No. 40 The latest achievements of the AI big model open source project
❶ Project Name: Qwen-VL
★ Open source implementation code/model weighting
★Qwen-VL can take images, text, detection boxes as input, and text and detection boxes as output. ★Features of the Qwen-VL series models include: powerful performance, multilingual dialogue model, multi-graph interleaved dialogue, the first general-purpose model supporting Chinese open domain positioning, fine-grained recognition and understanding
# Alibaba Cloud's large-scale visual language model
❷ Project name: Lemur
★ Open source implementation code/model weighting
★ Released Lemur-70B-v1 and Lemur-70B-chat-v1, state-of-the-art open pre-trained and supervised fine-tuning large language models that balance text and code intelligence.
★The Lemur-70B-v1 model is initialized based on LLaMa-2 70B and further trained on 100B text and code data; The Lemur-70B-chat-v1 model is initialized and continues training based on supervised fine-tuning data
# Open pre-trained large language models
❸ Project name: Giraffe
★ Open source implementation code/model weighting
★ The long context model trained on LLaMa-13B and LLaMa2-13B can handle long sequences and complex language tasks, excelling on various language tasks and datasets, especially in handling long sequences and extending context length
★Giraffe innovatively proposes truncation methods, and provides a variety of context length extrapolation methods, which can be selected according to different tasks and datasets, and can be easily used through the HuggingFace API
# Open source commercial large model with a context length of 32K
❹Project Name: CMLM-ZhongJing
★ Open source implementation code
★ Drawing on the process of human memory knowledge, using professional tables, with the help of the language representation ability of large language models, strict setting of specific prompt templates, so that the model generates 15 scenarios based on the table data of TCM gynecological prescriptions, so as to promote the model's reasoning ability on TCM prescription data and diagnostic thinking logic
# The first TCM language model "Zhongjing"
❺ Project name: AutoAgents
★ Open source implementation code
★ An experimental open-source application of LLM-based automatic agent generation, driven by LLM, autonomously generates multiple agents to achieve the set goals
# Multi-agent automatic generation framework
