[B! data] deejayrokaã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯

deejayroka id:deejayroka

dataã«é–¢ã™ã‚‹deejayrokaã®ãƒ–ãƒƒã‚¯ãƒžãƒ¼ã‚¯ (33)

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

${{author_name}}$
{{author_name}}{{created}}
{{ #comment }}{{ comment }}{{ /comment }}
- {{ label }}

${{author_name}}$

{{{comment_expanded}}}

{{label}}

{{#is_bookmark}}ãƒªã‚¹ãƒˆ{{/is_bookmark}}{{^is_bookmark}}ãƒªãƒ³ã‚¯{{/is_bookmark}}

MetaãŒ1å…†5000å„„å††ã‹ã‘ã¦åŒç¤¾æœ€å¤§è¦æ¨¡ã®AIãƒ‡ãƒ¼ã‚¿ã‚»ãƒ³ã‚¿ãƒ¼ã‚’ãƒ«ã‚¤ã‚¸ã‚¢ãƒŠå·žã«å»ºè¨äºˆå®š
MetaãŒãƒ«ã‚¤ã‚¸ã‚¢ãƒŠå·žåŒ—éƒ¨ãƒªãƒƒãƒãƒ©ãƒ³ãƒ‰ã«ã‚ã‚‹400ä¸‡å¹³æ–¹ãƒ•ã‚£ãƒ¼ãƒˆ(ç´„37ä¸‡å¹³æ–¹ãƒ¡ãƒ¼ãƒˆãƒ«)ã®åœŸåœ°ã«ã€åŒç¤¾æœ€å¤§è¦æ¨¡ã¨ãªã‚‹AIãƒ‡ãƒ¼ã‚¿ã‚»ãƒ³ã‚¿ãƒ¼ã‚’è¨ˆç”»ã—ã¦ã„ã‚‹ã“ã¨ãŒã‚ã‹ã‚Šã¾ã—ãŸã€‚ãƒ«ã‚¤ã‚¸ã‚¢ãƒŠå·žã«ã‚ˆã‚‹ã¨ã€ã“ã®è¨ˆç”»ã¯ã€Œãƒ«ã‚¤ã‚¸ã‚¢ãƒŠå·žå²ä¸Šæœ€å¤§è¦æ¨¡ã®æ°‘é–“è³‡æœ¬æŠ•è³‡ã€ã«ãªã‚‹ã¨ã®ã“ã¨ã§ã€è²»ç”¨ã¯100å„„ãƒ‰ãƒ«(ç´„1å…†5000å„„å††)è¦æ¨¡ã«ãªã‚‹è¦‹è¾¼ã¿ã§ã™ã€‚ Landry Announces Meta Selects North Louisiana as Site of $10 Billion Artificial Intelligence Optimized Data Center | Office of Governor Jeff Landry https://gov.louisiana.gov/news/4697 s Meta selects Entergy, Northeast Louisiana as site of
deejayroka 2024/12/08
ai

ãƒ‡ãƒ¼ã‚¿

data
ãƒªãƒ³ã‚¯
Introducing LlamaCloud and LlamaParse â€” LlamaIndex, Data Framework for LLM Applications
Today is a big day for the LlamaIndex ecosystem: we are announcing LlamaCloud, a new generation of managed parsing, ingestion, and retrieval services, designed to bring production-grade context-augmentation to your LLM and RAG applications. Using LlamaCloud as an enterprise AI engineer, you can focus on writing the business logic and not on data wrangling. Process large volumes of production data,
deejayroka 2024/08/29
LLM

data

RAG
ãƒªãƒ³ã‚¯
LF AI & Data Landscape
LF AI & Data Foundation Interactive Landscape The LF AI & Data Foundation landscape (png, pdf) is dynamically generated below. It is modeled after the CNCF landscape and based on the same open source code. Please open a pull request to correct any issues. Greyed logos are not open source. Last Updated: 2024-09-11T07:29:18.
deejayroka 2024/07/26
ai

data

MLOps

ML
ãƒªãƒ³ã‚¯
4 Advanced RAG Algorithms to Optimize Retrieval
Welcome to Lesson 5 of 11 in our free course series, LLM Twin: Building Your Production-Ready AI Replica. Youâ€™ll learn how to use LLMs, vector DVs, and LLMOps best practices to design, train, and deploy a production ready â€œLLM twinâ€ of yourself. This AI character will write like you, incorporating your style, personality, and voice into an LLM. For a full overview of course objectives and prerequi
deejayroka 2024/06/09
â€œRAGâ€

data
ãƒªãƒ³ã‚¯
Microsoft Fabric Vs Databricks: Choosing the Right Data Solution
deejayroka 2024/04/01
â€Microsoft Fabric Vs Databricksâ€

