{"article":{"annotation":{"article_id":"ai2-gpu-scheduler-reallocates-cluster-time-october-2026","available_at":"2026-10-15T19:00:51.209824Z","geography_revision":"ef6ee0010dda5fe7b240fd514749b43e214c57da73222d2e40ec83b276dd0977","id":"b6503a4b6ef1bc217d370b7013ba67bc6508f5b34bb5acd690f0318342c99b5c","input_sha256":"80735c7e2be862204b0dc836335907cc0c29cbd94a40d9eadf3d329c5aca26b3","limitations":["Rule-derived subjects are candidates, not verified editorial labels or calibrated model probabilities.","Coverage is distinct from event location. Unknown country/city/coordinates remain absent; no automatic geocoding.","Article geometries keep their stated precision and methodology; reproducing them is not independent coordinate verification."],"method":"controlled-rules-v1","places":[{"admin1":null,"city":null,"coordinate_note":null,"coordinate_reference_system":"OGC:CRS84","coordinate_source":null,"country":"US","evidence":[{"field":"country","language":"en","note":"Explicit preserved editorial country/region; no city or national centroid inferred","source_urls":["https://huggingface.co/blog/allenai/impactful-scheduling"]}],"geometry":null,"id":"iso3166-1:US","kind":"country-or-area","labels":{"ar":"الولايات المتحدة الأمريكية","en":"United States of America","es":"Estados Unidos de América","fr":"États-Unis d’Amérique"},"match_status":"candidate","precision":"regional","role":"coverage","uncertainty_m":null}],"rules_sha256":"fb5ad3c27c91cb840f7a0a05f7f64a5cb26ad0374409ff9457f277c7092e22b4","schema_version":1,"subjects":[{"concept_id":"medtop:13000000","confidence":null,"confidence_method":null,"evidence":[{"field":"category+topics","language":"en","note":"Preserved editorial category/topic; broad candidate only","source_urls":["https://huggingface.co/blog/allenai/impactful-scheduling"]}],"method":"controlled-rules-v1","rule_id":"technology-broad","status":"candidate"}],"taxonomy_revision":"1f2db31d811fca07e99b16d402c10e2cc2965c9df1c7b41d9be348b40a24e7fb","taxonomy_version":"IPTC Media Topics 2026-Q2"},"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=fr","category":"technology","content":{"key_points":[],"paragraphs":["L’Allen Institute for AI (Ai2), basé aux États-Unis, a remplacé la planification par priorités dans ses clusters de recherche par un système qui attribue des budgets de temps GPU aux équipes et projets, répartit la capacité selon des règles hiérarchiques de partage équitable et peut remettre les tâches en file après une durée minimale déclarée. L’institut indique que ses clusters comptent de 88 à 1 024 GPU et qu’environ 150 chercheurs demandent à tout moment deux à trois fois la capacité disponible.","Au cours des 30 jours suivant le début du déploiement progressif fin juillet, les équipes ont reçu 98 % des heures GPU qui leur étaient dues selon la demande et les allocations ; l’occupation des clusters est restée à 98 % avant et après le changement. Ai2 signale aussi une baisse de 74 % des réparations nécessitant une intervention humaine et une réduction de cinq minutes à 24 secondes de l’attente médiane dans son plus grand cluster H100. L’institut continue de surveiller les interruptions de sessions interactives et le risque de fragmentation de capacité pour les grands travaux."],"sections":[]},"corrections":[],"country":"US","data":{"blocks":[{"description":{"ar":"القياسات المعلنة لنظام الميزانيات وتقاسم الموارد وتقسيم الوقت الجديد.","de":"Gemeldete Messwerte des neuen Systems aus Budgets, fairer Zuteilung und Zeitscheiben.","en":"Reported measurements from the new budgeting, fair-share and time-slicing system.","es":"Mediciones publicadas del nuevo sistema de presupuestos, reparto equitativo y división del tiempo.","fr":"Mesures publiées pour le nouveau système de budgets, de partage équitable et de découpage temporel.","pt":"Medições divulgadas do novo sistema de orçamentos, partilha justa e divisão de tempo.","tr":"Yeni bütçe, adil paylaşım ve zaman dilimleme düzeninin bildirilen ölçümleri."