Skip to content
System One

laya-gfx1030

Runs Laya's English typed-decision checkpoint on a Windows machine with an AMD Radeon RX 6900 XT (gfx1030) GPU through a pinned PyTorch ROCm build, with an optional Jev-compatible HTTP loopback service.

Thotheris/laya-gfx1030 on GitHub

Stars
0
Forks
0
Language
Python
Last push
22 Sep 2026

Post on X

Views
30
Likes
0
Reposts
0

Read from GitHub and X on . Counts change daily.

How it uses Jev

entry point, and it is opt-in.

Nothing here reaches the cloud, with or without credentials: the transport is
replaced, so ``api.typesafe.ai`` and ``openrouter.ai`` are never contacted, and
the socket guard would stop them if they were.

Two limits on that sentence, both asserted in the tests rather than assumed:

* the transport, ``keystore`` and the HTTP helpers validate the URL *themselves*
  (literal ``127.0.0.1``, ``http``, an explicit port, no userinfo/query/fragment,
  no path on the server URL) and use an opener with **no environment proxy** and
  **no redirect following** -- so a stray ``HTTP_PROXY`` or a ``302`` in a reply
  cannot route a "local" request to a remote host even when the socket guard is
  not installed (``guard=False``, or a direct call to ``make_local_transport``);
* the socket guard is a per-process guard on ``socket.socket.connect``, not a

View in scripts/jev_local.py

Open on GitHub

More like this

Keep browsing