Ivan Fioravanti ᯅ (@ivanfioravanti)

공식 mlx-lm PR 병합을 기다리는 중이며, 로컬 AI에서 Qwen 27B는 느리지만 좋고 MiniMax M2.7은 M3 Ultra에서 강력하다고 언급했다. 배치 추론으로 백그라운드 처리할 수 있는 작업이 현재 로컬 AI에 특히 적합하다는 점을 강조했다.

https://x.com/ivanfioravanti/status/2050932921767235822

#mlxlm #qwen #localai #batchinference #minimax

Ivan Fioravanti ᯅ (@ivanfioravanti) on X

@atomtanstudio No, waiting for the PR to be fully merged first, with official mlx-lm I use Qwen 27B slow but great locally, MiniMax M2.7 strong on M3 Ultra. Anything that can go in background using batch inference is perfect nowadays for local AI!

X (formerly Twitter)
A Developer's Guide to Fine-Tuning GPT-4o for Image Classification on Azure AI Foundry | Azure AI Foundry Blog

Learn how to fine-tune OpenAI GPT-4o on Azure to classify dog breeds, compare it with its base model, and optimize for accuracy, latency, and cost.

Azure AI Foundry Blog
Amazon Bedrock の Batch Inference が GA になったから試したよ! - Qiita

二番煎じですが…実行時間も含めた多少は独自性のある記事にしてます。https://qiita.com/maekawataiki/items/9f5dc0d83446e7bf55a9公式アナウン…

Qiita