},"id":"scheduler-rollout-results","items":[{"basis":"observed","label":{"ar":"حصة ساعات GPU المستحقة التي تم تسليمها","de":"Anteil der bereitgestellten zustehenden GPU-Stunden","en":"Share of owed GPU hours delivered","es":"Proporción de horas de GPU adeudadas que se entregaron","fr":"Part des heures GPU dues effectivement fournies","pt":"Parcela das horas de GPU devidas que foi fornecida","tr":"Hak edilen GPU saatlerinin teslim oranı"},"measure":{"aggregation":"rate","frequency":"irregular","geography":"united-states","id":"ai2-gpu-hours-owed-delivered-share","periodRole":"announcement-date","unitCode":"percent"},"period":"2026-10-09","unit":"percent","value":98},{"basis":"observed","label":{"ar":"انخفاض الإصلاحات التي تتطلب تدخلاً بشرياً","de":"Rückgang der Reparaturen mit erforderlichem menschlichem Eingriff","en":"Reduction in repairs requiring human intervention","es":"Reducción de las reparaciones que requieren intervención humana","fr":"Baisse des réparations nécessitant une intervention humaine","pt":"Redução das reparações que exigem intervenção humana","tr":"İnsan müdahalesi gereken onarımlardaki azalma"},"measure":{"aggregation":"change","frequency":"irregular","geography":"united-states","id":"ai2-human-intervention-repair-reduction","periodRole":"announcement-date","unitCode":"percent"},"period":"2026-10-09","unit":"percent","value":74}],"methodology":{"ar":"أبلغت Ai2 عن حصة ساعات GPU المستحقة للفرق وفق الطلب والتخصيصات خلال فترة 30 يوماً بعد بدء النشر التدريجي على المجموعات في نهاية يوليو، وعن التغير في الإصلاحات التي تتطلب تدخلاً بشرياً.","de":"Ai2 meldete den Anteil der GPU-Stunden, die den Teams gemessen an Nachfrage und Zuteilung während eines 30-Tage-Zeitraums zustanden, nach einem Ende Juli begonnenen schrittweisen Rollout sowie die Veränderung bei Reparaturen mit erforderlichem menschlichem Eingriff.","en":"Ai2 reported the share of GPU hours teams were owed based on demand and allocations during a 30-day period after a cluster-by-cluster rollout that began at the end of July, plus the change in repairs requiring human intervention.","es":"Ai2 informó de la proporción de horas de GPU que correspondían a los equipos según su demanda y asignaciones durante un periodo de 30 días tras un despliegue gradual iniciado a finales de julio, además del cambio en las reparaciones que requieren intervención humana.","fr":"Ai2 a indiqué la part des heures GPU dues aux équipes selon leur demande et leurs allocations sur une période de 30 jours après un déploiement progressif commencé fin juillet, ainsi que l’évolution des réparations nécessitant une intervention humaine.","pt":"A Ai2 informou a parcela de horas de GPU devida às equipas conforme a procura e as alocações num período de 30 dias após uma implantação gradual iniciada no fim de julho, além da variação nas reparações que exigem intervenção humana.","tr":"Ai2, Temmuz sonunda başlayan küme bazlı devreye alımın ardından 30 günlük dönem için ekiplerin talep ve tahsislerine göre hak ettikleri GPU saatlerinin teslim oranını ve insan müdahalesi gerektiren onarım işlerinin değişimini bildirdi."},"sourceUrls":["https://huggingface.co/blog/allenai/impactful-scheduling"],"title":{"ar":"نتائج مجدول وحدات معالجة الرسوميات في Ai2","de":"Ergebnisse des GPU-Schedulers von Ai2","en":"Ai2 GPU scheduler results","es":"Resultados del planificador de GPU de Ai2","fr":"Résultats du planificateur de GPU d’Ai2","pt":"Resultados do escalonador de GPUs da Ai2","tr":"Ai2 GPU zamanlayıcısının sonuçları"},"type":"metrics","updatedAt":"2026-10-15T18:50:44+00:00"}],"version":1},"data_downloads":[{"formats":{"csv":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026/data/scheduler-rollout-results?lang=fr&format=csv","json":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026/data/scheduler-rollout-results?lang=fr&format=json"},"id":"scheduler-rollout-results"}],"data_types":["metrics"],"editorial_illustration":{"alt":"Illustration conceptuelle de la technologie, des processeurs et des informations connectées.","asset_id":"newsverum-v1-technology","avif_variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.avif","width":480},{"height":480,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-720.avif","width":720},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.avif","width":960},{"height":800,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1200.avif","width":1200},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.avif","width":1536}],"credit":"Newsverum","documentary":false,"height":1024,"kind":"symbolic-editorial-illustration","language":"fr","scope":"topic","url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.webp","width":480},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.webp","width":960},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","width":1536}],"width":1536},"event_id":"event-0cb92b5e-66b1-5c4a-a51d-d51934f40382","event_period":"2026-10-09","event_period_basis":"source-review","id":"ai2-gpu-scheduler-reallocates-cluster-time-october-2026","image":null,"importance":0.6,"language":"fr","modified_at":"2026-10-15T19:00:44+00:00","publication_basis":"publication-record","published_at":"2026-10-15T19:00:44+00:00","reading_minutes":2,"region":"world","section_illustration":{"alt":"Illustration