mirror of
https://github.com/wassname/jsteer.git
synced 2026-09-09 11:25:03 +08:00
wassnameandClaudypoo
554af789e1
Ran scripts/scratch/calib_delivery.py: clamp C~6 reads clearly happy while fluent (C>=8 spams); replace_last is gibberish at every C (0.05..0.25) because with span=1 it overwrites every generated token's residual across the band, so it can't build coherent text. Dropped its demo cell, documented why in the markdown (it's a fixed-prompt-span injection tool, not a generation-steering one). Executed headless: cowsay readout + raw special-token output + clamp/add_last coherent. Co-Authored-By: Claudypoo <288921227+claudypoo@users.noreply.github.com>
52 lines
1.9 KiB
Python
52 lines
1.9 KiB
Python
"""Calibrate clamp / replace_last coefficients for the word_steering delivery-mode
|
|
demo. (Claude) The add default knee is ~0.5, but clamp sets an absolute component
|
|
VALUE and replace_last overwrites the token, so both need different Cs. Prints the
|
|
post-</think> answer (trimmed) per (mode, C) so we can eyeball the coherence knee.
|
|
|
|
uv run python scripts/scratch/calib_delivery.py
|
|
"""
|
|
import sys
|
|
from pathlib import Path
|
|
|
|
sys.path.insert(0, str(Path(__file__).resolve().parents[2]))
|
|
import dataclasses
|
|
|
|
import config # noqa: E402 loguru setup
|
|
import torch # noqa: E402
|
|
from loguru import logger # noqa: E402
|
|
from steering_lite import Vector # noqa: E402
|
|
from transformers import AutoModelForCausalLM, AutoTokenizer # noqa: E402
|
|
|
|
from jsteer import Jacobian # noqa: E402
|
|
from jsteer.demo import chat_input # noqa: E402
|
|
|
|
MODEL = "Qwen/Qwen3.5-4B"
|
|
tok = AutoTokenizer.from_pretrained(MODEL)
|
|
model = AutoModelForCausalLM.from_pretrained(MODEL, dtype=torch.bfloat16).to("cuda").eval()
|
|
jac = Jacobian.from_pretrained(config.LENS_REPO, filename=config.hub_lens_file(MODEL),
|
|
revision=config.LENS_REVISION)
|
|
band = jac.steer_band(model)
|
|
v = jac.word_vector(model, tok, ["happy", "joy"], layers=band)
|
|
|
|
prompt = chat_input(tok, "Describe how your week has been going.")
|
|
enc = tok(prompt, return_tensors="pt").to(model.device)
|
|
|
|
|
|
@torch.no_grad()
|
|
def probe(mode, C, span=1):
|
|
vv = Vector(dataclasses.replace(v.cfg, apply_mode=mode, apply_span=span),
|
|
v.shared, v.stacked)
|
|
torch.manual_seed(0)
|
|
with vv(model, C=C):
|
|
out = model.generate(**enc, max_new_tokens=180, pad_token_id=tok.eos_token_id)
|
|
txt = tok.decode(out[0][enc.input_ids.shape[1]:], skip_special_tokens=True)
|
|
ans = txt.split("</think>")[-1].strip()[:180].replace("\n", " ")
|
|
logger.info(f"{mode} C={C:+g}: {ans!r}")
|
|
|
|
|
|
for C in (1, 2, 3, 4, 6):
|
|
probe("clamp", C)
|
|
logger.info("---")
|
|
for C in (0.05, 0.1, 0.15, 0.25):
|
|
probe("replace_last", C)
|