SuperAPI AI Directory
Partner with us中文AI newsFavorites

AI developer platforms

3,631 results · Page 70 of 101

ForgeTrain

AI developer platforms

ForgeTrain is a large language model pretraining framework released by OpenBMB, whose entire codebase was written end-to-end by an autonomously running AI Agent loop, with zero human editing. The project achieves 44.13% MFU (Model FLOPs Utilization) on NVIDIA H100, about a 10% improvement over the Megatron-LM baseline, and has been production-validated with a complete pretraining run of MiniCPM4-0.5B. The current v0.1.0 version supports BF16 training for MiniCPM4-0.5B (data parallelism only) and MiniCPM4-8B (tensor parallelism=2), and the accompanying Harness scaffold (used to let Agents automatically generate training frameworks) will be released soon.

Read more

Forta

AI developer platforms

Forta is an institutional-grade risk management and security platform for the onchain era. Its core product, Forta Firewall, is powered by advanced AI detection models and can integrate with various protocols and rollups to identify and block malicious activity before transaction execution. The company claims it can prevent over 99% of hacks. The platform provides continuous monitoring of asset dependencies, mapping every direct or indirect dependency and quantifying it in USD, monitoring block by block, and sending real-time alerts via email, webhook, or API. Forta also supports transaction screening at the sequencer layer, enabling it to block malicious and non-compliant transactions before execution as well as take response measures after execution. The network is powered by the $FORT token; token holders can stake and earn rewards while participating in maintaining network security.

Read more

Foundations-of-LLMs

AI developer platforms

Foundations-of-LLMs is a foundational textbook on large language models written and open-sourced by a team from Zhejiang University, aiming to systematically explain the basic knowledge and cutting-edge technologies related to large models. The first edition of this book covers six major chapters: traditional language models, the architectural evolution of large language models, Prompt engineering, parameter-efficient fine-tuning, model editing, and retrieval-augmented generation. Each chapter uses an animal as a background for examples and is accompanied by a Paper List to track the latest progress. The project is maintained through monthly updates and provides the complete PDF as well as chapter-by-chapter PDFs and related papers. The author team has also open-sourced the multi-agent development framework Agent-Kernel. This book is suitable for readers interested in large language models as a learning reference.

Read more