mert-kurttutan/torchview

Health v2 · maintenance only

Adoption not part of the score

1072 stars · 49 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

torchview is a Python library that visualizes PyTorch models as visual graphs, showing tensors, modules, torch functions, and input/output shapes. It is the PyTorch equivalent of Keras's plot_model, supporting meta tensors, RNNs, skip connections, and Hugging Face models.

Use cases

• visualize pytorch model architecture as a graph
• show input and output shapes of pytorch model layers
• plot pytorch model like keras plot_model
• inspect operations between tensors in a neural network
• visualize large models without consuming memory using meta tensors
• generate architecture diagrams for hugging face models

When to choose

• you need architecture diagrams of PyTorch models with layer shapes
• you want a Keras plot_model equivalent for PyTorch
• you need to visualize very large models without memory overhead via meta tensors
• you want to see tensor operations, not just module calls

When to avoid

• you need training metrics dashboards or profiling rather than architecture graphs
• you work outside PyTorch (e.g., TensorFlow or JAX)
• you cannot install the Graphviz system dependency

Facets

library · maturity active

data-visualization developer-tools deep-learning machine-learning data-visualization python cross-platform pytorch graphviz model-visualization neural-networks torchviz

Member repositories

For agents

markdown · JSON · MCP: product_card(name="mert-kurttutan/torchview")