{"passport":{"unfragile":{"@version":"1.0","version":"2026-05","artifact":{"id":"awesome-practical-deep-learning-for-coders-fast-ai","slug":"practical-deep-learning-for-coders-fast-ai","name":"Practical Deep Learning for Coders - fast.ai","type":"product","url":"https://course.fast.ai/","page_url":"https://unfragile.ai/practical-deep-learning-for-coders-fast-ai","categories":["productivity"],"tags":[],"pricing":{"model":"unknown","free":false,"starting_price":null},"status":"inactive","verified":false},"capabilities":[{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_0","uri":"capability://text.generation.language.top.down.deep.learning.curriculum.with.practical.first.pedagogy","name":"top-down deep learning curriculum with practical-first pedagogy","description":"Teaches deep learning by starting with high-level applications (image classification, NLP) and progressively revealing underlying mathematics and theory, rather than bottom-up linear algebra foundations. Uses Jupyter notebooks embedded in the course platform to interleave video lectures, code examples, and interactive exercises in a single learning context. The curriculum is structured around real datasets and competitions (ImageNet, MNIST variants) to anchor abstract concepts in concrete problems.","intents":["I want to learn deep learning without spending months on linear algebra prerequisites","I need to understand how to apply neural networks to real problems before diving into theory","I want to see working code examples alongside explanations in the same environment"],"best_for":["software developers transitioning into ML/AI with existing programming skills","practitioners who learn better from examples than from mathematical formalism","teams building production ML systems who need rapid prototyping skills"],"limitations":["Assumes Python programming competency — not suitable for non-programmers","Heavy focus on computer vision and NLP; limited coverage of reinforcement learning or time-series forecasting","Course material is updated periodically; older versions may reference deprecated library APIs"],"requires":["Python 3.7+","Jupyter notebook environment (local or cloud-based like Kaggle/Colab)","GPU access recommended but not required for initial lessons","fastai library (PyTorch-based, installed via pip)"],"input_types":["video lectures","Jupyter notebooks with executable code cells","datasets (CSV, images, text)","pre-trained model weights"],"output_types":["trained neural network models","predictions on new data","visualizations of learned features","code implementations of architectures"],"categories":["text-generation-language","educational-content","curriculum-design"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_1","uri":"capability://image.visual.transfer.learning.based.image.classification.with.minimal.data","name":"transfer-learning-based image classification with minimal data","description":"Teaches and provides code patterns for leveraging pre-trained convolutional neural networks (ResNet, EfficientNet, Vision Transformers) trained on ImageNet, then fine-tuning only the final layers on custom datasets with as few as 10-100 images per class. The fastai library implements discriminative learning rates (lower learning rates for early layers, higher for later layers) and progressive unfreezing to stabilize training on small datasets. Includes techniques like data augmentation and learning rate scheduling to prevent overfitting.","intents":["I need to build an image classifier for a custom domain (medical images, product defects) but only have 50 labeled examples","I want to avoid training a CNN from scratch which requires massive datasets and GPU time","I need to understand which layers of a pre-trained model to freeze vs fine-tune"],"best_for":["startups and small teams with limited labeled data budgets","domain experts (radiologists, quality engineers) building specialized classifiers","rapid prototyping scenarios where model accuracy needs to be validated in days, not months"],"limitations":["Transfer learning assumes source domain (ImageNet) is sufficiently similar to target domain; fails for highly specialized domains like microscopy or satellite imagery without domain-specific pre-trained models","Fine-tuning on very small datasets (<50 images) still risks overfitting despite regularization techniques","Requires GPU for reasonable training speed; CPU training on large images is prohibitively slow"],"requires":["Python 3.7+","fastai library (built on PyTorch)","GPU with 4GB+ VRAM (NVIDIA, Apple Silicon, or cloud GPU)","labeled image dataset (minimum 10-20 images per class for proof-of-concept)"],"input_types":["image files (JPEG, PNG, TIFF)","directory structure with class labels as folder names","CSV with image paths and labels"],"output_types":["trained PyTorch model checkpoint","class predictions with confidence scores","confusion matrix and per-class metrics","feature visualizations from intermediate layers"],"categories":["image-visual","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_10","uri":"capability://data.processing.analysis.dataset.creation.and.annotation.workflows","name":"dataset creation and annotation workflows","description":"Teaches