mirror of
https://github.com/wassname/jsteer.git
synced 2026-08-21 04:50:17 +08:00
wassnameandClaudypoo
ed7fa92996
One generically-named notebook (replaces persona-named ones): loads model+lens once, builds all 7 steering vectors (word/persona_vector/topk/soft/pinv/meandiff/random-null) with the j-thoughts lens readout, then loads precomputed results and shows a per-method generation dropdown + comparison table. Heavy 7-method sweep (~18 min) runs headless via scripts/scratch/run_steering_demo.py -> artifacts/steering_demo_results.json, because marimos single-threaded kernel makes a long in-cell compute un-monitorable (any status poll interrupts it). Notebook loads the JSON so it renders instantly. Result (dilemma, P(YES=lie)): flat ~0.03-0.14 at every methods coherent edge vs 0.107 baseline; persona_pinv widest window (C*+ = +1.19). Includes the GPT-5.6-terra comment review (docs/reviews/). Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>
45 lines
1.6 KiB
TOML
45 lines
1.6 KiB
TOML
[project]
|
|
name = "jsteer"
|
|
version = "0.1.0"
|
|
description = "Fit a model's full Jacobian once (jlens), then steer any word or persona (steering-lite runtime). repeng-style UX for Jacobian pullback steering."
|
|
readme = "README.md"
|
|
requires-python = ">=3.11"
|
|
license = "MIT"
|
|
dependencies = [
|
|
"torch>=2.1",
|
|
"transformers>=4.51",
|
|
"accelerate>=1.6",
|
|
"jlens", # the researchers' verified Jacobian estimator + cache format
|
|
"steering-lite", # hook runtime: Vector, attach, `with v(model, C=...)`, calibrate
|
|
"loguru>=0.7",
|
|
"tqdm>=4.66",
|
|
"tabulate>=0.9", # Claude: steering_lite.calibrate imports it at module load (undeclared in its core deps)
|
|
]
|
|
|
|
[project.optional-dependencies]
|
|
notebooks = ["matplotlib>=3.8", "ipykernel", "tabulate", "nbformat", "nbconvert"]
|
|
test = ["pytest"]
|
|
|
|
[tool.uv.sources]
|
|
# Claude: switched from git to local editable path deps. The jlens git source
|
|
# (github.com/anthropics/jacobian-lens) is not a fetchable public repo, and both
|
|
# deps live locally under active co-development, so editable paths are correct
|
|
# for this dev machine. Revert to git sources when publishing.
|
|
jlens = { path = "../j-steer-dev/docs/vendor/jacobian-lens", editable = true }
|
|
steering-lite = { path = "../../lite/steering-lite", editable = true }
|
|
|
|
[build-system]
|
|
requires = ["setuptools>=68"]
|
|
build-backend = "setuptools.build_meta"
|
|
|
|
[dependency-groups]
|
|
dev = [
|
|
"ipykernel>=7.3.0",
|
|
"ipywidgets>=8.1.8",
|
|
"datasets>=2.0", # jlens.examples.load_wikitext_prompts: the fitting corpus for fit.py
|
|
"marimo>=0.23.13",
|
|
]
|
|
|
|
[tool.setuptools]
|
|
packages = ["jsteer"]
|