perspectiva aiqbenchProduse AI
analiza aiqbench
MLCommons has updated its MLPerf Training benchmark suite to v6.1, introducing two new workloads that reflect current industry priorities: a recommendation task using HSTU architecture and a massive LLM post-training workload based on Group Relative Policy Optimization. These additions expand the evaluation matrix to include emerging hardware configurations and diverse deployment scenarios, offering a more comprehensive view of real-world training performance beyond standard inference benchmarks.
3
Surse care acoperăHPCwire, AMD, CoreWeave
MLCommons Releases MLPerf Training v6.1 Results, Spotlighting Recent Advances in AI Training
HPCwire1 min de cititArticolul și perspectiva sunt în engleză.

Alte relatări ale acestei știri
Subiecte conexe

Cipuri & infrastructură
EuroHPC Requires Bidders to Resubmit AI Gigafactory Proposals
HPCwire · 6 nov. 2026

Cipuri & infrastructură
AWS Targets Metadata-Intensive HPC Workloads with FSx for Lustre Intelligent-Tiering Update
HPCwire · 6 nov. 2026

Calcul cuantic
NQCC and STFC Select 13 Projects for UK Quantum Computing Grants
HPCwire · 6 nov. 2026

Calcul cuantic
CESGA and IGFAE Expand Quantum Computing Access Through HADRON-2030
HPCwire · 2 nov. 2026