best practices for creating high-quality training datasets, including data collection strategies, annotation guidelines, and quality control. Covers how to use annotation tools (LabelImg, CVAT, Prodigy), manage annotation workflows with multiple annotators, and measure inter-annotator agreement. Discusses the importance of dataset diversity, handling class imbalance, and avoiding common pitfalls like data leakage. Includes practical guidance on data augmentation to increase effective dataset size.","intents":["I need to create a labeled dataset for a custom computer vision or NLP task","I want to set up an annotation workflow with multiple annotators and quality control","I need to understand how to measure and improve dataset quality"],"best_for":["teams building custom ML systems who need to create labeled datasets","data scientists managing annotation workflows and quality assurance","practitioners learning that dataset quality is often more important than model architecture"],"limitations":["Annotation is labor-intensive and expensive; no shortcut for creating high-quality labels","Course provides guidance but not automated annotation tools; requires external services or crowdsourcing platforms","Inter-annotator agreement metrics are discussed but not deeply covered; requires domain expertise to set meaningful thresholds","No coverage of active learning or semi-supervised learning for reducing annotation burden"],"requires":["Raw data (images, text, or other modalities)","Annotation tool (LabelImg, CVAT, Prodigy, or custom)","Annotation guidelines and quality control process","Budget for annotation (crowdsourcing or in-house labor)"],"input_types":["raw images, text, or other data","annotation guidelines (text or video)","reference examples"],"output_types":["labeled dataset (images with bounding boxes, text with tags, etc.)","annotation statistics (inter-annotator agreement, label distribution)","quality control reports"],"categories":["data-processing-analysis","automation-workflow"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_11","uri":"capability://planning.reasoning.learning.rate.scheduling.and.hyperparameter.optimization","name":"learning rate scheduling and hyperparameter optimization","description":"Teaches how to select learning rates and other hyperparameters to train deep learning models effectively. Covers the learning rate finder (plotting loss vs learning rate to identify optimal ranges), learning rate schedules (constant, step decay, cosine annealing), and momentum/weight decay tuning. Includes techniques like discriminative learning rates (different rates for different layers) and cyclical learning rates. Discusses the relationship between batch size, learning rate, and convergence speed.","intents":["I want to train my model faster and achieve better accuracy by tuning the learning rate","I need to understand how to use the learning rate finder to select a good starting learning rate","I want to know how to adjust hyperparameters when my model is overfitting or underfitting"],"best_for":["practitioners training deep learning models who want to improve convergence speed and final accuracy","teams with limited GPU budgets who need to maximize training efficiency","learners understanding the relationship between hyperparameters and model performance"],"limitations":["Optimal hyperparameters are dataset and model-dependent; no universal rules apply across all problems","Learning rate finder assumes a specific training loop structure; may not work with custom training code","Hyperparameter optimization is still somewhat of an art; requires experimentation and intuition","No coverage of Bayesian optimization or other automated hyperparameter search methods"],"requires":["Python 3.7+","fastai library with learning rate finder","Trained model and dataset","GPU recommended for fast iteration"],"input_types":["training dataset","model architecture","initial hyperparameter guesses"],"output_types":["learning rate finder plot (loss vs learning rate)","optimal learning rate range","trained model with tuned hyperparameters","training curves showing convergence"],"categories":["planning-reasoning","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_2","uri":"capability://text.generation.language.natural.language.processing.with.pre.trained.language.models.and.fine.tuning","name":"natural language processing with pre-trained language models and fine-tuning","description":"Teaches NLP using transfer learning with pre-trained language models (ULMFiT, BERT-style architectures) for tasks like text classification, sentiment analysis, and named entity recognition. The course covers the Universal Language Model Fine-tuning (ULMFiT) approach: pre-train on general text corpus, fine-tune on task-specific corpus, then fine-tune on labeled data. Includes practical patterns for handling variable-length sequences, building custom tokenizers, and interpreting model predictions via attention weights.","intents":["I need to classify customer support tickets or product reviews but only have 500 labeled examples","I want to understand how language models transfer knowledge from general text to domain-specific tasks","I need to extract