From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mp10.migadu.com ([2001:41d0:8:6d80::]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)) by ms9.migadu.com with LMTPS id CPFMIz3MLmSxVwEASxT56A (envelope-from ) for ; Thu, 06 Apr 2023 15:42:21 +0200 Received: from aspmx1.migadu.com ([2001:41d0:8:6d80::]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits)) by mp10.migadu.com with LMTPS id eD81Ij3MLmR6aAAAG6o9tA (envelope-from ) for ; Thu, 06 Apr 2023 15:42:21 +0200 Received: from lists.gnu.org (lists.gnu.org [209.51.188.17]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by aspmx1.migadu.com (Postfix) with ESMTPS id 4AC5A2C9EE for ; Thu, 6 Apr 2023 15:42:21 +0200 (CEST) Received: from localhost ([::1] helo=lists1p.gnu.org) by lists.gnu.org with esmtp (Exim 4.90_1) (envelope-from ) id 1pkPs7-00072P-9A; Thu, 06 Apr 2023 09:41:51 -0400 Received: from eggs.gnu.org ([2001:470:142:3::10]) by lists.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1pkPs4-00071r-SB for guix-devel@gnu.org; Thu, 06 Apr 2023 09:41:49 -0400 Received: from mout02.posteo.de ([185.67.36.66]) by eggs.gnu.org with esmtps (TLS1.2:ECDHE_RSA_AES_256_GCM_SHA384:256) (Exim 4.90_1) (envelope-from ) id 1pkPs2-0005Os-Hm for guix-devel@gnu.org; Thu, 06 Apr 2023 09:41:48 -0400 Received: from submission (posteo.de [185.67.36.169]) by mout02.posteo.de (Postfix) with ESMTPS id 8EC5124057C for ; Thu, 6 Apr 2023 15:41:44 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=posteo.net; s=2017; t=1680788504; bh=dAEd9kHudG/HhcZ0jK7b2lu1hm/klUirCPByJCuiHb8=; h=Date:From:To:Subject:From; b=rsRoVQeSxTFjwfDUXD73Qf40VbBBKpzpBcKedeuB5zyk22EI/wMCObZk9V7QJi2Ft tVFE0BaaqnzhSPu0AgY1PzLGgkvlJY3h7QCXnsln8WvRDrrJRCm7AgJNTfXhYwj7op yeWYb8wA72gG8qujtiGX+wI3RlWQsMVRvoKzcTT+/jVR4ntCxmPKBQJa61VVL/M9QW 3BlMOQMngvBwNWq7ZUdWWmPyKy36q7vjRYccDFdNttPoIpno9LzUnkih+30+87VwlW hRsVluiTZNZgMRvHvukZFWfx0Xpc4s1KIdfjsrXtkGFAPXzY7khQsduygwlL3D31Zv srQXDZzyarVSA== Received: from customer (localhost [127.0.0.1]) by submission (posteo.de) with ESMTPSA id 4PsjMq3H59z9rxB; Thu, 6 Apr 2023 15:41:43 +0200 (CEST) Date: Thu, 06 Apr 2023 13:41:40 +0000 From: Kyle To: guix-devel@gnu.org, Simon Tournier , Ryan Prior , Nicolas Graves , "licensing@fsf.org" Subject: =?US-ASCII?Q?Re=3A_Guidelines_for_pre-trained?= =?US-ASCII?Q?_ML_model_weight_binaries_=28Was?= =?US-ASCII?Q?_re=3A_Where_should_we_put_machi?= =?US-ASCII?Q?ne_learning_model_parameters=3F=29?= In-Reply-To: <868rf5e71j.fsf@gmail.com> References: <868rf5e71j.fsf@gmail.com> Message-ID: <3A47DA6E-C392-4989-AFD4-20660D968415@posteo.net> MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Received-SPF: pass client-ip=185.67.36.66; envelope-from=kyle@posteo.net; helo=mout02.posteo.de X-Spam_score_int: -43 X-Spam_score: -4.4 X-Spam_bar: ---- X-Spam_report: (-4.4 / 5.0 requ) BAYES_00=-1.9, DKIM_SIGNED=0.1, DKIM_VALID=-0.1, DKIM_VALID_AU=-0.1, DKIM_VALID_EF=-0.1, RCVD_IN_DNSWL_MED=-2.3, RCVD_IN_MSPIKE_H2=-0.001, SPF_HELO_NONE=0.001, SPF_PASS=-0.001 autolearn=ham autolearn_force=no X-Spam_action: no action X-BeenThere: guix-devel@gnu.