AAAI'25 Deadline Today! SD Core Members Open Source a More Powerful Text-based Graph Model Than Midjourney, Now Available for One-click Startup
Midjourney is really not popular anymore! After Stable Diffusion 3 was open sourced in June, Robin Rombach, a former core member of Stability AI, led a new team to launch the FLUX.1 image generation model family at the beginning of this month. Officials claim that FLUX.1 surpasses the head models of Midjourney v6.0 and DALL·E 3 in multiple key indicators, and it is open source. It sounds so powerful, are you also eager to try it?
The hyper.ai official website has now launched the "FLUX.1-schnell Vincent Figure Demo" in the tutorial section.Just clone it with one click and start playing. Scroll down the article to get the link~
From August 12 to August 16, hyper.ai official website updates:
* Selection of high-quality tutorials: 3
* High-quality public datasets: 10
* Community article selection: 4 articles
* Popular encyclopedia entries: 5
* Top conferences with deadline in August: 2
Visit the official website:hyper.ai
Selected Public Tutorials
1. FLUX.1-schnell Vincent Figure Demo
FLUX.1 is a large model with 12 billion parameters that can generate images from text descriptions. It has reached the state-of-the-art level in terms of real-time tracking, visual quality, image detail, and output diversity. This tutorial uses the FLUX.1 [schnell] version model. The model and environment are deployed. You can directly use the large model for inference generation according to the tutorial instructions.
Direct use:https://go.hyper.ai/peksE
ControlNet author Lvmin Zhang has developed a new project called Paints-Undo, which can quickly disassemble the painting process of any image, helping beginners better understand the painting techniques of different styles of images. HyperAI Super Neural has now launched the "Paints-Undo Demo of the entire painting process generated from one image". This tutorial has set up the environment for everyone. You don't need to enter any commands, just clone it with one click to start!
Direct use:https://go.hyper.ai/EwBE0
3. One-click deployment of the Puke chemical large model ChemLLM-7B-chat Demo
ChemLLM-7B-Chat is the first open source large-scale language model for chemistry and molecular sciences, "Puke Chemistry (ChemLLM)", which was open sourced by the Shanghai Artificial Intelligence Laboratory (Shanghai AI Laboratory) in 2024. This tutorial is a one-click deployment demo of the model. You only need to clone and start the container and directly copy the generated API address to experience the inference of the model.
Direct use:https://go.hyper.ai/X8V9z
Selected public datasets
1. ChemData Chemical Task Dataset
This dataset was open-sourced by the Shanghai Artificial Intelligence Laboratory together with its first scientific big model, the Pu Ke Chemistry Big Model (ChemLLM). It mainly includes 9 core chemical tasks, 730K high-quality question-and-answer large language model chemical ability instruction fine-tuning datasets.
Direct use:https://go.hyper.ai/94tF1
2. ChemBench4K Chemical Ability Evaluation Benchmark Dataset
The dataset consists of 9 tasks about chemical molecules and reactions, including 4,100 multiple-choice questions. The benchmark lays the foundation for objectively measuring the chemistry level of large language models.
Direct use:https://go.hyper.ai/itsdU
3. BRIGHT Text Retrieval Benchmark Dataset
The dataset collects 1,385 real queries from different fields (StackExchange, LeetCode, and math competitions), all of which come from real artificial data. The BRIGHT dataset is specifically designed to test whether the retrieval system can identify such deep logical relationships and find relevant academic articles or reports.
Direct use:https://go.hyper.ai/s735d
4. Multimodal ArXiv Scientific Understanding Dataset
Multimodal ArXiv consists of ArXivCap and ArXivQA to enhance the scientific understanding of LVLM. ArXivCap is a graph caption dataset containing 6.4 million images and 3.9 million captions. ArXivQA is a question-answering dataset generated by GPT-4V based on scientific graphs through prompts. The related results have been accepted by ACL 2024.
Direct use:https://go.hyper.ai/n64Jh
5. SPIQA Multimodal Scientific Paper Question Answering Dataset
This is the first large-scale QA dataset specifically designed to interpret complex figures and tables in scientific research articles in various fields of computer science, containing 270K questions divided into training, validation and 3 different evaluation parts. Through extensive experiments on 12 well-known basic models, the team evaluated the ability of current multimodal systems to understand subtle aspects of research articles.
Direct use:https://go.hyper.ai/qd7I2
6. MMEvaIPro Multimodal Benchmark Evaluation Dataset
MMEvalPro improves existing evaluation methods by adding two "anchor" questions (1 perception question and 1 knowledge question), forming "question triplets" that test different aspects of the model's multimodal understanding. The final benchmark contains 2,138 question triplets, for a total of 6,414 different questions covering different topics and difficulty levels.
