches a Spotlight interface to visualize the results and identify clusters.

```python
# The Imports  
from renumics import spotlight  
from sliceguard.data import create_imagedataset_from_bing  
from sliceguard.models.huggingface import finetune_image_classifier, generate_image_pred_probs_embeddings  
from sliceguard.embeddings import generate_image_embeddings  
  
# Create an Image Dataset from Bing  
class_names = [  
    "Blue Tang",  
    "Clownfish",  
    "Spotted Eagle Ray",  
    "Longnose Butterfly Fish",  
    "Moorish Idol",  
    "Royal Gramma Fish",  
]  
df = create_imagedataset_from_bing(  
    class_names, 25, "data", test_split=0.2, license="Free to share and use"  
)  
  
# DataFrame Format:  
#+----+-------------------------------------+-------------------+---------+---------+  
#|    | image                               | label_str         |   label | split   |  
#|----+-------------------------------------+-------------------+---------+---------|  
#|  6 | data/Blue Tang/Image_15.jpg         | Blue Tang         |       0 | val     |  
#| 73 | data/Spotted Eagle Ray/Image_11.jpg | Spotted Eagle Ray |       2 | train   |  
#| 51 | data/Spotted Eagle Ray/Image_13.jpg | Spotted Eagle Ray |       2 | train   |  
#| 57 | data/Spotted Eagle Ray/Image_7.jpg  | Spotted Eagle Ray |       2 | val     |  
#| 31 | data/Clownfish/Image_10.jpg         | Clownfish         |       1 | train   |  
#| 27 | data/Clownfish/Image_13.jpg         | Clownfish         |       1 | train   |  
#+----+-------------------------------------+-------------------+---------+---------+  
  
# Fine-tune a ViT Model with the data (in 1-2 minutes on a GPU)  
finetune_image_classifier(  
    df[df["split"] == "train"],  
    model_name="google/vit-base-patch16-224-in21k",  
    output_model_folder="./model_folder",  
    epochs=15,  
)  
  
# Enrich the DataFrame with Predictions, Probabilities and Embeddings  
df["prediction"], df["probs"], df["embeddings"] = generate_image_pred_probs_embeddings(  
    df["image"].values, model_name="./model_folder"  
)  
# Check the result and detect problematic clusters  
spotlight.show(  
    df, layout="https://spotlight.renumics.com/resources/image_classification_v1.0.json"  
)
```

--------------------------------

### Display HDF5 Dataset with Spotlight

Source: https://renumics.com/docs/loading-data/hdf5

Loads and displays an HDF5 dataset using the Spotlight API. This function can be used to view the data stored in the HDF5 file.

```python
spotlight.show(OUTPUT_DATASET)
```

--------------------------------

### Load CIFAR-100 Dataset and Convert to Pandas DataFrame

Source: https://renumics.com/docs/data-centric-ai/playbook/duplicates-annoy

Loads the CIFAR-100 dataset from the Huggingface hub using the 'datasets' library and converts it into a Pandas DataFrame. This is the first step in preparing the data for nearest neighbor analysis.

```python
dataset = datasets.load_dataset("renumics/cifar100-enriched", split="train")  
df = dataset.to_pandas()  
```

--------------------------------

### Write Mixed Data Types to HDF5 with Spotlight

Source: https://renumics.com/docs/loading-data/hdf5

Creates an HDF5 file containing various data types: boolean, integer, float, string, categorical, and datetime. The dataset is then displayed.

```python
from datetime import datetime

with spotlight.Dataset("example.h5", "w") as dataset:
    dataset.append_bool_column("boolean", [True, False, False, True])
    dataset.append_int_column("integer", range(4))
    dataset.append_float_column("float", 1.0)
    dataset.append_string_column("string", "foo")
    dataset.append_categorical_column(
        "categorical", ["test", "train", "test", "train"]
    )
    dataset.append_datetime_column("datetime", datetime.now())

spotlight.show("example.h5")
```