conceptuelle de la technologie, des processeurs et des informations connectées.","asset_id":"newsverum-v1-technology","avif_variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.avif","width":480},{"height":480,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-720.avif","width":720},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.avif","width":960},{"height":800,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1200.avif","width":1200},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.avif","width":1536}],"credit":"Newsverum","documentary":false,"height":1024,"kind":"symbolic-editorial-illustration","language":"fr","scope":"section","url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","variants":[{"height":320,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-480.webp","width":480},{"height":640,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-960.webp","width":960},{"height":1024,"url":"https://newsverum.com/assets/newsverum/illustrations/v1/technology-1536.webp","width":1536}],"width":1536},"slug":"ai2-gpu-scheduler-reallocates-cluster-time-october-2026","sources":[{"checkedAt":"2026-10-15","name":"Allen Institute for AI (Ai2)","publishedAt":"2026-10-09","url":"https://huggingface.co/blog/allenai/impactful-scheduling"}],"summary":"Ai2 indique que les équipes ont reçu 98 % des heures GPU qui leur étaient dues lors d’une évaluation de 30 jours d’un nouveau système de budgets et de partage équitable.","text":"Ai2 indique que les équipes ont reçu 98 % des heures GPU qui leur étaient dues lors d’une évaluation de 30 jours d’un nouveau système de budgets et de partage équitable.\n\nL’Allen Institute for AI (Ai2), basé aux États-Unis, a remplacé la planification par priorités dans ses clusters de recherche par un système qui attribue des budgets de temps GPU aux équipes et projets, répartit la capacité selon des règles hiérarchiques de partage équitable et peut remettre les tâches en file après une durée minimale déclarée. L’institut indique que ses clusters comptent de 88 à 1 024 GPU et qu’environ 150 chercheurs demandent à tout moment deux à trois fois la capacité disponible.\n\nAu cours des 30 jours suivant le début du déploiement progressif fin juillet, les équipes ont reçu 98 % des heures GPU qui leur étaient dues selon la demande et les allocations ; l’occupation des clusters est restée à 98 % avant et après le changement. Ai2 signale aussi une baisse de 74 % des réparations nécessitant une intervention humaine et une réduction de cinq minutes à 24 secondes de l’attente médiane dans son plus grand cluster H100. L’institut continue de surveiller les interruptions de sessions interactives et le risque de fragmentation de capacité pour les grands travaux.","title":"Ai2 présente un planificateur qui modifie le partage de ses clusters GPU","topics":[],"translations":[{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=tr","language":"tr","url":"https://newsverum.com/tr/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=en","language":"en","url":"https://newsverum.com/en/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=ar","language":"ar","url":"https://newsverum.com/ar/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=fr","language":"fr","url":"https://newsverum.com/fr/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=es","language":"es","url":"https://newsverum.com/es/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=de","language":"de","url":"https://newsverum.com/de/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},{"api_url":"https://newsverum.com/api/v1/articles/ai2-gpu-scheduler-reallocates-cluster-time-october-2026?lang=pt","language":"pt","url":"https://newsverum.com/pt/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"}],"url":"https://newsverum.com/fr/news/ai2-gpu-scheduler-reallocates-cluster-time-october-2026"},"version":"1"}