torchview

• Fix readme format on pypi
• Replace flake8 with ruff (only relevant for development issue)

• Fixed the issue about isolated Tensor nodes
  • Issue: #73 , PR: #74
• Complete coverage of tensor creation ops
  • PR: #74

• Fixed the bug for modules that changes mutable inputs, e.g. list of tensors PR
  • PR: #70 , Issue: #63
• Very limited support for traced modules via #71 , see why here #66
• #66 also hints at the solution for support for traced modules for those who want to implement and contribute the project

• Added support for change the direction of graph
• Fixed issues for modules that create tensor during forward prop

• Support for more flexible architectures (Reuse of input, unused input tensors)
• Increased support for Huggingface models including, bert, gpt2, detectron

Increased support for some huggingface tokenizer

• new display feature for nested modules ('expand_nested') very similar to keras plot_model
• new test models
• code cleaning