--------------------------------

### Inspect and Detect Problematic Data Segments with Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/cv-issues

This code demonstrates how to inspect errors and detect problematic data segments using Spotlight. It prepares the DataFrame, fetches a layout configuration, and then displays the data with specified image and embedding data types.

```python
df_show = df.drop(columns=['embedding', 'probabilities'])  
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/cv_issues.json"  
response = requests.get(layout_url)  
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))  
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)  

```

--------------------------------

### Inspect Outliers with Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/outliers-cleanlab

Prepares the DataFrame for visualization by removing 'embedding' and 'probabilities' columns. It then fetches a layout configuration from a URL and uses `renumics-spotlight.show` to display the data, specifying data types for 'image' and 'embedding_reduced', and applying the fetched layout for interactive outlier inspection.

```python
df_show = df.drop(columns=['embedding', 'probabilities'])  
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/outlier_cleanlab.json"  
response = requests.get(layout_url)  
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))  
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)  

```

--------------------------------

### Detect and Visualize Audio Classification Dataset Issues with Python

Source: https://renumics.com/docs/use-cases/audio-classification

This Python script uses SliceGuard to detect issues in an audio classification dataset loaded from Hugging Face. It then visualizes these issues using Spotlight. The script requires the renumics-spotlight, sliceguard, and scikit-learn libraries.

```python
# The Imports
from renumics import spotlight
from sliceguard import SliceGuard
from sliceguard.data import from_huggingface
from sklearn.metrics import accuracy_score

# Load an Example Dataset as DataFrame
df = from_huggingface("renumics/emodb")

# DataFrame Format:
# +---------------------+---------+
# |        audio        | emotion |
# +---------------------+---------+
# | /path/to/audio1.wav | joy     |
# | /path/to/audio2.wav | anger   |
# | /path/to/audio3.wav | joy     |
# | ...                 |         |
# +---------------------+---------+

# Detect Issues Using sliceguard
sg = SliceGuard()
issues = sg.find_issues(df, features=["audio"], y="emotion", metric=accuracy_score)
report_df, spotlight_data_issues, spotlight_dtypes, spotlight_layout = sg.report(
    no_browser=True
)

# Visualize Detected Issues in Spotlight:
spotlight.show(
    report_df,
    dtype=spotlight_dtypes,
    issues=spotlight_data_issues,
    layout=spotlight_layout,
)

```

--------------------------------

### Load Hugging Face Dataset with Memory Mapping

Source: https://renumics.com/docs/use-cases/motorsports-telemetry

Loads a Hugging Face Dataset from disk, which is stored in Arrow format. This process utilizes memory mapping, resulting in negligible RAM usage for loading, making it highly efficient for large datasets.

```python
import datasets  
import psutil  
  
print('memory used in MB:', psutil.virtual_memory()[3]/ 1024 ** 2)  
ds = datasets.Dataset.load_from_disk('f1telemetry')  
print('memory used in MB:', psutil.virtual_memory()[3]/ 1024 ** 2)  

```

--------------------------------

### Detect and Visualize Dataset Issues with SliceGuard and Spotlight (Python)

Source: https://renumics.com/docs/use-cases/image-classification

Loads an image classification dataset from Hugging Face, detects problematic data clusters using SliceGuard, and visualizes the findings with Spotlight. Requires a DataFrame with 'image' paths and 'label' columns. Outputs a report and Spotlight visualization.

```python
# The Imports
from renumics import spotlight
from sliceguard import SliceGuard
from sliceguard.data import from_huggingface
from sklearn.metrics import accuracy_score

# Load an Example Dataset as DataFrame
df = from_huggingface("Matthijs/snacks")

# DataFrame Format:
# +-------------------+---------+
# |       image       |  label  |
# +-------------------+---------+
# | /path/to/img1.png | popcorn |
# | /path/to/img2.png | muffin  |
# | /path/to/img3.png | cake    |
# | ...               |         |
# +-------------------+---------+

# Detect Issues Using sliceguard
sg = SliceGuard()
issues = sg.find_issues(df, features=["image"], y="label", metric=accuracy_score)
report_df, spotlight_data_issues, spotlight_dtypes, spotlight_layout = sg.report(
    no_browser=True
)

# Visualize Detected Issues in Spotlight:
spotlight.show(
    report_df,
    dtype=spotlight_dtypes,
    issues=spotlight_data_issues,
    layout=spotlight_layout,
)

