GPUMODE/KernelBook

1June 9 Researcher Reciprocity License2Version 1.03dated June 9, 20264 5This is a license (the "License") between you ("You") and GPU Mode and the6KernelBook dataset contributors ("Licensor"). This License adapts the Open7Responsible AI License Data ("Open RAIL-D") pattern for a dataset artifact and8adds the Researcher Reciprocity use restriction in Attachment A. It is intended9to have an open and permissive character while preserving reciprocal research10access when the Dataset is used to train or improve AI systems.11 12If you train on it, you let us generate.13 14Section I: Preamble15 16`KernelBook` is a dataset of pairs of PyTorch programs and equivalent Triton17code generated by TorchInductor. The Dataset contains code pairs, metadata,18and related materials for GPU kernel research and development.19 20Licensor wishes to promote collaboration, open research, education,21benchmarking, and broad reuse of the Dataset. Licensor also wishes to avoid a22one-way bargain in which researchers and contributors publish ideas and code23that are used to improve AI systems, while the providers of those AI systems24then prohibit those same researchers from generating outputs, evaluating the25systems, benchmarking them, publishing research, or exploring their own ideas.26 27This License therefore grants broad rights to use the Dataset, subject to28attribution and the use-based restriction in Attachment A.29 30Section II: Definitions31 321. "License" means these terms and conditions for use, reproduction, and33Distribution.34 352. "Dataset" means the files, records, metadata, documentation, and other36materials distributed with this License.37 383. "Output" means the results of operating a model, service, application, or39other system.40 414. "Model" means any machine-learning or artificial-intelligence based42assemblies, including model weights, checkpoints, parameters, optimizer states,43adapters, embedding systems, agents, APIs, hosted services, or other systems44that are trained, tuned, evaluated, benchmarked, or otherwise used in connection45with the Dataset.46 475. "Derivatives of the Dataset" means all modifications, transformations,48annotations, translations, extracts, subsets, compilations, arrangements, or49other works based on the Dataset.50 516. "Derivatives of a Model" means all modifications to a Model, works based on52a Model, or any other model that is created or initialized by transfer of53patterns of weights, parameters, activations, embeddings, outputs, or other54representations of the Model, including distillation methods and methods based55on synthetic data generated by the Model.56 577. "Training Use" means using the Dataset, in whole or in part, to train,58pretrain, fine-tune, post-train, align, distill, evaluate for training,59benchmark for training, generate synthetic data for training, construct60embeddings for training, rank or filter examples for training, or otherwise61improve the weights, behavior, capabilities, or performance of a Model or62Derivatives of a Model.63 648. "Covered Model" means any Model or Derivatives of a Model that is trained,65fine-tuned, distilled, aligned, evaluated for training, benchmarked for66training, or otherwise improved through Training Use of the Dataset.67 689. "Distribution" means any transmission, reproduction, publication, hosting,69or other sharing of the Dataset, Derivatives of the Dataset, a Covered Model, or70Derivatives of a Covered Model to a third party, including making any of them71available by electronic or remote means, such as API-based or web access.72 7310. "Licensor" means GPU Mode, the dataset maintainers, and any contributor who74has authority to license their contribution under these terms.75 7611. "You" or "Your" means an individual or legal entity exercising permissions77granted by this License or making use of the Dataset for any purpose.78 7912. "Third Parties" means individuals or legal entities that are not under80common control with Licensor or You.81 8213. "Authorized Researchers" means GPU Mode, the dataset maintainers, dataset83contributors, and any researchers or organizations that GPU Mode designates in84writing for purposes of generating outputs from, evaluating, benchmarking,85auditing, criticizing, or publishing research about a Covered Model.86 8714. "Ordinary Users" means the general class of users to whom You make a88Covered Model available, including through a public product, commercial product,89research release, API, hosted service, preview, beta, or gated access program.90 91Section III: Intellectual Property Rights92 932. Grant of Copyright License. Subject to the terms and conditions of this94License, each Licensor grants You a worldwide, non-exclusive, no-charge,95royalty-free copyright license to reproduce, prepare derivative works of,96publicly display, publicly perform, sublicense, and distribute the Dataset and97Derivatives of the Dataset.98 993. No Patent License. This License does not grant any patent license.100 101Section IV: Conditions of Usage, Distribution, and Redistribution102 1034. Distribution and Redistribution. You may reproduce and distribute copies of104the Dataset or Derivatives of the Dataset in any medium, with or without105modifications, provided that You meet the following conditions:106 1074.1. You must give Third Party recipients of the Dataset or Derivatives of the108Dataset a copy of this License or a clear link to it.109 1104.2. You must retain reasonable copyright, license, and attribution notices,111excluding notices that do not pertain to any part of the Dataset or Derivatives112of the Dataset.113 1144.3. You must give reasonable attribution to GPU Mode and the KernelBook dataset.115Reasonable attribution includes, where practical, the dataset name, a link to116the dataset source, and any citation requested in the Dataset documentation.117 1184.4. You must cause any modified files, datasets, or documentation that You119Distribute to carry prominent notices stating that You changed them.120 1214.5. You may add Your own copyright statement to Your modifications and may122provide additional or different license terms for Your independent additions,123annotations, analyses, software, models, outputs, or other works, provided