PulseAugur
EN
LIVE 05:22:17

New Software Enables Transformer Training on Inference-Only NPUs

By PulseAugur Editorial · [1 sources] ·

VIDRAFT has developed VKORE, a software layer designed to enable transformer model training on inference-only NPUs, challenging the industry norm that reserves training for GPUs. This innovation reportedly boosts single-user inference speed by 3.8x and server throughput by 4.1x on existing NPU hardware. Furthermore, VKORE enables LoRA training on these NPUs with 0.9999 parity to standard PyTorch, achieving speeds approximately 170x faster than a CPU on the same server, and about 54% of a Nvidia B200's throughput on comparable hardware. AI

IMPACT Enables more efficient use of existing NPU hardware for training, potentially reducing the need for dedicated GPU infrastructure.

RANK_REASON The item details a novel software approach to enable transformer training on inference-only NPUs, including performance metrics and comparisons. [lever_c_demoted from research: ic=1 ai=1.0]

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

New Software Enables Transformer Training on Inference-Only NPUs

How we ranked this

Signal score
31 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The item details a novel software approach to enable transformer training on inference-only NPUs, including performance metrics and comparisons. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · AI OpenFree ·

    Training Transformers on 'Inference-Only' NPUs: How VKORE Reached 0.9999 Parity

    <p><strong>TL;DR:</strong> NPUs are sold as inference-only accelerators. VIDRAFT built VKORE, a software acceleration layer that leaves the chip untouched and (1) raises single-user inference 3.8x and server throughput 4.1x on the same NPU server, and (2) runs real transformer Lo…

RELATED ENTITIES

RELATED TOPICS