1298c8e74b
5596d500c2bf2dee886a1641710479bb7c0d1ed5 (v0.1.1) -> b087cf158905e62b16c02e7bb1aa38691186ea6b (v0.1.2). Approval gate now freezes a data snapshot with nonce + digest bound to the detached runner; run_name/epochs/batch/lr/calib/seed validated before the gate. Checks at the pin: pytest 37 passed; validate/doctor/compat OK.
30 lines
1.3 KiB
YAML
30 lines
1.3 KiB
YAML
name: thomas
|
|
repo: https://github.com/keppy/hermes-plugin-thomas
|
|
sha: b087cf158905e62b16c02e7bb1aa38691186ea6b
|
|
description: >-
|
|
Fine-tune a small calibrated encoder classifier on Modal and rate it with gonogo. Every GPU launch
|
|
goes through the human approval gate with the exact config shown; data checks and evaluation run
|
|
locally for free. Disclosure — thomas_encoder_train spends the user's Modal GPU credits and uploads
|
|
the chosen training JSONL rows to Modal after the human approval gate; the base model is downloaded
|
|
from Hugging Face on the Modal side. Training and inference run in a separate Python interpreter the
|
|
user configures via THOMAS_PYTHON (with the author's thomas package, torch and modal installed); the
|
|
Hermes venv only gets gonogo-eval. Runs and artifacts are written under HERMES_HOME/thomas/runs; the
|
|
eval tool writes report files to the path the model supplies.
|
|
maintainer: keppy
|
|
tier: community
|
|
category: tools
|
|
requires_hermes: ">=0.21"
|
|
platforms: [linux, macos]
|
|
version: "0.1.2"
|
|
docs_url: https://github.com/keppy/hermes-plugin-thomas#readme
|
|
capabilities:
|
|
provides_tools:
|
|
- thomas_check_data
|
|
- thomas_encoder_train
|
|
- thomas_run_status
|
|
- thomas_encoder_eval
|
|
provides_hooks:
|
|
- pre_tool_call
|
|
provides_middleware: []
|
|
requires_env: []
|