entities or relationships from unstructured text without building a rule-based system"],"best_for":["NLP practitioners building production text classifiers with limited labeled data","teams migrating from rule-based NLP (regex, keyword matching) to learned models","researchers prototyping new NLP architectures without implementing from scratch"],"limitations":["Course material predates transformer explosion (2018-2019); BERT and GPT coverage is limited compared to modern LLM courses","Fine-tuning approach assumes access to unlabeled domain-specific text for intermediate fine-tuning step; not all domains have sufficient unlabeled corpora","No coverage of prompt engineering, in-context learning, or instruction-tuned models (GPT-3 style)"],"requires":["Python 3.7+","fastai NLP module","Pre-trained language model weights (provided or downloaded from Hugging Face)","GPU with 4GB+ VRAM for training; 2GB+ for inference","Text dataset (unlabeled for pre-training, labeled for fine-tuning)"],"input_types":["plain text files or CSV with text column","pre-tokenized sequences","raw documents (PDFs, web pages) with preprocessing"],"output_types":["class predictions with confidence scores","attention weight visualizations","token-level predictions (NER tags)","fine-tuned model checkpoint"],"categories":["text-generation-language","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_3","uri":"capability://data.processing.analysis.collaborative.filtering.and.recommendation.systems.with.matrix.factorization","name":"collaborative filtering and recommendation systems with matrix factorization","description":"Teaches recommendation systems using collaborative filtering, specifically matrix factorization with embeddings. The approach learns latent representations for users and items by factorizing the user-item interaction matrix, then predicts ratings or rankings by computing dot products of learned embeddings. The course covers both explicit feedback (ratings) and implicit feedback (clicks, purchases), regularization techniques to prevent overfitting, and how to handle cold-start problems with content-based fallbacks.","intents":["I need to build a movie/product recommendation system but don't have explicit ratings, only user behavior","I want to understand how Netflix and Spotify generate personalized recommendations at scale","I need to recommend items to new users who have no interaction history"],"best_for":["e-commerce platforms and content platforms building personalization features","teams with user-item interaction data but limited content metadata","practitioners learning collaborative filtering before moving to deep learning-based recommenders"],"limitations":["Matrix factorization assumes static user preferences; doesn't capture temporal dynamics or concept drift","Cold-start problem (new users/items with no interactions) requires hybrid approaches not fully covered","Doesn't scale to billions of users/items without distributed training infrastructure","No coverage of modern deep learning recommenders (neural collaborative filtering, attention-based models)"],"requires":["Python 3.7+","fastai library","User-item interaction data (ratings, clicks, purchases) in sparse matrix format","GPU optional but recommended for large matrices (>1M users or items)"],"input_types":["user-item interaction matrix (sparse, CSR format)","CSV with columns: user_id, item_id, rating/interaction","implicit feedback (binary: clicked/not clicked)"],"output_types":["predicted ratings or ranking scores","top-N recommendations per user","user and item embedding vectors","similarity scores between items"],"categories":["data-processing-analysis","memory-knowledge"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_4","uri":"capability://data.processing.analysis.structured.data.modeling.with.embeddings.and.tabular.neural.networks","name":"structured data modeling with embeddings and tabular neural networks","description":"Teaches how to apply deep learning to tabular/structured data (CSV files with mixed categorical and continuous features) using entity embeddings and shallow neural networks. The approach learns dense vector representations for categorical variables (like country, product category) rather than one-hot encoding, then concatenates embeddings with continuous features and passes through a small MLP. Includes techniques for handling missing values, feature scaling, and regularization via dropout and batch normalization.","intents":["I have a CSV with customer demographics, purchase history, and want to predict churn or lifetime value","I want to move beyond XGBoost/random forests for tabular data and leverage deep learning","I need to learn representations of categorical features (e.g., city embeddings) that capture similarity"],"best_for":["data scientists working with business/financial datasets in CSV format","teams building predictive models for structured data where tree-based models plateau","practitioners learning how embeddings work before applying them to NLP or recommendations"],"limitations":["Deep learning on tabular data often underperforms well-tuned XGBoost or LightGBM on small datasets (<100K rows)","Requires careful hyperparameter tuning (embedding