```

--------------------------------

### Load CIFAR-100 Dataset

Source: https://renumics.com/docs/data-centric-ai/playbook/huggingface-embedding

Loads the CIFAR-100 dataset from the Huggingface hub and converts it into a Pandas DataFrame for further processing. This step prepares the data for embedding computation.

```python
dataset = datasets.load_dataset("renumics/cifar100-enriched", split="train")
df = dataset.to_pandas()
```

--------------------------------

### Concatenate and Visualize Data with Spotlight

Source: https://renumics.com/docs/use-cases/motorsports-telemetry

Combines lap metadata with extracted telemetry data into a single pandas DataFrame and then visualizes this combined dataset using the renumics-spotlight library. This allows for interactive exploration of the F1 race data.

```python
from renumics import spotlight  
import pandas as pd

# Assuming df_metadata and df_telemetry are already defined from previous steps
df_metadata = pd.DataFrame(laps)
df = pd.concat([df_metadata, df_telemetry], axis=1)  
  
spotlight.show(df)  

```

--------------------------------

### Manually Assign Data Types for Spotlight Visualization

Source: https://renumics.com/docs/loading-data/huggingface

Manually assigns data types for image and categorical labels when they are ambiguous or not specified in the Hugging Face dataset. This allows Spotlight to correctly interpret and visualize specific features like images and fine labels. It takes the dataset and a dictionary of data type mappings as input.

```python
label_mapping = dict(zip(ds.features['fine_label'].names, range(len(ds.features['fine_label'].names))))
spotlight.show(ds, dtype={'img': spotlight.Image, 'fine_label': spotlight.dtypes.CategoryDType(categories=label_mapping)})
```

--------------------------------

### Specify Categorical Data Type with Mapping

Source: https://renumics.com/docs/loading-data/data-types

Shows how to use the `CategoryDType` for categorical variables in RENUMICS Spotlight, including the ability to provide a label mapping. This is particularly useful for assigning numerical representations to string labels.

```python
label_mapping = {'cat': 0, 'dog': 1, 'mouse': 2}
spotlight.show(ds, dtype={'img': spotlight.Image, 'label': spotlight.dtypes.CategoryDType(categories=label_mapping)})
```

--------------------------------

### Style TextLengthLens with Tailwind CSS and twin.macro

Source: https://renumics.com/docs/development/lenses

Enhances the 'TextLengthLens' component by applying styling using tailwindcss utility classes through twin.macro. It adds basic styling to the displayed text length, including font size, weight, and padding.

```typescript
import 'twin.macro';  
import { Lens } from '../types';  
  
const TextLengthLens: Lens = ({ value }) => {  
    return <div tw="text-sm font-semibold px-1 py-1">{`${value}`.length}</div>;  
};  
...  

```

--------------------------------

### Inspect Decision Boundary with Spotlight

Source: https://renumics.com/docs/data-centric-ai/playbook/decision-boundary

Visualizes the data with computed decision boundary scores using the renumics-spotlight library. It drops unnecessary columns, fetches a layout configuration, and then displays the DataFrame with specified data types and layout.

```python
df_show = df.drop(columns=['embedding', 'probabilities'])  
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/decision_boundary_layout.json"  
response = requests.get(layout_url)  
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))  
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)  

```

--------------------------------

### Create TextLengthLens Component in TypeScript

Source: https://renumics.com/docs/development/lenses

Defines a basic React Lens component named 'TextLengthLens' using TypeScript. It specifies metadata like key, data types, and dimensions. This Lens is intended to be registered and used within the RENUMICS platform.

```typescript
import { Lens } from "../types";  
  
const TextLengthLens: Lens = () => {  
    return <>TextLengthView</>;  
};

TextLengthLens.key = "TextLengthView";  
TextLengthLens.dataTypes = [  
    "int",  
    "float",  
    "bool",  
    "str",  
    "datetime",  
    "Category",  
];  
TextLengthLens.defaultHeight = 22;  
TextLengthLens.minHeight = 22;  
TextLengthLens.maxHeight = 64;  
TextLengthLens.displayName = "Text Length";  
  
export default TextLengthLens;  