AI

Microsoft

Azure

AWS

data

cloud

Analytics
ãƒªãƒ³ã‚¯
Video generation models as world simulators
We explore large-scale training of generative models on video data. Specifically, we train text-conditional diffusion models jointly on videos and images of variable durations, resolutions and aspect ratios. We leverage a transf ormer architecture that operates on spacetime patches of video and image latent codes. Our largest model, Sora, is capable of generating a minute of high fidelity video. Ou
deejayroka 2024/02/16
data

ai

sora
ãƒªãƒ³ã‚¯
ã€npmã€‘ã‚ãƒ£ãƒƒã‚·ãƒ¥ãŒå£Šã‚Œã¦npm installã«å¤±æ•—ã™ã‚‹ - Qiita
ã¯ã˜ã‚ã« ä¹…ã—ã¶ã‚Šã«create-react-appã‚’å©ã„ãŸã‚‰ã€å¤±æ•—ã—ã¾ã—ãŸã€‚ èª¿ã¹ãŸã“ã¨ã‚’ç°¡å˜ã«ãƒ¡ãƒ¢ãƒ¡ãƒ¢ã€‚ â€»èªè˜é–“é•ã£ã¦ã„ã‚‹æ‰€ãŒã‚ã‚Œã°ã€ã”æŒ‡æ‘˜ãŠé¡˜ã„ã—ã¾ã™ã€‚ ç¾è±¡ ä¸‹è¨˜ã‚³ãƒžãƒ³ãƒ‰å®Ÿè¡Œæ™‚ã«ã‚¨ãƒ©ãƒ¼ãŒç™ºç”Ÿã€‚ ã‚³ãƒžãƒ³ãƒ‰ï¼š npx create-react-app {ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆå} --template typescript ã‚¨ãƒ©ãƒ¼ï¼š ERR! cb() never called! ç’°å¢ƒã¯ä»¥ä¸‹ã®é€šã‚Šã€‚ ãƒ»OS ï¼šDockerä¸Šã®Ubuntu:20.04 ãƒ»nvm ï¼š0.38.0 ãƒ»nodeï¼š14.17.4 ãƒ»npm ï¼š6.14.14 åŽŸå› ã¾ã¨ã‚ã¦ãã ã•ã£ã¦ã„ã‚‹è¨˜äº‹ã«ã‚ˆã‚‹ã¨ã€ã©ã†ã‚„ã‚‰npmã®ã‚ãƒ£ãƒƒã‚·ãƒ¥ã®ä¸å…·åˆã‚‰ã—ã„ã€‚ï¼ˆã‚ãƒ£ãƒƒã‚·ãƒ¥ãŒå£Šã‚Œã¦ã„ã‚‹ï¼Ÿï¼‰ ã“ã®ç¾è±¡ã¯create-react-appã«é™ã£ãŸã‚‚ã®ã§ã¯ãªãã€npm installã§ã‚‚ç™ºç”Ÿã™ã‚‹ãã†ã§ã™ã€‚ npmã®ã‚ãƒ£ãƒƒã‚·ãƒ¥ã¨ã¯ï¼Ÿ npmã®Docã«
deejayroka 2024/02/08
â€œnpm cache verify â€

data

react

qiita
ãƒªãƒ³ã‚¯
cifar10 ã§ Test Time Augmentation (TTA) ã®å®Ÿé¨“ - Qiita
Deep Learning ã®ã€ç‰¹ã«ç”»åƒåˆ†é¡žç³»ã®ã‚¿ã‚¹ã‚¯ã§ã‚ˆãä½¿ç”¨ã•ã‚Œã‚‹ã€ Test Time Augmentation (TTA) ã¨ã„ã† Augmentation ã®æ‰‹æ³•ã‚’å‹äºº (@hattan0523) ã«èžã„ãŸã®ã§ã€å®Ÿéš›å¤‰ã‚ã‚‹ã‚‚ã®ã‹ã¨æ€ã£ã¦ã‚«ã‚¸ãƒ¥ã‚¢ãƒ«ã«è©¦ã—ã¦ã¿ã¾ã—ãŸã€‚ ãªã‚“ã‹ã‚³ãƒ³ãƒšã¨ã‹ã§ã¯å¿…é ˆã‚‰ã—ã„ã€‚ â€» é–“é•ã„ãªã©ã”æŒ‡æ‘˜æ“è¿Žã§ã™ã€‚ Test Time Augmentation (TTA) ã¨ã¯ æ©Ÿæ¢°å¦ç¿’ã€ç‰¹ã« Deep Learning ã§ã¯ã€å¤§ããªãƒ¢ãƒ‡ãƒ«ã‚’å¦ç¿’ã•ã›ã‚‹ãŸã‚ã«ãŸãã•ã‚“ã®è¨“ç·´ãƒ‡ãƒ¼ã‚¿ãŒå¿…è¦ã«ãªã‚Šã¾ã™ã€‚ ãƒ‡ãƒ¼ã‚¿ãŒå¢—ãˆã‚‹ã»ã©ç²¾åº¦ãŒä¸ŠãŒã£ãŸã‚Šã€è¤‡é›‘ãªãƒ¢ãƒ‡ãƒ«ã‚‚å¦ç¿’ã§ãã‚‹ã‚ˆã†ã«ãªã‚‹ã“ã¨ãŒçŸ¥ã‚‰ã‚Œã¦ã„ã¾ã™ã€‚ ãã“ã§ã€è¨“ç·´ç”¨ãƒ‡ãƒ¼ã‚¿ã‚’å°‘ã—åŠ å·¥ã—åˆ¥ã®ãƒ‡ãƒ¼ã‚¿ã¨ã—ã¦ãƒ‡ãƒ¼ã‚¿ã‚»ãƒƒãƒˆã«åŠ ãˆã‚‹ã“ã¨ã§ã€å¦ç¿’ã«ä½¿ãˆã‚‹ãƒ‡ãƒ¼ã‚¿ã®é‡ã‚’å¢—ã‚„ã™è©¦ã¿ãŒã‚ˆãè¡Œã‚ã‚Œã¦ãŠã‚Šã€ Data Augmentation ã¨å‘¼ã°ã‚Œã¾ã™
deejayroka 2024/02/06
â€œæŽ¨è«–æ™‚ã®ãƒ‡ãƒ¼ã‚¿ã‚’åŠ å·¥ã™ã‚‹ã“ã¨ã§ã€1ã¤ã®ãƒ‡ãƒ¼ã‚¿ã‚’æ•°ç¨®é¡žä½œæˆã—ã¾ã™ã€‚ ãã®æ•°ç¨®é¡žã®ãƒ‡ãƒ¼ã‚¿ã«å¯¾ã—ã¦å¦ç¿’ã—ãŸãƒ¢ãƒ‡ãƒ«ã§æŽ¨è«–ã‚’è¡Œã„ã€çµæžœã®å¹³å‡ã‚„å¤šæ•°æ±ºã‚’ã¨ã£ã¦ãã‚Œã‚’æŽ¨è«–çµæžœã¨ã™ã‚‹ã“ã¨ã§ç²¾åº¦ã‚’ä¸Šã’ã¾ã™â€