Direct use:https://go.hyper.ai/Hw8JA
7. PubMedVision Large-Scale Medical VQA Dataset
PubMedVision is a large-scale, high-quality medical multimodal dataset. The research team used sophisticated data processing methods to screen out medical-related images and informative image descriptions from papers in PubMed international medical journals, effectively filtering out a large number of medical-irrelevant images and context-irrelevant content.
Direct use:https://go.hyper.ai/Uy8XM
8. Multi modal Self instruct Multimodal benchmark dataset
The dataset contains a total of 11,193 abstract images with relevant questions, covering 8 major categories including dashboards, roadmaps, charts, tables, flowcharts, relationship diagrams, visual puzzles and 2D floor plans, in addition to an additional 62,476 data for fine-tuning the model.
Direct use:https://go.hyper.ai/FwGuz
9. Assetto Corsa Gym Large-scale autonomous racing simulation benchmark
The dataset collects 64 million steps of racing driving data, of which 2.3 million steps are from human drivers of different driving skills and the rest are from Soft Actor-Critic (SAC) policies.
Direct use:https://go.hyper.ai/6tfuM
10. MiraData: A Large-Scale Video Dataset with Long Duration and Structured Captions
MiraData focuses on uncut video clips of 1 to 2 minutes (average duration 72.1 seconds), each video is accompanied by a structured description from different angles, with an average description length of 318 words, ensuring a comprehensive presentation of the video content. This dataset provides valuable resources and new challenges for researchers in the fields of long video generation, video content understanding and generation.
Direct use:https://go.hyper.ai/2LmEJ
For more public datasets, please visit:
Community Articles
On August 14, various awards of ACL 2024 were announced one by one. A total of 7 results won the Best Paper. "Deciphering Oracle Bone Language with Diffusion Models" jointly published by Huazhong University of Science and Technology, University of Adelaide, Anyang Normal University, and South China University of Technology won the award. HyperAI Super Neural Network brought everyone a detailed interpretation.
View the full report:https://go.hyper.ai/t5Zon
Recently, Pratyusha Sharma of MIT and researchers from CETI used machine learning to analyze sperm whale recordings and confirmed that the sounds made by sperm whales are structured and composed of different features. They also isolated the sperm whale pronunciation alphabet, which is highly similar to the human language expression system. This article is a detailed interpretation and sharing of the research paper.
View the full report:https://go.hyper.ai/nA23S
Recently, the research team of Academician Dai Qionghai and Professor Fang Lu of Tsinghua University has grasped the symmetry of photon propagation, equating the forward and backward propagation in neural network training to the forward propagation of light, and developed a method of full forward mode learning. This article is a detailed interpretation and sharing of the research paper.
View the full report:https://go.hyper.ai/lxNhj
In order to combine the neural network algorithm and the DFT algorithm more organically, the research group of Xu Yong and Duan Wenhui from Tsinghua University proposed a neural network density functional theory (neural-network DFT) framework. This framework unifies the minimization of the loss function in the neural network and the optimization of the energy functional in the density functional theory. Compared with the traditional supervised learning method, it has higher accuracy and efficiency, and has opened up a new path for the development of deep learning DFT methods. This article is a detailed interpretation and sharing of the research paper.
View the full report:https://go.hyper.ai/oE7nH
Popular Encyclopedia Articles
1. Paired t-Test
2. Reciprocal sorting fusion RRF
3. Pareto Front
4. Large-scale Multi-task Language Understanding (MMLU)
5. Data Augmentation
Here are hundreds of AI-related terms compiled to help you understand "artificial intelligence" here:
One-stop tracking of top AI academic conferences:https://go.hyper.ai/event
The above is all the content of this week’s editor’s selection. If you have resources that you want to include on the hyper.ai official website, you are also welcome to leave a message or submit an article to tell us!
See you next week!
About HyperAI
HyperAI (hyper.ai) is the leading artificial intelligence and high-performance computing community in China.We are committed to becoming the infrastructure in the field of data science in China and providing rich and high-quality public resources for domestic developers. So far, we have:
* Provide domestic accelerated download nodes for 1300+ public data sets
* Includes 400+ classic and popular online tutorials
* Interpretation of 100+ AI4Science paper cases
* Support 500+ related terms search
* Hosting the first complete Apache TVM Chinese documentation in China
Visit the official website to start your learning journey: