JoinDiscord

Notebooks

The notebook reference guide for Argilla tutorials.

This section contains:

Notebooks

• 🔐 Backup and version Argilla Datasets using DVC
  • Introduction
  • Running Argilla
  • Setup packages
  • Configure .git
  • Configure DVC
  • Define Background Process
  • Retrieve data versions
  • Summary
• 🚀 Run Argilla with a Transformer in an active learning loop and a free GPU in your browser
  • Initial setup on Google Colab
  • Install Elastic Search
  • Start the Argilla localhost in a terminal
  • Create a public link to Argilla localhost with ngrok
  • Log data to argilla and start your active learning loop with small-text
  • Start annotating in the browser via the ngrok link
  • Extract annotated data for downstream use
  • Summary
• 💾 Monitor FastAPI model endpoints
  • Introduction
  • Running Argilla
  • Setup
  • 1. Loading models
  • 2. Convert output to Argilla format
  • 3. Create prediction endpoint
  • 4. Add Argilla logging middleware to the application
  • 5. NER endpoint with spaCy
  • Launching and testing the API
  • Summary
  • Apendix: main.py full code
• 🧸 Using LLMs for Text Classification and Summarization Suggestions with spacy-llm
  • Introduction
  • Running Argilla
  • Setup
  • spacy-llm pipeline
  • Load Data
  • FeedbackDataset
• 🗺️ Add bias-equality features to datasets with disaggregators
  • Introduction
  • Running Argilla
  • Setup
  • 📰 Load news summary data
  • 🗺️ Apply disaggregator features
  • 📊 Analyze potential biases
  • Summary
• 💡 Build and evaluate a zero-shot sentiment classifier with GPT-3
  • Running Argilla
  • Setup
  • 0. The dataset: banking77 sentiment dataset
  • 1. Building a zero-shot sentiment classifier with GPT-3
  • 2. Evaluating our zero-shot sentiment classifier
• 💨 Label data with semantic search and Sentence Transformers
  • Introduction
  • Running Argilla
  • Setup
  • 💾 Downloading and embedding your dataset
  • 🙃 Upload dataset into Argilla
  • 🏷 Bulk labeling with the find similar action
  • Summary
• 📸 Bulk Labeling Multimodal Data
  • Running Argilla
  • Setup
  • Introduction
  • A ‘real-world’ multimodal dataset
  • 🔫 Zero-Shot Classification
  • Consolidate data labeling
  • Bulk Labeling with embeddings
  • Upload to Argilla
  • Fewshot Classification
  • Summary
• 🧱 Augment weak supervision rules with Sentence Transformers
  • Introduction
  • Running Argilla
  • Setup
  • Detailed Workflow
  • The dataset
  • 1. Create an Argilla dataset with unlabelled data and test data
  • 2. Defining rules
  • 3. Building and analyzing weak labels
  • 4. Using the weak labels
  • 5. Extending the weak labels
  • 6. Training a downstream model
  • Summary
  • Appendix: Visualize changes
  • Appendix: Optimizing the thresholds
  • Appendix: Plot extension
• 🔫 Zero-shot and few-shot classification with SetFit
  • Introduction
  • Running Argilla
  • Setup
  • 💾 Embedding your dataset
  • 🔫 Zero-shot predictions with SetFit
  • 🏷 Bulk labeling with the find similar and zero-shot predictions
  • 📏 Evaluate the zero-shot model
  • 🦾 Train a few-shot SetFit model
  • Summary
• 🗂 Multi-label text classification with weak supervision
  • Running Argilla
  • Setup
  • GoEmotions
  • Research topic dataset
  • Summary
  • Appendix A
  • Appendix B
• 📰 Train a text classifier with weak supervision
  • Introduction
  • Running Argilla
  • Setup
  • 1. Load test and unlabelled datasets into Argilla
  • 2. Define Rules
  • 3. Denoise weak labels with Snorkel’s Label Model
  • 4. Prepare our training set
  • 5. Train a downstream model with scikit-learn
  • Summary
  • Appendix I: Log datasets to the Hugging Face Hub
• 🗂️ Assign records to your annotation team
  • Introduction
  • Running Argilla
  • Setup
  • 📥 Loading a dataset
  • 🔀 Make a random assignment
  • 💾 Log your dataset and assignments in Argilla
  • ✍️ Annotate records assigned to you
  • Summary
• 🩹 Delete labels from a Token or Text Classification dataset
  • Setup
  • First steps
  • Remove the unwanted label from the records
  • Update dataset settings
  • Summary
• 🔫 Evaluate a zero-shot NER with Flair
  • Introduction
  • Running Argilla
  • Setup
  • 1. Load the wnut_17 dataset
  • 2. Configure Flair TARSTagger
  • 3. Predict over wnut_17 and log into argilla
  • Summary
• 🐭 Train a NER model with skweak
  • Introduction
  • Running Argilla
  • Setup
  • 1. Log the dataset into Argilla
  • 2. Use Argilla to write skweak heuristic rules
  • 3. Evaluate the precision of our rules
  • 4. Annotate the training data and aggregate the weak labels
  • 5. Evaluate the baseline
  • 6. Train and evaluate our model
  • Summary
• 💫 Explore and analyze spaCy NER predictions
  • Introduction
  • Running Argilla
  • Our dataset
  • Logging spaCy NER entities into Argilla
  • Exploring and comparing en_core_web_sm and en_core_web_trf models
  • Explore the IMDB dataset
  • Summary
  • Appendix: Log datasets to the Hugging Face Hub
• 🔗 Using LLMs for Few-Shot Token Classification Suggestions with spacy-llm
  • Introduction
  • Running Argilla
  • Setup
  • Load dataset
  • spacy-llm pipeline
  • Dataset
• 🧐 Find label errors with cleanlab
  • Introduction
  • Running Argilla
  • 1. Load datasets
  • 2. Train model
  • 3. Get label error candidates
  • 4. Uncover and correct label errors
  • Summary
  • Appendix I: Find label errors in your train data using cross-validation
  • Appendix II: Log datasets to the Hugging Face Hub
• 🥇 Compare Text Classification Models
  • Introduction
  • Running Argilla
  • Setup
  • Make Zero Shot Text Classification predictions using transformers
  • Make zero-shot text classification using a trained SetFit classifier
• 🕵️‍♀️ Analyze predictions with explainability methods
  • Running Argilla
  • Token attributions and what do highlight colors mean?
  • Using Transformers Interpret
  • Using Shap
• 🧼 Clean labels using your model’s loss
  • Introduction
  • Running Argilla
  • Preliminaries
  • 1. Load the fine-tuned model and the training dataset
  • 2. Computing the loss
  • 3. Log high-loss examples into Argilla
  • 4. Using Argilla UI for inspection and relabeling
  • 5. Sharing the dataset in the Hugging Face Hub
  • Summary
• 🤔 Fine-tunning a NER model with BERT for Beginners
  • Introduction
  • Running Argilla
  • Setup
  • 🚀 Exploring our dataset
  • ⏳ Preprocessing the data
  • 🔍 Fine-tunning the model
  • 📝✔️ Summary
• Text classification active learning with classy-classification
  • Introduction
  • Running Argilla
  • 💿 Load data into Argilla
  • ⏱ Train a few-shot classifier
  • 🕵🏽‍♂️ Active learning heuristics
  • 🔁 Set up an active learning loop
  • 🎥 A live demo video
  • Summary
• 🤔 Text Classification active learning with ModAL
  • Introduction
  • Running Argilla
  • 1. Loading and preparing data
  • 2. Defining our classifier and Active Learner
  • 3. Active Learning loop
  • Summary
  • Bonus: Compare query strategies, random vs max uncertainty
  • Appendix: How did we obtain the train/test data?
• 🤯 Few-shot classification with SetFit
  • Summary
  • Running Argilla
  • Load unlabelled dataset in Argilla
  • Manual labeling
  • Train and evaluate SetFit model
  • Conclusion
• 🤗 Train a sentiment classifier with SetFit
  • Introduction
  • Running Argilla
  • Source dataset: banking77
  • 1. Load the dataset and label a few examples
  • 2. Hand labeling
  • 3. Train our SetFit sentiment classifier
  • Summary
• 👂 Text Classification: Active Learning with small-text
  • Introduction
  • Running Argilla
  • The TREC dataset
  • Preprocessing the dataset
  • Setting up the active learner
  • Argilla and you: the perfect oracle
  • Starting the active learning loop
  • Can we stop?
  • Summary
• 🏷️ Fine-tune a sentiment classifier with your own data
  • Introduction
  • Running Argilla
  • Preliminaries
  • 1. Run the pre-trained model over the dataset and log the predictions
  • 2. Explore and label data with the pre-trained model
  • 3. Fine-tune the pre-trained model
  • 4. Testing the fine-tuned model
  • 5. Run our fine-tuned model over the dataset and log the predictions
  • 6. Explore and label data with the fine-tuned model
  • 7. Fine-tuning with the extended training dataset
  • Summary
• 🕸️ Train a summarization model with Unstructured and Transformers
  • Introduction
  • Running Argilla
  • Section 1: Data Collection and Staging with unstructured
  • Section 2: Label Verification with argilla
  • Section 3: Model Training with transformers