Clojurians Log v2

Your code defines symbol datasets twice. The require-python does so and the def does so. In general Clerk does not handle always re-definition of symbols well. Try to use a different name for the def

Eventually there is slight difference in different lib-pythoin versions regarding the precise symbol created on

(require-python `datasets)

As far as I know, both are nearly doing the same, namely creating a symbol called dlai.clerk-error-minimal/datasets Which you can use to call the functions of the module. So you only need one of the two. Probably the second alone works, as it would overwrite in any case the first definition. (at least outside Clerk)

Thanks for looking into this, @carsten.behring. The require-python defines dataset*s* (plural), while the def defines dataset singular, so there should be no collision / double def . My own investigation seems to show that depending on the way I do Python interop, some python source files seem to end up processed by parse-file ( https://github.com/nextjournal/clerk/blob/main/src/nextjournal/clerk/parser.cljc#L397 ). Did not have more time to look into it but would definitely like to understand this more.

I see both in plural in your code

Note, that using a different way of Python interop resolves the issue

(ns dlai.clerk-error-minimal
  (:require [libpython-clj2.require :refer [require-python]]
            [libpython-clj2.python :as py]
            [dlai.core :as core]))

;; require python stuff
(require-python 'datasets)

(def datasets (py/import-module "datasets"))

(py/py.. datasets (load_dataset  "knkarthick/dialogsum"))

could you file an issue ideally with a repro?

Sure thing