Skip to content
Publishing a binding

clay-tensor-host

0.0.1

Does the number crunching for machine learning, and turns text into tokens and back.

NativeCPUmacOS arm64OfflineHandler

Runs on 0 of 2 nodes

that you can see

Shared

Private

See everything that runs on cloud-nbg1-1

Details · answers, reads, artifact, licence, provenance
HandlerTensor, Tokenizer

A small native program that does the number crunching for machine learning, and turns text into tokens and back.

Runs in
Server
macOS on Apple silicon only. Other platforms would be separate bindings.
Form
Native program, 4.9 MB
Same answer every timeSame answer
May change between versions
Network
Never
Licence
None declared
Signed
By Clay, development keyBy Clay, dev key

Abilities it answers

A program that declares one of these runs this handler on a server. In tests, a handler written in Clay can answer instead.

Tensor#q42ehuofl9q3topq…
matmul, add, softmax, load and 20 more
Tokenizer#nn3cjmcgge5kmp38…
encode, decode

Hashes are the structural ability hashes the kernel computes for the declarations in packages/tensor-host/clay (tensor.clay, gpt2.clay), as the page and the host see them when a program runs. The Tensor hash covers type tensor, which is structural, so every store computes the same hash (wishlist W35). A new operation is a new ability version and a new hash.

Works with

gpt2-small-weightsData · read by Tensor.load

Exact bytes and where they came from

Upstream · github.com/robmurtagh/clay at 0.0.1
Project
https://github.com/robmurtagh/clay
Version
0.0.1
Commit or revision
8fa21355ecac656a2ea725f16b352130268a2720
Licence
None declared
How it was built · repo-local Rust 1.98.1 (.tools), release profile opt-level 3, LTO, 1 codegen unit
Toolchain
repo-local Rust 1.98.1 (.tools), release profile opt-level 3, LTO, 1 codegen unit
Built by
repo maintainer
Built at
2026-11-11T06:30:08Z
Notes
upstream.commit is the repository HEAD when this entry was made (the build is newer than the last change to packages/tensor-host/src). determinism is the closest class in this vocabulary: output bits are NOT promised equal across devices or builds (docs/ml-workloads.md section 6). The build is not reproducible byte for byte across machines. Licence null: the repo declares none yet. Dependencies listed are the direct ones plus the regex engine tokenizers links; the full transitive tree is not listed.
Memory
Not measured
CPU
aarch64; this build is a macOS arm64 Mach-O executable
Source path
packages/tensor-host/target/release/examples/tensor_run
Format
file
  1. cd packages/tensor-host && cargo build --release --example tensor_run
  2. Versions are those in packages/tensor-host/Cargo.lock; licences are read from each crate's Cargo.toml in the local cargo registry.
Built from (3 libraries)
PackageVersionLicence
matrixmultiply0.3.11MIT/Apache-2.0
tokenizers0.22.2Apache-2.0
onig6.5.3MIT
Signatures · 1, development key

Checked when this page was built, against the trust set.

  • c5878029baeb4574 Clay, development key: valid (ok)
Manifest id
a028894e40e576cf442a4f25c6570fba4c7f9cde61c69bcd1d0cbb5164cfc972
Status
ok
Loading it from a runner · checkLoadable, with the code

Fetch /v1/manifests/a028894e40e576cf442a4f25c6570fba4c7f9cde61c69bcd1d0cbb5164cfc972.json, get the artifact bytes by hash, then call checkLoadable(manifest, bytes, policy). It checks the signature against the trusted keys, the revocation list, the fingerprints and sizes, and the policy’s allowed forms and determinism classes, and returns a reason when it refuses. How a runner loads a binding