SuperAPI AI Directory
Partner中文AI newsFavorites

All AI tools

35,738 results · Page 552 of 993

baby-llama2-chinese

AI developer platforms

baby-llama2-chinese is an open-source project for beginners of Chinese large language models. Its goal is to pretrain from scratch and SFT a small-parameter Chinese LLaMa2 model that can run on a single 24G GPU, yielding a chat-llama2 with simple Chinese question-answering capabilities. The project covers the complete pipeline of pretraining, SFT instruction fine-tuning, reward model, and reinforcement learning (marked as to-do), and provides code and explanations for Chinese pretraining corpus processing, data cleaning, tokenization processing, and model comparison. The author states that a complete set of LLM learning materials is also being organized (in progress), hoping to help LLM beginners get started as quickly as possible. The project chooses the tokenizer of ChatGLM2-6B (vocabulary size 64793) and open-sources a pretraining corpus processed by this tokenizer, totaling 63.4 billion Tokens. The repository provides a unified inference entry point, infer.py, which supports pretrained and SFT weights, can automatically infer the model dimension, number of layers, vocabulary size, and number of KV heads, and automatically selects CUDA, Apple Silicon MPS, or CPU. The project has also released multiple comparison models with different parameter counts and pretraining data scales, as well as a medical question-answering model fine-tuned on medical vertical-domain SFT data.

Read more

BabyAGI

AI agents

BabyAGI is an experimental self-building autonomous agent framework built by Yohei, and is not a production-grade tool. The original BabyAGI released in March 2023 introduced task planning as a development method for autonomous agents; that version has been archived and migrated to the babyagi_archive repository (September 2024 snapshot). The core of the current latest version of BabyAGI is a function framework called functionz, used to store, manage, and execute functions from a database, providing a graph-based structure to track imports, dependent functions, and authentication keys, and supporting automatic loading and comprehensive logging. It also comes with a dashboard for managing functions, running updates, and viewing logs. Installation is via pip install babyagi, and it can be imported in Python code and used to start the dashboard, accessed at http://localhost:8080/dashboard for use. The framework includes two preloaded function packs: default_functions provides function execution, key management, triggers, and logging; ai_generator provides AI description generation, embeddings, and prompt-based function selection. It also includes two experimental self-building agents: process_user_input in the code_writing_functions pack and self_build in the self_build pack. The project uses the MIT license. The author explicitly notes that the framework is not designed for production environments and should be used with caution.

Read more

BabyCommandAGI

AI agents

BabyCommandAGI is an open-source project based on BabyAGI, aimed at exploring the potential of combining a command-line interface (CLI) with large language models (LLMs). By having the LLM converse with the CLI, it automatically executes and plans tasks, enabling scenarios such as automatic programming and environment setup. Users need to configure a Claude or OpenAI API key and run it in a Docker environment. The system is currently in an early stage and may loop execution or damage the environment; it is recommended to use it in a virtual environment and to be mindful of API costs.

Read more

BabyFilter

AI images

Historical third-party summary: The saved BabyFilter card and a One’s Vibe index connect the original babyfilter.art address with a playful image effect that turns portrait photos into a baby-like or younger appearance. Only that visual-transformation purpose is summarized. Outputs are not treated as real childhood photos,age assessments or identity evidence;facial fidelity and privacy promises are not verified. The extra reference is an index;source full text not verified. The directory batch is marked2026-10-12 and was checked on2026-10-18. The name and URL remain. No portrait was uploaded,and processing,downloads,pricing and licensing were not tested.

Read more

Backlink Exchange MarketPlace

AI marketing

Backlink Exchange MarketPlace is a page on the personal website of independent developer Praneet Brar (https://praneetbrar.com/backlink-marketplace），其个人主页自述为「独立黑客与全栈开发者，打造 SEO products, AI-driven tools, and practical web projects." Apart from this self-description and the page link, the available information does not provide specific details about this tool's features, usage, pricing model, or supported platforms, so an introduction cannot be expanded at this time.

Read more