data

TensorFlow

ãƒ¢ãƒ‡ãƒ«

å¦ç¿’

ãƒ‡ãƒ¼ã‚¿

ãƒ†ã‚¹ãƒˆ
ãƒªãƒ³ã‚¯
ãƒ‡ãƒ¼ã‚¿ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ç•Œéšˆã§è©±é¡Œã®dbtï¼ˆdata build toolï¼‰ã®ã¾ã¨ã‚ - Qiita
æ¦‚è¦ ãƒ‡ãƒ¼ã‚¿ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ç•Œéšˆã§è©±é¡Œã¨ãªã£ã¦ããŸdbtã«é–¢ã™ã‚‹èª¿æŸ»å†…å®¹ã‚’å…±æœ‰ã—ã¾ã™ã€‚ OSSç‰ˆã®dbtã§ã‚‚æ©Ÿèƒ½ãŒå……å®Ÿã—ã¦ãŠã‚Šã€ãƒ‡ãƒ¼ã‚¿ã‚¨ãƒ³ã‚¸ãƒ‹ã‚¢ã®æ–¹ã«ã¯ä¸€éƒ¨ã®æ©Ÿèƒ½ã®åˆ©ç”¨ã ã‘ã§ã‚‚ä¾¡å€¤ãŒã‚ã‚‹ãƒ„ãƒ¼ãƒ«ã§ã™ã€‚ å¿…è¦ã«å¿œã˜ã¦ã€ã“ã®è¨˜äº‹ã‚’æ›´æ–°ã—ã¾ã™ã€‚ dbtï¼ˆData Build Toolï¼‰ã¨ã¯ dbtã¨ã¯Data Build Toolã®ç•¥ã§ã‚ã‚Šã€dbt Labsç¤¾ãŒé–‹ç™ºã—ã¦ã„ã‚‹ãƒ‡ãƒ¼ã‚¿é€£æºã®ELTã«ãŠã‘ã‚‹Tï¼ˆãƒ‡ãƒ¼ã‚¿ã‚¹ãƒˆã‚¢å†…ã§ã®ãƒ‡ãƒ¼ã‚¿å¤‰æ›ï¼‰ã®å®Ÿæ–½ã‚’ã‚µãƒãƒ¼ãƒˆã™ã‚‹ãƒ„ãƒ¼ãƒ«ã§ã™ã€‚ æ©Ÿèƒ½ã‚’ã–ã£ãã‚Šæ•´ç†ã™ã‚‹ã¨ä¸‹è¨˜ã®ã‚‚ã®ãŒã‚ã‚Šã€SQLæ–‡ã®ã‚³ãƒ³ãƒãƒ¼ãƒãƒ³ãƒˆåŒ–ã‚„ãƒ‰ã‚ãƒ¥ãƒ¡ãƒ³ãƒˆã®è‡ªå‹•ç”Ÿæˆã«ç‰¹ç•°æ€§ãŒã‚ã‚Šã¾ã™ã€‚ æœ‰å„Ÿç‰ˆã§ã‚ã‚‹dbt cloudã¨OSSç‰ˆã§ã‚ã‚‹dbt CoreãŒã‚ã‚Šã¾ã™ã€‚dbt cloudã«ã¦ã€developerç‰ˆã¨ã„ã†å€‹äººæ¤œè¨¼ç”¨ã®ç„¡å„Ÿãƒ©ã‚¤ã‚»ãƒ³ã‚¹ã‚‚ã‚ã‚‹ã®ã§ã€ã¨ã‚Šã‚ãˆãšè§¦ã‚ŠãŸã„æ–¹ã¯ã“ã¡ã‚‰ãŒãŠã™ã™ã‚ã§ã™ã€‚ å¼•ç”¨å…ƒï¼šdbt Pri
deejayroka 2023/12/11
ãƒ‡ãƒ¼ã‚¿

