{"id":51294469,"url":"https://github.com/knowm/ktram-neural-core","last_synced_at":"2026-06-30T13:02:43.522Z","repository":{"id":368018588,"uuid":"1265504015","full_name":"knowm/ktram-neural-core","owner":"knowm","description":"Neural core kT-RAM implementation with unit-crossbars and the 2-1 kT-bit configuration. ","archived":false,"fork":false,"pushed_at":"2026-06-28T16:58:22.000Z","size":2318,"stargazers_count":1,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-06-28T18:23:12.389Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/knowm.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-10T20:46:37.000Z","updated_at":"2026-06-28T16:58:25.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/knowm/ktram-neural-core","commit_stats":null,"previous_names":["knowm/ktram-neural-core"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/knowm/ktram-neural-core","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knowm%2Fktram-neural-core","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knowm%2Fktram-neural-core/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knowm%2Fktram-neural-core/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knowm%2Fktram-neural-core/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/knowm","download_url":"https://codeload.github.com/knowm/ktram-neural-core/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/knowm%2Fktram-neural-core/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":34967640,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-06-30T02:00:05.919Z","response_time":92,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-06-30T13:02:42.770Z","updated_at":"2026-06-30T13:02:43.512Z","avatar_url":"https://github.com/knowm.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"# ktram-neural-core\n\n**kT-RAM** — Thermodynamic RAM — is a construct: addressable memristor-pair synapses\n(kT-bits) driven by a small instruction set, where reading the memory *is* computing with it.\nNo separate fetch, no ALU, no trip out to DRAM. This package emulates one kT-RAM topology —\nthe **2-1 neural lane** — in Python. It is the software half of a build: the emulator runs the\ninstruction set and the experiments ahead of the physical memristor-crossbar circuits it\nmodels, so the two can be developed side by side.\n\nThe emulator is built for teaching and intuition, not production. It pairs with the Knowm blog\nseries that works through the kT-RAM instruction set chapter by chapter — start with\n[Chapter 4: The Neural Lane](https://knowm.ai/blog/building-the-neural-lane/) and its companion\n[Chapter 4b: The Neural Lane Emulator](https://knowm.ai/blog/the-neural-lane-emulator/). It is\nnot an EDA component and it is not meant for large-scale simulation.\n\nThe emulator's internal structure mirrors the hardware it models — unit crossbars, differential\npairs, neural lanes — selected by Activation Address Tuples (AATs), rather than an idealized\nflat-synapse construct.\n\n```\nCore\n └─ NeuralLane[]                      2-1 differential readout: y = Σ(Ga−Gb) / Σ(Ga+Gb)\n      └─ UnitCrossbarPair[]           a differential pair = one signed synapse (kT-bit)\n           ├─ UnitCrossbar  a-side    carries the selected memristor (Ideal fidelity)\n           └─ UnitCrossbar  b-side    the differential partner\n```\n\nA **Core** is specified the way the hardware is — by unit-crossbar geometry and operating model —\nand addressed only by AATs. A single synapse is the smallest lane.\n\n## Install\n\n```bash\npip install ktram-neural-core        # once published\n# or, from a checkout:\npip install -e python/\n```\n\nRuntime dependency is NumPy only. Plotting (`matplotlib`) is an examples-only extra:\n`pip install -e \"python/[examples,dev]\"`.\n\n## Quickstart — a single synapse\n\n```python\nfrom ktram_neural_core import Core\n\ncore = Core(1, 1, spaces_per_lane=1, num_lanes=1, model=\"byte\", init=\"medium\")\nlane = core.lane(0)\nz = (0,)                              # the AAT: address 0 in the one space\n\ny = lane.evaluate(z, \"FF\")           # read; sets and returns y\n    lane.evaluate(z, \"RH\")           # write; one instruction per call\nga, gb = core.read_gab(0, z)         # debug/visualization only\n```\n\nThe four device models — `float`, `byte`, `mss`, `rs` — are all available. Drive voltages\nand pulse width are model-aware Core defaults and fully settable\n(`set_voltages(...)`, `set_pulse_width(dt)`).\n\n## Examples \u0026 benchmarks\n\nRunnable lessons — each a `figures.py` plus, for most, a Colab-ready notebook — and the\nclassifier benchmarks live under [`python/examples/`](python/examples); see its\n[README](python/examples/README.md) for the catalog and Colab links.\n\n```bash\ncd python \u0026\u0026 pip install -e \".[examples,dev]\"\npython examples/single-synapse/figures.py     # writes examples/single-synapse/figures/ (gitignored)\npython examples/iris-classifier/benchmark.py   # a classifier benchmark report\npytest                                          # the unit + behavioral suite\n```\n\nEach `figures.py` defaults to its own `figures/` subdir; pass a path to write elsewhere.\n\n## Documentation\n\nFull docs are in [`docs/`](docs/) — the [object model and instruction set](docs/architecture.md),\n[how to run everything](docs/running.md), and the [repo layout \u0026 conventions](docs/repo-structure.md).\n\n## License\n\nMIT (see [`LICENSE`](LICENSE)) grants copyright. A separate [`PATENTS`](PATENTS) file reserves\nKnowm's US hardware-patent rights: **software emulation is permitted with no patent license;\nhardware realization in the US requires a separate license.**\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fknowm%2Fktram-neural-core","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fknowm%2Fktram-neural-core","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fknowm%2Fktram-neural-core/lists"}