Fork me on GitHub
#data-science
<
2022-12-30
>
respatialized17:12:18

https://open.substack.com/pub/garymarcus/p/is-chatgpt-really-a-code-red-for I think physics exams questions can probably be reworked to reveal the lack of common sense reasoning abilities in LLMs.

aaelony22:12:41

Hemingway famously said there are two ways things play out; gradually first, then suddenly ;)

Benjamin09:12:43

{:paths ["src"]
 :deps {techascent/tech.ml.dataset {:git/url "[email protected]:techascent/tech.ml.dataset.git"
                                    :git/sha "e58a1b8aed097ce7cd84a46f9cfe686fd37c18f8"}}}
(ns hello)

(require '[tech.v3.dataset :as ds])
1. Caused by java.lang.ClassNotFoundException tech.v3.dataset.Text URLClassLoader.java: 445 http://java.net.URLClassLoader/findClass is this a legit bug or am I missing something? had a similar error with 6.104 but another class

Benjamin10:12:04

mabye 6.103 works because

scicloj/scicloj.ml {:mvn/version "0.2.0"}
loads correctly

respatialized15:12:12

In dtype-next, is there an efficient way of filtering the rows of a 2d tensor by a 1d tensor of boolean values, such that only the true indices are selected, in a manner similar to numpy slicing?

respatialized15:12:01

using tensor/select or datatype/indexed-buffer with boolean values casts them into 0s and 1s, returning the values at those index positions.