data

OSS

ãƒ„ãƒ¼ãƒ«
ãƒªãƒ³ã‚¯
Transformerã«ã‚ˆã‚‹æ™‚ç³»åˆ—ãƒ‡ãƒ¼ã‚¿äºˆæ¸¬ã®ã”ç´¹ä»‹ - Platinum Data Blog by BrainPad
æœ¬è¨˜äº‹ã¯ã€å½“ç¤¾ã‚ªã‚¦ãƒ³ãƒ‰ãƒ¡ãƒ‡ã‚£ã‚¢ã€ŒDoorsã€ã«ç§»è»¢ã—ã¾ã—ãŸã€‚ ç´„5ç§’å¾Œã«è‡ªå‹•çš„ã«ãƒªãƒ€ã‚¤ãƒ¬ã‚¯ãƒˆã—ã¾ã™ã€‚ ç™ºå±•ã‚’ç¶šã‘ã‚‹ã€Œè‡ªç„¶è¨€èªžå‡¦ç†ã€æŠ€è¡“ã€‚ãã®ä¸ã§ã‚‚å¹…åºƒãä½¿ç”¨ã•ã‚Œã‚‹æ·±å±¤å¦ç¿’ãƒ¢ãƒ‡ãƒ«Transf ormerã¯ã€è‡ªç„¶è¨€èªžå‡¦ç†ä»¥å¤–ã®éŸ³æ¥½ç”Ÿæˆã‚„ç‰©ä½“æ¤œå‡ºãªã©ã®é ˜åŸŸã§ã‚‚é©ç”¨ã§ãã¾ã™ã€‚æœ¬ãƒ–ãƒã‚°ã§ã¯ã€ Transf ormerã‚’æ™‚ç³»åˆ—ãƒ‡ãƒ¼ã‚¿ã«é©ç”¨ã™ã‚‹æ–¹æ³•ã‚’ã”ç´¹ä»‹ã—ã¾ã™ã€‚ ã“ã‚“ã«ã¡ã¯ã€AIã‚½ãƒªãƒ¥ãƒ¼ã‚·ãƒ§ãƒ³ã‚µãƒ¼ãƒ“ã‚¹éƒ¨ã®äº•å‡ºã¨ç”³ã—ã¾ã™ã€‚ ã“ã®è¨˜äº‹ã§ã¯ã€ç‰¹ã«è‡ªç„¶è¨€èªžå‡¦ç†åˆ†é‡Žã§å¹…åºƒãä½¿ç”¨ã•ã‚Œã‚‹æ·±å±¤å¦ç¿’ãƒ¢ãƒ‡ãƒ«Transf ormerã‚’æ™‚ç³»åˆ—ãƒ‡ãƒ¼ã‚¿ã¸é©ç”¨ã™ã‚‹æ–¹æ³•ã«é–¢ã—ã¦ã”ç´¹ä»‹ã—ã¾ã™ã€‚ ä»¥å‰ã®è¨˜äº‹ã§ã¯ã€Transf ormerã®æ§‹é€ ã‚„ç‰¹å¾´ãªã©ã«ã¤ã„ã¦ã€è‡ªç„¶è¨€èªžå‡¦ç†åˆ†é‡Žã®æ©Ÿæ¢°ç¿»è¨³ã‚’ä¾‹ã¨ã—ã¦ã”ç´¹ä»‹ã—ã¦ãŠã‚Šã¾ã™ã€‚ã¯ã˜ã‚ã«ã€ã“ã¡ã‚‰ã®è¨˜äº‹ã‚’ã”ä¸€èªã—ã¦ã„ãŸã ãã“ã¨ã§ã€ã‚ˆã‚Šæœ¬è¨˜äº‹ã§ã”ç´¹ä»‹ã™ã‚‹å†…å®¹ã«å¯¾ã™ã‚‹ç†è§£ãŒæ·±ã¾ã‚‹ã‹ã¨æ€ã„ã¾ã™ã€‚ Transf orm
deejayroka 2023/12/11
ãƒ‡ãƒ¼ã‚¿

data

deep learning
ãƒªãƒ³ã‚¯
Jina AI Launches World's First Open-Source 8K Text Embedding, Rivaling OpenAI
Embed images and sentences into fixed-length vectors with CLIP
deejayroka 2023/11/06
ã‚ã¨ã§èªã‚€

ai

data

è‡ªç„¶è¨€èªžå‡¦ç†

NLP
ãƒªãƒ³ã‚¯
Evaluating the Ideal Chunk Size for a RAG System using LlamaIndex â€” LlamaIndex, Data Framework for LLM Applications
IntroductionRetrieval-augmented generation (RAG) has introduced an innovative approach that fuses the extensive retrieval capabilities of search systems with the LLM. When implementing a RAG system, one critical parameter that governs the systemâ€™s efficiency and performance is the chunk_size. How does one discern the optimal chunk size for seamless retrieval? This is where LlamaIndex Response Eval
deejayroka 2023/11/02
LLM