```

--------------------------------

### Load F1 Race Session Data with FastF1

Source: https://renumics.com/docs/use-cases/motorsports-telemetry

Loads a specific Formula 1 race session's data, including telemetry, laps, and other relevant information, using the fastf1 library. This function retrieves session details for a given year, location, and session type.

```python
import fastf1  
  
session = fastf1.get_session(2023, 'Montreal', 'Race')  
  
session.load(telemetry=True, laps=True)  
  
laps = session.laps  

```

--------------------------------

### Add PCA Embeddings to HDF5 Dataset with Spotlight

Source: https://renumics.com/docs/loading-data/hdf5

Appends Principal Component Analysis (PCA) embeddings to an existing HDF5 dataset. This is done by opening the dataset in append mode ('a').

```python
from sklearn.decomposition import PCA

pca_embeddings = PCA(8).fit_transform(digits.data)
with spotlight.Dataset(OUTPUT_DATASET, "a") as dataset:
    dataset.append_embedding_column("pca", pca_embeddings)
```

--------------------------------

### Detect and Visualize Text Classification Dataset Issues with SliceGuard and Spotlight (Python)

Source: https://renumics.com/docs/use-cases/text-classification

Loads a dataset from Hugging Face, detects problematic clusters using SliceGuard based on text features and labels, and visualizes these issues using Spotlight. It requires the 'renumics-spotlight', 'sliceguard', and 'scikit-learn' libraries.

```python
# The Imports
from renumics import spotlight
from sliceguard import SliceGuard
from sliceguard.data import from_huggingface
from sklearn.metrics import accuracy_score

# Load an Example Dataset as DataFrame
df = from_huggingface("dair-ai/emotion")

# DataFrame Format:
# +-------+-------+
# | text  | label |
# +-------+-------+  
# | text1 | joy   |
# | text2 | anger |
# | text3 | joy   |
# | ...   |       |
# +-------+-------+

# Detect Issues Using sliceguard
sg = SliceGuard()
issues = sg.find_issues(df, features=["text"], y="label", metric=accuracy_score)
report_df, spotlight_data_issues, spotlight_dtypes, spotlight_layout = sg.report(
    no_browser=True
)

# Visualize Detected Issues in Spotlight:
spotlight.show(
    report_df,
    dtype=spotlight_dtypes,
    issues=spotlight_data_issues,
    layout=spotlight_layout,
)

```

--------------------------------

### Compute and Concatenate k-Nearest Neighbor Distances

Source: https://renumics.com/docs/data-centric-ai/playbook/drift-kcore

Computes the k-nearest neighbor distances for the loaded CIFAR-100 dataset using the `compute_k_core_distances` function and then concatenates the resulting distances and indices back to the original DataFrame.

```python
df_kcore = compute_k_core_distances(df)  
df = pd.concat([df, df_kcore], axis=1)  

```

--------------------------------

### Load Pandas DataFrame via Python API

Source: https://renumics.com/docs/loading-data

Loads an in-memory Pandas DataFrame for interactive exploration using the Spotlight Python API. This is particularly useful within notebook environments.

```python
from renumics import spotlight

spotlight.show(df)
```

--------------------------------

### Inspect and Remove Leakage with Spotlight (Python)

Source: https://renumics.com/docs/data-centric-ai/playbook/leakage-annoy

Visualizes the data and identifies leakage using renumics-spotlight. It drops unnecessary columns, fetches a layout configuration from a URL, and then displays the DataFrame with specified data types and layout.

```python
df_show = df.drop(columns=['embedding', 'probabilities'])  
layout_url = "https://raw.githubusercontent.com/Renumics/spotlight/playbook_initial_draft/playbook/rookie/leakage_annoy.json"  
response = requests.get(layout_url)  
layout = spotlight.layout.nodes.Layout(**json.loads(response.text))  
spotlight.show(df_show, dtype={"image": spotlight.Image, "embedding_reduced": spotlight.Embedding}, layout=layout)  
```
Tokens:
Raw