dimensions, network depth, dropout rates) to avoid overfitting","No built-in support for time-series features or temporal dependencies in tabular data","Interpretability is harder than tree-based models; feature importance requires additional techniques"],"requires":["Python 3.7+","fastai tabular module","Pandas for data loading and preprocessing","CSV or database table with mixed categorical and continuous columns","GPU optional; CPU training is feasible for datasets <1M rows"],"input_types":["CSV files with mixed data types","Pandas DataFrames","SQL queries returning tabular data"],"output_types":["class predictions (classification) or continuous predictions (regression)","learned embedding vectors for categorical features","feature importance estimates","model checkpoint for inference"],"categories":["data-processing-analysis","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_5","uri":"capability://image.visual.generative.modeling.with.gans.and.diffusion.models","name":"generative modeling with gans and diffusion models","description":"Teaches generative deep learning using Generative Adversarial Networks (GANs) and diffusion models for image synthesis. Covers the adversarial training loop (generator vs discriminator), loss functions (Wasserstein, spectral normalization), and practical stabilization techniques. Includes applications like style transfer, super-resolution, and image-to-image translation. The course explains how diffusion models iteratively denoise random noise to generate images, contrasting with GAN training dynamics.","intents":["I want to generate synthetic images for data augmentation or creative applications","I need to understand how GANs and diffusion models work before using pre-trained models like Stable Diffusion","I want to implement style transfer or image-to-image translation for a custom domain"],"best_for":["researchers and practitioners exploring generative models beyond classification","teams building creative AI applications (art generation, design tools)","learners wanting to understand adversarial training and mode collapse before using modern diffusion models"],"limitations":["GAN training is notoriously unstable; course covers stabilization techniques but doesn't guarantee convergence on custom datasets","Requires large, high-quality image datasets (10K+ images) for good results; small datasets often produce mode collapse","Diffusion model coverage is limited compared to modern courses; no guidance on classifier-free guidance or LoRA fine-tuning","Computationally expensive; requires high-end GPU (V100, A100) for reasonable training times"],"requires":["Python 3.7+","fastai library with GAN modules","PyTorch for custom architecture implementation","GPU with 8GB+ VRAM (16GB+ recommended for large models)","Large image dataset (10K+ images for training from scratch)"],"input_types":["image files (JPEG, PNG) for training","seed noise vectors for generation","paired images for image-to-image translation"],"output_types":["generated images (synthetic samples)","style-transferred images","super-resolved images","generator and discriminator model checkpoints"],"categories":["image-visual","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_6","uri":"capability://image.visual.object.detection.and.instance.segmentation.with.convolutional.architectures","name":"object detection and instance segmentation with convolutional architectures","description":"Teaches object detection (localizing and classifying multiple objects in an image) and instance segmentation (pixel-level masks for each object) using architectures like Faster R-CNN, RetinaNet, and Mask R-CNN. Covers the two-stage detection pipeline (region proposal network + classification/localization head), anchor-based vs anchor-free approaches, and loss functions (focal loss for class imbalance). Includes practical patterns for handling variable image sizes, non-maximum suppression, and evaluation metrics (mAP).","intents":["I need to detect and localize objects in images (e.g., defects in manufacturing, vehicles in traffic)","I want to segment individual instances of objects, not just classify the whole image","I need to understand how modern object detectors balance speed and accuracy"],"best_for":["computer vision engineers building production detection systems","teams working on autonomous vehicles, robotics, or surveillance applications","practitioners moving from image classification to more complex vision tasks"],"limitations":["Requires annotated bounding boxes or segmentation masks; annotation is expensive and time-consuming","Performance degrades on small objects or cluttered scenes without careful architecture tuning","Real-time inference requires GPU; CPU inference is too slow for most applications","Course material predates modern efficient detectors (YOLO v5+, EfficientDet); coverage may be outdated"],"requires":["Python 3.7+","fastai vision module with detection support","Annotated dataset with bounding boxes (COCO, Pascal VOC format, or custom)","GPU with 6GB+ VRAM for training; 2GB+ for inference","Annotation tool (LabelImg, CVAT, or similar) to create training data"],"input_types":["images (JPEG, PNG)","bounding box annotations (XML, JSON, or CSV format)","segmentation masks (PNG with class IDs)"],"output_types":["bounding box predictions with class labels and confidence scores","instance segmentation masks","mAP and per-class metrics","trained model checkpoint"],"categories":["image-visual","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_7","uri":"capability://data.processing.analysis.time.series.forecasting.with.recurrent.and.attention.based.architectures","name":"time-series forecasting with recurrent and attention-based architectures","description":"Teaches time-series prediction using recurrent neural networks (LSTMs, GRUs) and attention mechanisms (Transformers). Covers sequence-to-sequence models for multi-step forecasting, handling variable-length sequences, and techniques like teacher forcing during training. Includes practical patterns for feature engineering (lag features, rolling statistics), handling missing values, and evaluating forecasts with metrics like RMSE and MAE. Discusses the trade-off between model complexity and interpretability.","intents":["I need to forecast stock prices, sales, or demand for inventory planning","I want to understand how LSTMs and attention mechanisms capture temporal dependencies","I need to predict multiple steps ahead (e.g., next 7 days of sales) from historical data"],"best_for":["data scientists and analysts working with financial, sales, or operational time-series data","teams building forecasting systems for supply chain or resource planning","practitioners learning sequence modeling before applying it to NLP or video"],"limitations":["RNNs struggle with very long sequences (>500 timesteps) due to vanishing gradients; Transformers are more stable but require more data","Assumes stationarity or requires explicit detrending; non-stationary series need careful preprocessing","No coverage of probabilistic forecasting (uncertainty quantification) or conformal prediction","Course material predates modern time-series Transformers (Informer, N-BEATS); coverage may be limited"],"requires":["Python 3.7+","fastai library with time-series support","Pandas for time-series manipulation","Historical time-series data (minimum 100-200 observations for training)","GPU optional; CPU training is feasible for small datasets"],"input_types":["CSV with timestamp and value columns","multivariate time-series (multiple features per timestep)","irregular time-series (non-uniform sampling)"],"output_types":["point forecasts (single predicted value per timestep)","multi-step ahead predictions (sequence of future values)","prediction intervals or uncertainty estimates","trained model checkpoint"],"categories":["data-processing-analysis","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_8","uri":"capability://safety.moderation.model.interpretation.and.feature.importance.analysis","name":"model interpretation and feature importance analysis","description":"Teaches techniques for understanding what a trained deep learning model has learned and which features drive predictions. Covers gradient-based methods (saliency maps, integrated gradients), attention visualization, permutation importance, and SHAP values. For images, includes techniques like Class Activation Maps (CAM) and Grad-CAM to visualize which regions influence predictions. Emphasizes the importance of model interpretability for debugging, building trust, and meeting regulatory requirements.","intents":["I need to debug why my model makes incorrect predictions on certain inputs","I want to explain model predictions to non-technical stakeholders or regulators","I need to identify which features are most important for a prediction task"],"best_for":["practitioners building production ML systems where interpretability is required (healthcare, finance, legal)","teams debugging model failures and understanding failure modes","researchers studying what neural networks learn and how they generalize"],"limitations":["Interpretation techniques are often heuristic and don't provide formal guarantees about correctness","Some methods (SHAP, integrated gradients) are computationally expensive for large models or datasets","Visualization techniques are primarily designed for images and text; less developed for tabular data","Interpretability can be misleading if not used carefully; requires domain expertise to validate"],"requires":["Python 3.7+","fastai library with interpretation modules","Trained deep learning model (image, text, or tabular)","Test data or specific examples to interpret","Optional: SHAP library for advanced feature importance"],"input_types":["trained model checkpoint","input samples (images, text, or tabular data)","ground truth labels (optional, for error analysis)"],"output_types":["saliency maps or attention visualizations","feature importance scores","confusion matrices and error analysis","SHAP value explanations"],"categories":["safety-moderation","code-generation-editing"],"confidence":0.5,"matches":0,"success_rate":0},{"id":"awesome-practical-deep-learning-for-coders-fast-ai__cap_9","uri":"capability://automation.workflow.production.deployment.and.inference.optimization","name":"production deployment and inference optimization","description":"Teaches how to take a