data

ãƒ™ã‚¯ã‚¿ãƒ¼
ãƒªãƒ³ã‚¯
Data, Design and Government
æ—¥æœ¬ã®ãƒ‡ã‚¸ã‚¿ãƒ«ç¤¾ä¼šå®Ÿç¾ã®å¸ä»¤å¡”ã¨ã—ã¦ã€ãƒ‡ã‚¸ã‚¿ãƒ«åºã¯æ”¿åºœå†…ã«ãŠã„ã¦ã€ãƒ‡ãƒ¼ã‚¿ã¨æ ¹æ‹ ã«åŸºã¥ã„ãŸæ”¿ç–åˆ¤æ–ãƒ»åŠ¹æžœã®å¯è¦–åŒ–ã®æŽ¨é€²ã‚’å…ˆå°Žã™ã‚‹å½¹å‰²ã‚’æ‹…ã£ã¦ã„ãã¾ã™ã€‚ãã®ä¸€ç’°ã¨ã—ã¦ã€æ”¿ç–ã«é–¢ã‚ã‚‹ãƒ‡ãƒ¼ã‚¿ã‚’ã€Œæ”¿ç–ãƒ‡ãƒ¼ã‚¿ãƒ€ãƒƒã‚·ãƒ¥ãƒœãƒ¼ãƒ‰ã€ã¨ã—ã¦å…¬é–‹ã—ã¦ã„ã¾ã™ã€‚æœ¬ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆã‚’ä¸»ç®¡ã—ã¦ã„ã‚‹ãƒ•ã‚¡ã‚¯ãƒˆï¼†ãƒ‡ãƒ¼ã‚¿ãƒ¦ãƒ‹ãƒƒãƒˆé•·ã®æ¨«ç”°ã¨ãƒ‡ã‚¶ã‚¤ãƒŠãƒ¼ã®å¿—æ°´ãŒæ”¿ç–ã«é–¢ã‚ã‚‹ãƒ‡ãƒ¼ã‚¿ã‚’å…¬é–‹ã™ã‚‹æ„å›³ã‚„ãƒ€ãƒƒã‚·ãƒ¥ãƒœãƒ¼ãƒ‰ã®é–‹ç™ºãƒ—ãƒã‚»ã‚¹ã‚’ç´¹ä»‹ã—ã¾ã™ã€‚ æ”¿ç–ãƒ‡ãƒ¼ã‚¿ãƒ€ãƒƒã‚·ãƒ¥ãƒœãƒ¼ãƒ‰ï¼š https://www.digital.go.jp/resources/govdashboard
deejayroka 2023/10/13
ãƒ‡ã‚¸ã‚¿ãƒ«

data

design

ãƒ‡ãƒ¼ã‚¿
ãƒªãƒ³ã‚¯
Mastering Customer Segmentation with LLM
Letâ€™s see a brief description of the columns of our dataset: age (numeric)job : type of job (categorical: â€œadmin.â€ ,â€unknownâ€,â€unemployedâ€, â€managementâ€, â€housem aidâ€, â€entrepreneurâ€, â€studentâ€, â€œblue-collarâ€, â€self-employedâ€, â€retiredâ€, â€technicianâ€, â€servicesâ€)marital : marital status (categorical: â€œmarriedâ€,â€divorcedâ€,â€singleâ€; note: â€œdivorcedâ€ means divorced or widowed)education (categorical: â€œ
deejayroka 2023/10/01
LLM

data

t-SNE

SHAP

ä¸»æˆåˆ†åˆ†æž
ãƒªãƒ³ã‚¯
ç”ŸæˆAIãƒ»LLMã®ãƒ„ãƒ¼ãƒ«æ‹¡å¼µã«é–¢ã™ã‚‹è«–æ–‡ã®å‹•å‘èª¿æŸ» 2023å¹´8æœˆç‰ˆ - Platinum Data Blog by BrainPad
æœ¬è¨˜äº‹ã¯ã€å½“ç¤¾ã‚ªã‚¦ãƒ³ãƒ‰ãƒ¡ãƒ‡ã‚£ã‚¢ã€ŒDoorsã€ã«ç§»è»¢ã—ã¾ã—ãŸã€‚ ç´„5ç§’å¾Œã«è‡ªå‹•çš„ã«ãƒªãƒ€ã‚¤ãƒ¬ã‚¯ãƒˆã—ã¾ã™ã€‚ ãƒ–ãƒ¬ã‚¤ãƒ³ãƒ‘ãƒƒãƒ‰ã¯ã€LLM/Generative AIã«é–¢ã™ã‚‹ç ”ç©¶ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆã‚’ç«‹ã¡ä¸Šã’ã€ã“ã®ã€ŒPlatinum Data Blogã€ã‚’é€šã˜ã¦LLM/Generative AIã«é–¢ã™ã‚‹ã•ã¾ã–ã¾ãªæƒ…å ±ã‚’ç™ºä¿¡ã‚’ã—ã¦ã„ã¾ã™ã€‚ç¾åœ¨ã¯ã€é€±ã«1å›žç¨‹åº¦ã®é »åº¦ã§ã€ç¤¾å†…ã§å®Ÿæ–½ã—ã¦ã„ã‚‹ç”ŸæˆAIãƒ»LLMã«é–¢ã™ã‚‹è«–æ–‡ãƒ¬ãƒ“ãƒ¥ãƒ¼ä¼šã®å†…å®¹ã‚’ãƒ”ãƒƒã‚¯ã‚¢ãƒƒãƒ—ã®ã†ãˆé…ä¿¡ã—ã¦ã„ã¾ã™ã€‚ ä»Šå›žã¯ã€ãƒ„ãƒ¼ãƒ«æ‹¡å¼µã‚’ãƒ†ãƒ¼ãƒžã«ã€4ã¤ã®è«–æ–‡ã‚’ã”ç´¹ä»‹ã—ã¾ã™ã€‚ ç›®æ¬¡ ä»Šå›žã®ãƒ†ãƒ¼ãƒž Lost in the Middle: How Language Models Use Long Contexts é¸å®šç†ç”± è«–æ–‡æ¦‚è¦ ã©ã†ã‚„ã£ã¦å…¥åŠ›ã‚³ãƒ³ãƒ†ã‚ã‚¹ãƒˆã‚’ã©ã®ã‚ˆã†ã«ä½¿ç”¨ã™ã‚‹ã‹èª¿ã¹ãŸã®ï¼Ÿ çµå±€ã‚³ãƒ³ãƒ†ã‚ã‚¹ãƒˆãŒé•·ã„ã»ã©ã„ã„ã®ã‹ï¼Ÿ ãƒ¬ãƒ“ãƒ¥ãƒ¼ä¼šFB Q: Attention
deejayroka 2023/09/25
ã‚ã¨ã§èªã‚€

LLM

è«–æ–‡

ãƒ„ãƒ¼ãƒ«

ãƒ¢ãƒ‡ãƒ«

ai

data

ChatGPT

ãƒ‘ãƒ•ã‚©ãƒ¼ãƒžãƒ³ã‚¹
ãƒªãƒ³ã‚¯
Best Practices for LLM Evaluation of RAG Applications
Unified governance for all data, analytics and AI assets
deejayroka 2023/09/22
ã‚ã¨ã§èªã‚€
ãƒªãƒ³ã‚¯
ä¸€ç•ªæ‰‹æ—©ãè‡ªç¤¾ã‚µãƒ¼ãƒ“ã‚¹ã®è³¢ã„ã‚µãƒãƒ¼ãƒˆãƒœãƒƒãƒˆã‚’ä½œã‚‹æ–¹æ³•(openai.Embedding+pinecone+SlashGPT)
ä¸€ç•ªæ‰‹æ—©ãè‡ªç¤¾ã‚µãƒ¼ãƒ“ã‚¹ã®è³¢ã„ã‚µãƒãƒ¼ãƒˆãƒœãƒƒãƒˆã‚’ä½œã‚‹æ–¹æ³•(openai.Embedding+pinecone+SlashGPT) è‡ªç¤¾ã‚µãƒ¼ãƒ“ã‚¹ã®æœ€æ–°æƒ…å ±ã«ã‚‚ã¨ã¥ãã‚µãƒãƒ¼ãƒˆãƒœãƒƒãƒˆã‚’ä½œã‚ŠãŸã„ã¨ã„ã†ã®ã¯ã‚ˆãã‚ã‚‹è©±ã ã¨ãŠã‚‚ã„ã¾ã™ã€‚ã“ã‚Œã‚’ã€€openai.embedding+pinecone+SlashGPTã§ã‚µã‚¯ãƒƒã¨å®Ÿç¾ã™ã‚‹æ–¹æ³•ã‚’ç´¹ä»‹ã—ã¾ã™ã€‚ æº–å‚™ SlashGPT repo ã®clone https://github.com/snakajima/SlashGPT OpenAI API KEY open ai ã§ã‚¢ã‚«ã‚¦ãƒ³ãƒˆã‚’ä½œæˆå¾Œä¸‹è¨˜ã§API-Keyã‚’ç™ºè¡Œã—ã¾ã™ https://platform.openai.com/account/api-keys ç„¡æ–™æž ã¯5$ã¾ã§ã§ã™ã€€https://openai.com/pricing pinecone API KEY vector DB SaaS ã‚’ã¤ã‹ã†ã¨ã¨ã¦
deejayroka 2023/09/20
â€œSlashGPTã¯ä¸å³¶è¡ãŒé–‹ç™ºã—ãŸChatGPTãªã©ã®LLMã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆã‚’æ‰‹è»½ã«é–‹ç™ºã™ã‚‹ãŸã‚ã®ãƒ„ãƒ¼ãƒ«ã§ã™ã€‚SlashGPTã‚’ä½¿ãˆã°ã€jsonãƒ•ã‚¡ã‚¤ãƒ«ã‚’è¨˜è¿°ã™ã‚‹ã ã‘ã§ChatGPTã‚’ä½¿ã£ãŸLLMã‚¨ãƒ¼ã‚¸ã‚§ãƒ³ãƒˆã‚„ãƒãƒ£ãƒƒãƒˆã‚¢ãƒ—ãƒªã‚’æ‰‹è»½ã«ã€ç°¡å˜ã«ã¤ãã‚‹ã“ã¨ãŒã§

