neuraltalk
Source
- Source kind:
github-repository - URL: https://github.com/karpathy/neuraltalk
- Discovery source: https://github.com/karpathy/neuraltalk
- License:
NOASSERTION - Distribution policy:
public-summary-local-archive-only - Public mirror status:
summary-only - Content hash:
9f2a25709a3afc32521e1f0312540b26415e874496b56568b422796c7d8d47f1 - First seen: 2026-05-15
- Last changed: 2026-05-15
Classification
- Primary category: Neural network fundamentals
- Corpus source note: 2026-05-15-karpathy-public-corpus
- Project taxonomy: karpathy-project-taxonomy
- Idea map: karpathy-idea-map
- Topic hub: karpathy-public-work
Summary
NeuralTalk Warning: Deprecated. Hi there, this code is now quite old and inefficient, and now deprecated. I am leaving it on Github for educational purposes, but if you would like to run or train image captioning I warmly recommend my new code release 1 and is SIGNIFICANTLY (I mean, ~100x+) faster because it is batched and runs on the GPU. It also supports CNN finetuning, which helps a lot with performance. This project contains Python+numpy source code for learning Multimodal Recurrent Neural Networks that describe images with sentences. This line of work was recently featured in a…
What This Teaches
- How core neural network ideas can be rebuilt from first principles.
- Useful for grounding later LLM work in gradients, activations, optimization, and model internals.
Why It Matters
This is high-priority for Vipin because it supports durable first-principles understanding instead of shallow API use.
Repository Snapshot
- Full name:
karpathy/neuraltalk - Default branch:
master - HEAD:
c36bd0aa23907638a1217f03a13bdba5e1a8cada - Stars at crawl: 5488
- Forks at crawl: 1330
- File count: 62
- README path:
Readme.md - License path: “
- Created: 2014-11-27T23:22:39Z
- Updated: 2026-05-13T15:31:59Z
- Pushed: 2020-12-22T22:26:36Z
Top-Level Structure
example_images: 26[root]: 9imagernn: 9vis_resources: 7matlab_features_reference: 4python_features: 3cv: 1data: 1eval: 1status: 1
File Extension Profile
.jpg: 20.py: 13.md: 8.js: 5.html: 3.m: 2.prototxt: 2.txt: 2.css: 1.gitignore: 1.json: 1.map: 1
Tags / Release-Like Markers
- No git tags found in the shallow local clone.
Sample File Tree
.gitignorecv/Readme.mddata/README.mddriver.pyeval/multi-bleu.perleval_sentence_predictions.pyexample_images/7EGRMwN.jpgexample_images/89pUfSc.jpgexample_images/animals.jpgexample_images/cat.jpgexample_images/cobra.jpgexample_images/diving.jpgexample_images/dogdinner.jpgexample_images/frog.jpgexample_images/gWEHGwf.jpgexample_images/hole.jpgexample_images/japanese.jpgexample_images/jump.jpgexample_images/koala.jpgexample_images/mic.jpgexample_images/pope.jpgexample_images/pose.jpgexample_images/qjujW6d.pngexample_images/QmG3nS6.jpgexample_images/Readme.mdexample_images/result.htmlexample_images/result_struct.jsonexample_images/seal.jpgexample_images/tasks.txtexample_images/UbVIl1e.jpgexample_images/vgg_feats.matexample_images/work.jpgimagernn/__init__.pyimagernn/data_provider.pyimagernn/generic_batch_generator.pyimagernn/imagernn_utils.pyimagernn/lstm_generator.pyimagernn/Readme.mdimagernn/rnn_generator.pyimagernn/solver.pyimagernn/utils.pymatlab_features_reference/deploy_features.prototxtmatlab_features_reference/extract_features.mmatlab_features_reference/prepare_images_batch.mmatlab_features_reference/README.mdmonitorcv.htmlpredict_on_images.pypy_caffe_feat_extract.pypython_features/deploy_features.prototxtpython_features/extract_features.pypython_features/README.mdReadme.mdrequirements.txtstatus/Readme.mdvis_resources/d3.min.jsvis_resources/d3utils.cssvis_resources/d3utils.jsvis_resources/jquery-1.8.3.min.jsvis_resources/jsutils.jsvis_resources/underscore-min.jsvis_resources/underscore-min.mapvisualize_result_struct.html
Public Handling Notes
- EXTRACTED: This page records public metadata and a source-grounded summary.
- INFERRED: Full local preservation, when available, is for private/local use unless a license or explicit source policy makes public redistribution safe.
- Do not treat this page as permission to republish unlicensed source text or code wholesale.