that124Your use, reproduction, and Distribution of the Dataset otherwise complies with125this License.126 1275. Use-Based Restrictions. The restriction set forth in Attachment A is a128use-based restriction. You may not use the Dataset, Derivatives of the Dataset,129Covered Models, or Derivatives of Covered Models for the restricted use130specified in Attachment A.131 132For Training Use, the use-based restriction in Attachment A must be included as133an enforceable provision in any legal agreement, terms of use, acceptable use134policy, license, or other terms governing the use or Distribution of a Covered135Model or Derivatives of a Covered Model. You must give notice to subsequent136users that the Covered Model or Derivatives of the Covered Model are subject to137Attachment A.138 1396. Outputs. Except as stated in this License, Licensor claims no rights in the140Output You generate using a Covered Model. You are accountable for the Output141You generate and its subsequent uses. No use of the Output may contravene this142License.143 144Section V: Other Provisions145 1467. No Endorsement. Nothing in this License permits You to use Licensor's names,147logos, trademarks, or service marks to imply endorsement, sponsorship, or148approval.149 1508. Third-Party Rights. The Dataset may include material submitted by third151parties. This License applies only to rights that Licensor has authority to152license. You are responsible for complying with any third-party rights, privacy153obligations, laws, or regulations that apply to Your use.154 1559. Disclaimer of Warranty. Unless required by applicable law or agreed to in156writing, Licensor provides the Dataset on an "AS IS" BASIS, WITHOUT WARRANTIES157OR CONDITIONS OF ANY KIND, either express or implied, including warranties or158conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, FITNESS FOR A PARTICULAR159PURPOSE, ACCURACY, AVAILABILITY, OR ABSENCE OF DEFECTS. You are solely160responsible for determining the appropriateness of using or redistributing the161Dataset and assume any risks associated with Your exercise of permissions under162this License.163 16410. Limitation of Liability. To the maximum extent permitted by law, in no event165and under no legal theory, whether in tort, contract, or otherwise, unless166required by applicable law or agreed to in writing, shall any Licensor or167contributor be liable to You for damages, including direct, indirect, special,168incidental, consequential, exemplary, or punitive damages arising as a result of169this License or out of the use or inability to use the Dataset, even if such170Licensor or contributor has been advised of the possibility of such damages.171 17211. Accepting Warranty or Additional Liability. While redistributing the Dataset173or Derivatives of the Dataset, You may choose to offer, and charge a fee for,174acceptance of support, warranty, indemnity, or other liability obligations or175rights consistent with this License. However, in accepting such obligations, You176may act only on Your own behalf and on Your sole responsibility, not on behalf177of any Licensor or contributor, and only if You agree to indemnify, defend, and178hold each Licensor and contributor harmless for any liability incurred by, or179claims asserted against, such Licensor or contributor by reason of Your180accepting any such warranty or additional liability.181 18212. Termination. If You violate this License, Your rights under it terminate183automatically. For violations other than violations of Attachment A, Your rights184are reinstated if You cure the violation within 30 days after discovering it or185receiving written notice from Licensor. For violations of Attachment A involving186a Covered Model, Your Training Use rights terminate automatically as to the187affected Covered Model and may be reinstated only if Licensor provides written188reinstatement or waiver.189 19013. Severability. If any provision of this License is held invalid, illegal, or191unenforceable, the remaining provisions remain valid as if the provision had not192been set forth. The unenforceable provision will be interpreted or reformed only193to the minimum extent necessary to make it enforceable while preserving its194purpose.195 19614. Additional Permission. Licensor may grant additional permissions,197exceptions, waivers, commercial terms, or private licenses in writing. Those198permissions apply only to the recipient and scope stated in the written grant.199 200End of Terms and Conditions201 202Attachment A203Use Restriction: Researcher Reciprocity for Training Use204 205You agree not to use the Dataset or Derivatives of the Dataset for Training Use206if You make the resulting Covered Model or Derivatives of the Covered Model207available under terms, policies, technical measures, access rules, account208restrictions, acceptable-use rules, or other conditions that prohibit, penalize,209or materially burden Authorized Researchers from:210 2111. generating outputs from the Covered Model;212 2132. evaluating, auditing, red-teaming, or benchmarking the Covered Model;214 2153. comparing the Covered Model to other systems;216 2174. publishing research, criticism, measurements, benchmark results, or analysis218concerning the Covered Model; or219 2205. using the Covered Model to explore, test, or develop their own research221ideas.222 223This access must be available on materially equal terms to those offered to224Ordinary Users of the Covered Model, subject only to neutral limits that apply225equally to Ordinary Users, such as generally applicable rate limits, payment226terms, safety rules, security rules, and laws.227 228Any terms, policies, technical measures, access rules, account restrictions,229acceptable-use rules, or other conditions that conflict with this Attachment A230make the Covered Model ineligible for the Training Use grant unless Licensor has231waived the conflict in writing.232 233You may not suspend, ban, throttle, sue, threaten, or otherwise retaliate234against Authorized Researchers solely because they engage in the activities235listed in this Attachment A, provided that their activity complies with236generally applicable law and neutral safety or security rules that are also237applied to Ordinary Users.238