å¦ç¿’

chrome

data

ã‚µãƒ¼ãƒ“ã‚¹

ChatGPT
ãƒªãƒ³ã‚¯
ELYZAãŒå…¬é–‹ã—ãŸæ—¥æœ¬èªžLLMã€ŒELYZA-japanese-Llama-2-7bã€ã«ã¤ã„ã¦ã®è§£èª¬ : (1) äº‹å‰å¦ç¿’ç·¨
ELYZAãŒå…¬é–‹ã—ãŸæ—¥æœ¬èªžLLMã€ŒELYZA-japanese-Llama-2-7bã€ã«ã¤ã„ã¦ã®è§£èª¬ : (1) äº‹å‰å¦ç¿’ç·¨ ã¯ã˜ã‚ã« ã“ã‚“ã«ã¡ã¯ã€‚ELYZAã®ç ”ç©¶é–‹ç™ºãƒãƒ¼ãƒ ã®ä½ã€…æœ¨ (@hikomimo)ã€ä¸æ‘ (@tyo_yo_)ã€å €æ±Ÿ (@eemon18)ã€å¹³å· (@h__must__) ã§ã™ã€‚ å…ˆæ—¥å¼Šç¤¾æ ªå¼ä¼šç¤¾ELYZAã§ã¯ä»¥ä¸‹ã®ã‚ˆã†ãªãƒªãƒªãƒ¼ã‚¹ã‚’ã•ã›ã¦ã„ãŸã ãã¾ã—ãŸã€‚ Metaã®ã€ŒLlama 2ã€ã‚’ãƒ™ãƒ¼ã‚¹ã¨ã—ãŸå•†ç”¨åˆ©ç”¨å¯èƒ½ãªæ—¥æœ¬èªžLLMã€ŒELYZA-japanese-Llama-2-7bã€ã‚’å…¬é–‹ã—ã¾ã—ãŸ ä¸Šè¨˜ã®ãƒªãƒªãƒ¼ã‚¹ã«ã¯ã€Metaã®ã€ŒLlama 2ã€ã‚’ãƒ™ãƒ¼ã‚¹ã¨ã—ãŸä»¥ä¸‹ã®ãƒ¢ãƒ‡ãƒ«ãŒå«ã¾ã‚Œã¾ã™ã€‚ æ—¥æœ¬èªžè¿½åŠ äº‹å‰å¦ç¿’æ¸ˆã¿ãƒ¢ãƒ‡ãƒ« ELYZA-japanese-Llama-2-7b ELYZA-japanese-Llama-2-7b-fast ä¸Šè¨˜ã®äº‹å‰å¦ç¿’æ¸ˆã¿ãƒ¢ãƒ‡ãƒ«ã«äº‹å¾Œå¦ç¿’ (instr
deejayroka 2023/09/12
å¦ç¿’

ãƒ¢ãƒ‡ãƒ«

data

ai
ãƒªãƒ³ã‚¯
LLMã‚’åŠ¹çŽ‡çš„ã«å†å¦ç¿’ã™ã‚‹æ‰‹æ³•(PEFT)ã‚’è§£èª¬ - Platinum Data Blog by BrainPad
æœ¬è¨˜äº‹ã¯ã€å½“ç¤¾ã‚ªã‚¦ãƒ³ãƒ‰ãƒ¡ãƒ‡ã‚£ã‚¢ã€ŒDoorsã€ã«ç§»è»¢ã—ã¾ã—ãŸã€‚ ç´„5ç§’å¾Œã«è‡ªå‹•çš„ã«ãƒªãƒ€ã‚¤ãƒ¬ã‚¯ãƒˆã—ã¾ã™ã€‚ ã“ã®ãŸã³ãƒ–ãƒ¬ã‚¤ãƒ³ãƒ‘ãƒƒãƒ‰ã¯ã€LLM/Generative AIã«é–¢ã™ã‚‹ç ”ç©¶ãƒ—ãƒã‚¸ã‚§ã‚¯ãƒˆã‚’ç«‹ã¡ä¸Šã’ã€ã“ã®ã€ŒPlatinum Data Blogã€ã‚’é€šã˜ã¦LLM/Generative AIã«é–¢ã™ã‚‹ã•ã¾ã–ã¾ãªæƒ…å ±ã‚’ç™ºä¿¡ã‚’ã—ã¦ã„ã¾ã™ã€‚ä»Šå›žã¯ã€åŠ¹çŽ‡çš„ã«LLMã®ã‚ˆã†ãªäº‹å‰å¦ç¿’æ¸ˆã¿ãƒ¢ãƒ‡ãƒ«ã‚’å†å¦ç¿’ã™ã‚‹æ‰‹æ³•(PEFT)ã«ã¤ã„ã¦ã”ç´¹ä»‹ã—ã¾ã™ã€‚ å•é¡Œæ„è˜ ãƒ“ã‚¸ãƒã‚¹ã§åˆ©ç”¨ã™ã‚‹éš›ã«ä¹—ã‚Šè¶Šãˆã‚‹ã¹ãå£ PEFTã¨ã¯ä½•ã‹ï¼Ÿ PEFTã®ã‚³ãƒ³ã‚»ãƒ—ãƒˆåˆ†é¡ž ãƒˆãƒ¼ã‚¯ãƒ³è¿½åŠ åž‹ Prefix Tuning P Tuning Prompt Tuning Adapteråž‹ Adapter LoRAåž‹ LoRA Ada LoRA ã¾ã¨ã‚ å‚è€ƒæ–‡çŒ® ã“ã‚“ã«ã¡ã¯ã€ã‚¢ãƒŠãƒªãƒ†ã‚£ã‚¯ã‚¹ã‚µãƒ¼ãƒ“ã‚¹éƒ¨ã®è¾»ã§ã™ã€‚ ä»Šå›žã¯ã€LLMã‚’åŠ¹çŽ‡çš„ã«å†å¦ç¿’ã™ã‚‹æ‰‹æ³•ã¨ã—ã¦ä»Šå¾Œ
deejayroka 2023/09/04
å¦ç¿’

