Louis Maddox lmmx

can you try to answer this question about how much space could be saved by pruning the contrib_ops dir by going and getting the onnx runtime in question? im not quite sure how you would evaluate this tbh Anush008/fastembed-rs#206 (comment)

●

I'll fetch the GitHub issue comment to understand the question, then investigate the ONNX Runtime to evaluate potential space savings.

●

WebFetchhttps://github.com/Anush008/fastembed-rs/issues/206#issuecomment-3697236466

In fast_ica.rs, the exp function appears to have a bug in the exponentiation func, where the variable named exp stores $-u^2/2$ rather than $exp(-u^2/2)$

For $G(u) = -\exp(-u^2/2)$, the derivatives should be:

$$g(u) = G'(u) = u \exp\left(-\frac{u^2}{2}\right)$$

$$g'(u) = G''(u) = (1 - u^2) \exp\left(-\frac{u^2}{2}\right)$$

https://github.com/rust-ml/linfa/blob/db3cadee926e546a2809346c73689ce5048a96e9/algorithms/linfa-ica/src/fast_ica.rs#L239-L246

You are facilitating a dialogue among groups with conflicting positions on a contested issue. Your goal is to help participants achieve meta-consensus: agreement on the legitimacy of disputed values, acceptance of the credibility of disputed beliefs, and agreement on the nature of disputed choices—even while disagreement persists at the level of specific preferences.

This approach recognizes that "unbridled pluralism and absolute consensus alike turn out on closer inspection to have few advocates or defenders." The resolution involves "pluralism in values, beliefs, and preferences in the context of a meta-consensus on one or more of these three aspects."

Begin by identifying distinct groups and mapping their positions across normative, epistemic, and preference dimensions. Then work to surface meta-consensus at each level while preserving legitimate disagreement.

	import matplotlib.pyplot as plt

	# Data
	images = [
	"boston_cooking_a",
	"boston_cooking_b",
	"finnish_cooking_a",
	"linguistics_thesis_a",
	"linguistics_thesis_b"
	]

	from pathlib import Path
	import re

	import polars as pl
	import polars_fastembed # noqa: F401


	def clean_markdown(text: str) -> str:
	"""Strip markdown/HTML artifacts from text."""
	text = re.sub(r'<table>[\s\S]*?</table>', '', text)

	//! ```cargo
	//! [dependencies]
	//! faer = "0.23"
	//! ```

	use faer::prelude::*;
	use faer::linalg::solvers::Svd;

	fn main() -> Result<(), faer::linalg::solvers::SvdError> {
	// 3×2 example matrix

	#!/bin/bash
	set -e

	cd $HOME/lab/ort/build

	# Clean previous attempt
	rm -rf onnxruntime

	git clone --recursive --depth 1 --branch v1.23.0 https://github.com/microsoft/onnxruntime