{"slug": "show-hn-tensorviz-explore-pytorch-models-inside-vs-code", "title": "Show HN: TensorViz – Explore PyTorch models inside VS Code", "summary": "TensorViz launched as a VS Code tool that lets developers explore 29 interactive PyTorch architectures, including ResidualClassifier, without installing Python or any additional software. The tool offers guided tours of model internals spanning attention, convolution, residual paths, GANs, diffusion, and Evoformer designs, and each model ships with its source and a small reproduction recipe so users can open their own PyTorch classes in the editor.", "body_md": "[START HEREExplore graph ResidualClassifier4-step tour TensorViz](https://tensorviz.dev/explore/residual-classifier/) \n\n## The residual path\n\nWhat happens where the skip path rejoins?\n\nMODEL EXPLORER / CURATED BY TENSORVIZ\n\n29 interactive PyTorch architectures. Follow a guided tour, inspect captured shapes, and take the source back to your own project.\n\nNo installation · No Python needed\nCHOOSE A LEARNING PATH\n\nStart with attention, change one design choice at a time, then connect the pieces in LLaMA.\n\nFollow convolution, residual paths, U-shaped skips and image patches through landmark vision designs.\n\nExplore alignment, bidirectional pretraining, image–text pairing, low-rank adaptation and reasoning post-training.\n\nCompare the GAN game with noise prediction, latent-space denoising and a Transformer diffusion backbone.\n\nInspect selective recurrent state, Evoformer sequence/pair coupling and promptable mask decoding.\n\nA teaching order, not a claim that each later model adopted every earlier idea.\n\n29 models to explore\n\nWhere does an image become a classifier input?\n\nHow do queries, keys and values meet?\n\nComponent · 2019\n\nWhat does normalization change in one token vector?\n\nComponent · 2020\n\nWhy does this feed-forward block split into two branches?\n\nArchitecture · 2017\n\nHow does a decoder form a next-token distribution?\n\nWhat happens to the shortcut when normalization moves?\n\nComponent · 2023\n\nWhich attention heads can share a key/value cache?\n\nComponent · 2021\n\nWhere does token position enter the attention score?\n\nInference method · 2022\n\nCan exact attention avoid storing the entire score matrix?\n\nComponent · 2024\n\nWhat must MLA store to avoid reconstructing every key and value?\n\nComponent · 2017\n\nHow can a token use only part of a larger network?\n\nTraining method · 2021\n\nHow can two small matrices adapt a frozen linear layer?\n\nArchitecture · 2023\n\nWhere do RMSNorm, RoPE and SwiGLU meet in an actual decoder block?\n\nArchitecture · 2012\n\nHow do five convolutions turn an image into a classification vector?\n\nArchitecture · 2015\n\nWhat can flow through a shortcut when a residual branch changes?\n\nHow does a segmentation decoder recover features lost during downsampling?\n\nArchitecture · 2020\n\nWhere does an image become a sequence for a Transformer?\n\nComponent · 2014\n\nHow does a recurrent decoder decide which source positions to consult?\n\nArchitecture · 2018\n\nWhat changes when the encoder can see both left and right context?\n\nTraining method · 2014\n\nWhich network is being updated, and what is it trying to change?\n\nTraining method · 2020\n\nHow does a model learn to reverse corruption?\n\nArchitecture · 2021\n\nWhy denoise a latent, and where does text enter?\n\nArchitecture · 2022\n\nWhat changes when a Transformer denoises latent patches?\n\nWhere do the image and text meet?\n\nHow do image features and a user prompt become a mask?\n\nHow does the input decide what a recurrent state retains?\n\nHow do sequence and pair representations interact before structure is produced?\n\nTraining method · 2025\n\nWhat changes in the learning process while the base architecture stays the same?\n\nTry another idea, or return to the full collection.\n\nBRING IT BACK TO YOUR PROJECT\n\nEvery model includes its source and a small reproduction recipe. Open your own PyTorch class with tensorViz in VS Code.", "url": "https://wpnews.pro/news/show-hn-tensorviz-explore-pytorch-models-inside-vs-code", "canonical_source": "https://tensorviz.dev/explore/", "published_at": "2026-10-12 04:49:06+00:00", "updated_at": "2026-10-12 05:09:09.158150+00:00", "lang": "en", "topics": ["developer-tools", "artificial-intelligence", "machine-learning", "ai-tools"], "entities": ["TensorViz", "PyTorch", "VS Code", "ResidualClassifier", "LLaMA", "Evoformer"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/show-hn-tensorviz-explore-pytorch-models-inside-vs-code", "markdown": "https://wpnews.pro/news/show-hn-tensorviz-explore-pytorch-models-inside-vs-code.md", "text": "https://wpnews.pro/news/show-hn-tensorviz-explore-pytorch-models-inside-vs-code.txt", "jsonld": "https://wpnews.pro/news/show-hn-tensorviz-explore-pytorch-models-inside-vs-code.jsonld"}}