ãƒ¢ãƒ‡ãƒ«

data

è§£èª¬

å‹‰å¼·
ãƒªãƒ³ã‚¯
ç‹¬è‡ªãƒŠãƒ¬ãƒƒã‚¸ã‚’ãƒŽãƒ¼ã‚³ãƒ¼ãƒ‰ã§ChatGPTã«é€£æºï¼Azure OpenAIã€ŒAdd your dataã€
ç‹¬è‡ªãƒŠãƒ¬ãƒƒã‚¸ã‚’ãƒŽãƒ¼ã‚³ãƒ¼ãƒ‰ã§ChatGPTã«é€£æºï¼Azure OpenAIã€ŒAdd your dataã€ ã€Œã€Œã‚„ã£ã¦ã¿ãŸã€è¨˜äº‹ã¯ã‚‚ã†ååˆ†ã€ã¨ã„ã†æ–¹ã¯ã‚‚ã†å°‘ã—è©³ã—ãã¿ã¦ã¿ã‚‹ã‚»ã‚¯ã‚·ãƒ§ãƒ³ã‹ã‚‰ã”è¦§ãã ã•ã„ï¼ å…ˆæœˆã®Microsoft Buildã§ç™ºè¡¨ã•ã‚Œã¦æ³¨ç›®ã‚’é›†ã‚ã¦ã„ãŸã€Azure OpenAIã§ã‚³ãƒ¼ãƒ‡ã‚£ãƒ³ã‚°ãªã—ã§ç¤¾å†…ã®ãƒŠãƒ¬ãƒƒã‚¸ã‚’çµ„ã¿è¾¼ã‚“ã å›žç”ã‚’ã™ã‚‹ChatGPTãŒä½œæˆã§ãã‚‹æ©Ÿèƒ½ã€ŒAdd your dataã€ãŒãƒ‘ãƒ–ãƒªãƒƒã‚¯ãƒ—ãƒ¬ãƒ“ãƒ¥ãƒ¼ã«ãªã‚Šã¾ã—ãŸðŸŽ‰ðŸŽ‰ðŸŽ‰ðŸŽ‰ ä»Šã¾ã§ã¯ã“ã®ã‚ˆã†ãªã‚¢ãƒ¼ã‚ãƒ†ã‚¯ãƒãƒ£ã‚’ç‹¬è‡ªã§æ§‹ç¯‰ã™ã‚‹å¿…è¦ãŒã‚ã£ãŸãŸã‚ã€ãªã‹ãªã‹ãƒãƒ¼ãƒ‰ãƒ«ãŒé«˜ã„ã‚‚ã®ã§ã—ãŸã€‚ ã“ã¡ã‚‰ã®ã‚¢ãƒ¼ã‚ãƒ†ã‚¯ãƒãƒ£ã¯ã‚µãƒ³ãƒ—ãƒ«ã¨ã—ã¦å…¬é–‹ã•ã‚Œã¦ã„ã‚‹ã®ã§ã”èˆˆå‘³ãŒã‚ã‚Œã°è¦‹ã¦ã¿ã¦ãã ã•ã„ã€‚ ã‚„ã£ã¦ã¿ã‚‹ ã§ã¯ã‚»ãƒƒãƒˆã‚¢ãƒƒãƒ—ã—ã¦ã¿ã¾ã—ã‚‡ã†ã€‚ ä»Šå›žã¯ã™ã§ã«Cognitive Searchã®ãƒªã‚½ãƒ¼ã‚¹ãŒã‚ã‚‹å‰æã§é€²ã‚ã¾ã™ã€‚ Azure OpenAIã®ãƒ—
deejayroka 2023/07/22
â€œWeb Appsã®ã€Œãƒ‡ãƒ—ãƒã‚¤ã‚»ãƒ³ã‚¿ãƒ¼ã€ã‚’ã¿ã‚‹ã¨è‡ªå‹•ãƒ‡ãƒ—ãƒã‚¤ã•ã‚ŒãŸã‚¢ãƒ—ãƒªã®ã‚½ãƒ¼ã‚¹ã‚³ãƒ¼ãƒ‰ã¯GitHubã‹ã‚‰å¼•ã£å¼µã£ã¦ãã¦ã„ã‚‹ã“ã¨ãŒã‚ã‹ã‚Šã¾ã™â€

Azure

ai

data
ãƒªãƒ³ã‚¯
1 2 æ¬¡ã®ãƒšãƒ¼ã‚¸