Deep Learning with PyTorch
About This Course
Train neural networks that work on your own data, using PyTorch. Over three days you build models up to convolutional networks, sequence models and fine-tuned transformers, and you train them on GPUs, working throughout with real image, audio and time-series datasets rather than tidied-up examples.
The course opens with classical machine learning in scikit-learn, so you can tell when a neural network is the right tool and when a random forest would do the job for a fraction of the effort. From there you move into PyTorch proper, and into the judgement calls that decide whether a model is any good.
Day 3 is the messy part most courses skip. Real image and audio data arrives in odd shapes and sizes, so you learn to prepare it yourself, including data too large to hold in memory. You then fine-tune pre-trained HuggingFace models, which is what you would actually do rather than training from scratch.
Who This Course Is For
Data scientists, analysts, quants, researchers, engineers and software developers who want to build and train neural networks on their own data.
It suits people who have used Python for analysis and now need models that classical methods cannot deliver: image and audio classification, sequence and time-series prediction, or fine-tuned language and vision models.
Prerequisites
Working knowledge of Python. You should be comfortable with functions, loops and working with data in NumPy arrays or Pandas DataFrames.
High-school mathematics is enough. Basic probability and linear algebra help but are not required, and no prior deep learning experience is assumed.
Day 1 covers classical machine learning with scikit-learn. Skip it if you have recently completed Python for Machine Learning or Python for Scientists and Engineers, or if you already use scikit-learn regularly.
What You'll Learn
- Build and train neural networks in PyTorch, on GPUs and other accelerators.
- Judge when a neural network is the right tool and when classical machine learning will do.
- Diagnose training problems, and apply regularisation that improves results.
- Prepare your own image, audio and tabular data for training, including datasets too large to fit in memory.
- Fine-tune pre-trained models from HuggingFace rather than training from scratch.
- Deploy trained models for inference, and monitor them for degradation.
Course Syllabus
Day 1: Machine learning foundations
- Core machine learning concepts and the scikit-learn workflow
- Model validation and selection
- Feature engineering
- Putting models into production
Day 2: Neural networks in PyTorch
- Tensors, automatic differentiation and AI accelerators
- Designing and training neural networks in PyTorch
- Monitoring training runs and applying regularisation
- Convolutional networks for imagery
Day 3: Real data, sequences and transformers
- Sequence and time-series models
- Transformers
- Working with your own image and audio data
- Fine-tuning pre-trained models from HuggingFace
Appendices
- Higher-level training frameworks and alternative libraries