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: "Development of GNU Guix and the GNU System distribution." List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: guix-devel-bounces+larch=yhetil.org@gnu.org Sender: guix-devel-bounces+larch=yhetil.org@gnu.org X-Migadu-Country: US X-Migadu-Flow: FLOW_IN ARC-Seal: i=1; s=key1; d=yhetil.org; t=1680788541; a=rsa-sha256; cv=none; b=d2v/Rqw/hbrwuf4rth6kccTNuWDoDcKqRbk+F4k5z4HReSd/X4/tMqvcUzvQ5/hRNS5uAe 290a65ci9+h22Px//yb4THCKe6S1LCUr22fvaUGe5p4TCqBpRd5f1ZkhHLHIaGyIP7fIyu 8/IgxlbPWP5up51jdrmFAwv5auN4upeqi+DRq37F1pJMr+Uc6JKtVefwQP1SMGzc2Hn1hk 4otAZaDr2Tt2M6ZuKM0ogHkY8YGFIxn4z2lQAtpFg6VwWkhwu5Q6E4YddvgCqHzIdVxEoP +0wKOhwAlid4QZDi8rjUYxBKHGdVAG3L/07d5AxWk77w86klgDTtqt+mRP/3lg== ARC-Authentication-Results: i=1; aspmx1.migadu.com; dkim=pass header.d=posteo.net header.s=2017 header.b=rsRoVQeS; dmarc=pass (policy=none) header.from=posteo.net; spf=pass (aspmx1.migadu.com: domain of "guix-devel-bounces+larch=yhetil.org@gnu.org" designates 209.51.188.17 as permitted sender) smtp.mailfrom="guix-devel-bounces+larch=yhetil.org@gnu.org" ARC-Message-Signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=yhetil.org; s=key1; t=1680788541; h=from:from:sender:sender:reply-to:subject:subject:date:date: message-id:message-id:to:to:cc:mime-version:mime-version: content-type:content-type: content-transfer-encoding:content-transfer-encoding: in-reply-to:in-reply-to:references:references:list-id:list-help: list-unsubscribe:list-subscribe:list-post:dkim-signature; bh=OSQ9kFmC2AtQueIH6WDtuahs8wXD1NnFPjuSOIUOd5I=; b=EwpL8E/OThqfHSQRx6qmhEOOgLjdbdkT9SCqq7hf/WG41PmbfsOs7KAlYzO0jdu+Y8cRq9 l2kOkG2yStOHgGa8Tdx/WOg9rCHOO+l1U2kHmFEj46wKTIOzkZ+E+sk3OEKVumA3MiFj3F mfLIAMB3hhXo/0WC4r0wwVsAtVxh5bX1mb4imr1rR1A3Hy646r08qwMUN+t3J4Q0Y4qCnM eMdDOrIktAi847Q/vttOXCY1nXOc/KmyEvckNOkUMaixvjiSifHadoQyRRssNXvoSsajcZ +25tl3pu0/N2YRQ7/J5+K05hC0R/HJximAHYxnamQ94GJSIpQWHzSAEdHsiNdg== X-Migadu-Spam-Score: 2.99 X-Migadu-Scanner: scn1.migadu.com Authentication-Results: aspmx1.migadu.com; dkim=pass header.d=posteo.net header.s=2017 header.b=rsRoVQeS; dmarc=pass (policy=none) header.from=posteo.net; spf=pass (aspmx1.migadu.com: domain of "guix-devel-bounces+larch=yhetil.org@gnu.org" designates 209.51.188.17 as permitted sender) smtp.mailfrom="guix-devel-bounces+larch=yhetil.org@gnu.org" X-Spam-Score: 2.99 X-Migadu-Queue-Id: 4AC5A2C9EE X-TUID: 9D2NN2dpdnPj >Since it is computing, we could ask about the bootstrap of such >generated data=2E I think it is a slippery slope because it is totally >not affordable to re-train for many cases: (1) we would not have the >hardware resources from a practical point of view,, (2) it is almost >impossible to tackle the source of indeterminism (the optimization is >too entailed with randomness)=2E=20 I have only seen situations where the optimization is "too entailed with r= andomness" when models are trained on proprietary GPUs with specific settin= gs=2E Otherwise, pseudo-random seeds are perfectly sufficient to remove the= indeterminism=2E=20 =3D> https://discourse=2Ejulialang=2Eorg/t/flux-reproducibility-of-gpu-exp= eriments/62092 Many people think that "ultimate" reproducibility is not a practical eithe= r=2E It's always going to be easier in the short term to take shortcuts whi= ch make conclusions dependent on secret sauce which few can understand=2E =3D> https://hpc=2Eguix=2Einfo/blog/2022/07/is-reproducibility-practical/ From my point of view, pre-trained >weights should be considered as the output of a (numerical) experiment, >similarly as we include other experimental data (from genome to >astronomy dataset)=2E I think its a stretch to consider a data compression as an experiment=2E I= n experiments I am always finding mistakes which confuse the interpretation= hidden by prematurely compressing data, e=2Eg=2E by taking inappropriate a= verages=2E Don't confuse the actual experimental results with dubious data = processing steps=2E