Relation to other packages¶

How TorchLens relates to the tools people already use. Checked 2026-11-03.

• Which tool for this job?: the dated table, and when five lines of hooks are enough.
• Comparable tools, checked November 2026: each tool's status and source.
• Using TorchLens with other tools: the bridges, by direction.
• Where to go instead: jobs another tool does and TorchLens does not.

Coming from another tool¶

Every page has the same sections: the same task in both tools, a translation table, what TorchLens adds, what the other tool does that TorchLens does not, using both, and semantic differences.

• PyTorch building blocks: forward hooks, torch.fx, Hugging Face output_hidden_states and output_attentions.
• Interpretability and intervention: TransformerLens, nnsight, pyvene, baukit, AutoCircuit, SAELens, RepE, penzai.
• Attribution: Captum, Inseq, transformers-interpret.
• Feature extraction: torchvision's feature extractor, torchvision's IntermediateLayerGetter, timm features_only, thingsvision, torchextractor, surgeon-pytorch, model zoos.
• Inspection and drawing: torchinfo, flax tabulate, torchview, torchviz or hiddenlayer, BertViz or Ecco, torchsnooper, lovely-tensors.