trained fastai model and deploy it for inference in production environments. Covers model export (to ONNX, TorchScript), quantization (int8, float16) for reducing model size and latency, and deployment patterns (REST API, serverless functions, edge devices). Includes practical guidance on handling batch inference, caching, and monitoring model performance in production. Discusses the trade-off between model accuracy and inference speed.","intents":["I have a trained model and need to serve predictions via an API or mobile app","I need to reduce model size and latency for real-time inference on edge devices","I want to monitor model performance and detect data drift in production"],"best_for":["ML engineers and full-stack developers deploying models to production","teams building real-time inference systems (recommendation, fraud detection, content moderation)","practitioners optimizing models for resource-constrained environments (mobile, IoT)"],"limitations":["Deployment patterns are framework-agnostic but require DevOps expertise (Docker, Kubernetes) for production scale","Quantization can reduce accuracy; requires careful validation on representative test sets","No built-in monitoring or A/B testing infrastructure; requires integration with external tools","Course material may not cover latest deployment platforms (Hugging Face Spaces, Modal, Replicate)"],"requires":["Python 3.7+","Trained fastai model checkpoint","ONNX or TorchScript for model export","Deployment platform (Flask, FastAPI, AWS Lambda, Docker)","Optional: quantization libraries (TensorRT, ONNX Runtime)"],"input_types":["trained model checkpoint (.pth file)","inference requests (images, text, or tabular data)","batch data for optimization"],"output_types":["exported model in ONNX or TorchScript format","quantized model (reduced size)","REST API or serverless function","performance metrics (latency, throughput)"],"categories":["automation-workflow","tool-use-integration"],"confidence":0.5,"matches":0,"success_rate":0}],"trust":{"score":21,"verified":false,"data_access_risk":"high","permissions":["Python 3.7+","Jupyter notebook environment (local or cloud-based like Kaggle/Colab)","GPU access recommended but not required for initial lessons","fastai library (PyTorch-based, installed via pip)","fastai library (built on PyTorch)","GPU with 4GB+ VRAM (NVIDIA, Apple Silicon, or cloud GPU)","labeled image dataset (minimum 10-20 images per class for proof-of-concept)","Raw data (images, text, or other modalities)","Annotation tool (LabelImg, CVAT, Prodigy, or custom)","Annotation guidelines and quality control process"],"failure_modes":["Assumes Python programming competency — not suitable for non-programmers","Heavy focus on computer vision and NLP; limited coverage of reinforcement learning or time-series forecasting","Course material is updated periodically; older versions may reference deprecated library APIs","Transfer learning assumes source domain (ImageNet) is sufficiently similar to target domain; fails for highly specialized domains like microscopy or satellite imagery without domain-specific pre-trained models","Fine-tuning on very small datasets (<50 images) still risks overfitting despite regularization techniques","Requires GPU for reasonable training speed; CPU training on large images is prohibitively slow","Annotation is labor-intensive and expensive; no shortcut for creating high-quality labels","Course provides guidance but not automated annotation tools; requires external services or crowdsourcing platforms","Inter-annotator agreement metrics are discussed but not deeply covered; requires domain expertise to set meaningful thresholds","No coverage of active learning or semi-supervised learning for reducing annotation burden","builder identity is not verified yet","no observed match outcomes yet"],"rank_breakdown":{"adoption":0.05,"quality":0.24,"ecosystem":0.25,"match_graph":0.25,"freshness":0.5,"weights":{"adoption":0.25,"quality":0.25,"ecosystem":0.1,"match_graph":0.35,"freshness":0.05}},"observed_outcomes":{"matches":0,"success_rate":0,"avg_confidence":0,"top_intents":[],"last_matched_at":null},"maintenance":{"status":"inactive","updated_at":"2026-06-17T09:51:04.047Z","last_scraped_at":"2026-05-03T14:00:30.220Z","last_commit":null},"community":{"stars":null,"forks":null,"weekly_downloads":null,"model_downloads":null,"model_likes":null}},"distribution":{"claim_url":"https://unfragile.ai/submit?claim=practical-deep-learning-for-coders-fast-ai","compare_url":"https://unfragile.ai/compare?artifact=practical-deep-learning-for-coders-fast-ai"}},"signature":"XUv5jQxpEgFEkvzpy5glcAkPzYp3bd8mFcLHqcS2w29foNtaQLpiIZzk99ZJIUOFkUxEPzT5lJY2yIvyXN9VBg==","signedAt":"2026-06-20T16:06:25.877Z","signedBy":"unfragile.ai","version":1},"_links":{"self":"https://unfragile.ai/api/v1/passport/practical-deep-learning-for-coders-fast-ai","artifact":"https://unfragile.ai/practical-deep-learning-for-coders-fast-ai","verify":"https://unfragile.ai/api/v1/verify?slug=practical-deep-learning-for-coders-fast-ai","publicKey":"https://unfragile.ai/api/v1/trust-passport-public-key","spec":"https://unfragile.ai/trust","schema":"https://unfragile.ai/schema.json","docs":"https://unfragile.ai/docs"}}