diff --git a/knowledge/technology/large_language_model/granite/attribution.txt b/knowledge/technology/large_language_model/granite/attribution.txt new file mode 100644 index 000000000..58dac6486 --- /dev/null +++ b/knowledge/technology/large_language_model/granite/attribution.txt @@ -0,0 +1,5 @@ +Title of work: IBM Granite +Link to work: https://en.wikipedia.org/wiki/IBM_Granite +Revision: https://en.wikipedia.org/wiki/IBM_Granite&oldid=1246833397 +License of the work: CC-BY-SA-4.0 +Creator names: Wikipedia Authors diff --git a/knowledge/technology/large_language_model/granite/qna.yaml b/knowledge/technology/large_language_model/granite/qna.yaml new file mode 100644 index 000000000..9a18eebbe --- /dev/null +++ b/knowledge/technology/large_language_model/granite/qna.yaml @@ -0,0 +1,144 @@ +created_by: cmizurichibmcom +version: 3 +domain: large-language-model +document_outline: Knowledge contribution about the IBM Granite model +seed_examples: + - context: >- + IBM Granite is a series of decoder-only Al foundation models created by + IBM. It was announced on September 7, 2023, and an initial paper was + published 4 days later. + questions_and_answers: + - question: What is IBM Granite + answer: >- + IBM Granite is a series of decoder-only Al foundation models created + by IBM. + - question: When was IBM Granite announced? + answer: September 7, 2023 + - question: What's a series of IBM decoder-only Al foundation models? + answer: IBM Granite + - context: >- + ==Foundation models== + + A foundation model is an AI model trained on broad data at scale such that + it can be adapted to a wide range of downstream tasks. + + Granite's first foundation models were Granite.13b.instruct and + Granite.13b.chat. The "13b" in their name comes from 13 billion, the + amount of parameters they have as models, lesser than most of the larger + models of the time. Later models vary from 3 to 34 billion parameters. + + On May 6, 2024, IBM released the source code of four variations of Granite + Code Models under Apache 2, an open source permissive license that allows + completely free use, modification and sharing of the software, and put + them on Hugging Face for public use. According to IBM's own report, + Granite 8b outperforms Llama 3 on several coding related tasks within + similar range of parameters. + questions_and_answers: + - question: What is a foundation model? + answer: >- + A foundation model is an AI model trained on broad data at scale, + making it adaptable to a wide range of downstream tasks. + - question: >- + What are the names of the first foundation models in the Granite + series? + answer: The first models were Granite.13b.instruct and Granite.13b.chat. + - question: What does the "13b" in Granite.13b.instruct refer to? + answer: The "13b" refers to the 13 billion parameters in the model. + - context: >- + '''IBM Granite''' is a series of decoder-only [[AI]] [[foundation + model|foundation models]] created by [[IBM]]. It was announced on + September 7, 2023. + + ==Foundation models== + + A foundation model is an AI model trained on broad data at scale such that + it can be adapted to a wide range of downstream tasks. + + Granite's first foundation models were Granite.13b.instruct and + Granite.13b.chat. The "13b" in their name comes from 13 billion, the + amount of parameters they have as models, lesser than most of the larger + models of the time. Later models vary from 3 to 34 billion parameters. + + On May 6, 2024, IBM released the source code of four variations of Granite + Code Models under Apache 2, an open source permissive license that allows + completely free use, modification and sharing of the software, and put + them on Hugging Face for public use. According to IBM's own report, + Granite 8b outperforms Llama 3 on several coding related tasks within + similar range of parameters. + questions_and_answers: + - question: What is a foundation model? + answer: >- + A foundation model is an AI model trained on broad data at scale, + making it adaptable to a wide range of downstream tasks. + - question: >- + What are the names of the first foundation models in the Granite + series? + answer: The first models were Granite.13b.instruct and Granite.13b.chat. + - question: What does the "13b" in Granite.13b.instruct refer to? + answer: The "13b" refers to the 13 billion parameters in the model. + - context: >- + '''IBM Granite''' is a series of decoder-only created by [[IBM]]. It was + announced on September 7, 2023. + + ==Foundation models== + + A foundation model is an AI model trained on broad data at scale such that + it can be adapted to a wide range of downstream tasks. + + Granite's first foundation models were Granite.13b.instruct and + Granite.13b.chat. The "13b" in their name comes from 13 billion, the + amount of parameters they have as models, lesser than most of the larger + models of the time. Later models vary from 3 to 34 billion parameters. + + On May 6, 2024, IBM released the source code of four variations of Granite + Code Models under Apache 2, an open source permissive license that allows + completely free use, modification and sharing of the software, and put + them on Hugging Face for public use. According to IBM's own report, + Granite 8b outperforms Llama 3 on several coding related tasks within + similar range of parameters. + questions_and_answers: + - question: How many model parameters do Granite models vary between? + answer: Granite models vary from 3 billion to 34 billion parameters. + - question: >- + When did IBM release the source code for 4 variations of Granite Code + Models? + answer: IBM released the corresponding source code on May 6, 2024. + - question: Under which license type did IBM release the Granite Code Models? + answer: IBM released them under the Apache 2 open source license type. + - context: >- + IBM Granite is a series of decoder-only [[AI]] [[foundation + model|foundation models]] created by [[IBM]]. It was announced on + September 7, 2023. + + ==Foundation models== + + A foundation model is an AI model trained on broad data at scale such that + it can be adapted to a wide range of downstream tasks. + + Granite's first foundation models were Granite.13b.instruct and + Granite.13b.chat. The "13b" in their name comes from 13 billion, the + amount of parameters they have as models, lesser than most of the larger + models of the time. Later models vary from 3 to 34 billion parameters. + + On May 6, 2024, IBM released the source code of four variations of Granite + Code Models under Apache 2, an open source permissive license that allows + completely free use, modification and sharing of the software, and put + them on Hugging Face for public use. According to IBM's own report, + Granite 8b outperforms Llama 3 on several coding related tasks within + similar range of parameters. + questions_and_answers: + - question: Where did IBM make the Granite Code Models available for public use? + answer: IBM made them available on Hugging Face. + - question: >- + What does the Apache 2 license allow users to do with the Granite Code + Models? + answer: >- + The Apache 2 license allows free use, modification, and sharing of the + software. + - question: According to IBM, which model does Granite 8b outperform? + answer: Granite 8b outperforms Llama 3 on several coding-related tasks. +document: + repo: https://github.com/cmizurichibmcom/taxonomy-knowledge-docs + commit: d3ad06a9f47b426127533b0d3fc62a17b777d546 + patterns: + - IBM_Granite-20241021T104730005.md