CogVLM
vectorflow
CogVLM | vectorflow | |
---|---|---|
16 | 9 | |
5,193 | 640 | |
10.2% | - | |
9.0 | 8.2 | |
30 days ago | 9 days ago | |
Python | Python | |
Apache License 2.0 | Apache License 2.0 |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
CogVLM
-
Mixtral: Mixture of Experts
CogVLM is very good in my (brief) testing: https://github.com/THUDM/CogVLM
The model weights seem to be under a non-commercial license, not true open source, but it is "open access" as you requested.
-
IT Employment Grew by Just 700 Jobs in 2023, Down From 267,000 in 2022
increasing growth most places in world
https://twitter.com/elonmusk/status/1743028102446408026
heres a total feature map of what was released in 2023:
https://twitter.com/enriquebrgn/status/1740950767325024387
I think thats definitely a signal that the B and C teams werent needed, considering they cut 90% of staff LOL.
As for the bots, AI is making it easier than ever to bypass those systems. CogVLM is just sitting there menacingly on github https://github.com/THUDM/CogVLM
- Show HN: I built an open source AI video search engine to learn more about AI
-
CogAgent-18B – visual-based GUI Agent capabilities
Jump to heading for benchmarks and examples: https://github.com/THUDM/CogVLM/tree/main?tab=readme-ov-file...
-
What do you think. When should we expect the next SDXL version?
Honestly at this point there is no need for human for captioning except maybe for NSFW content. Img2text is just good enough for nearly all images. GPTVision or open source equivalent (like CogVLM https://github.com/THUDM/CogVLM ) are just good enough.
-
shinning the spotlight on CogVLM
A core Llama.cpp contributor, named cmp-nct, discovered stumbled upon what might be the next leap forward for vision/language models. CogVLM (which uses a Vicuna 7B language model combined with a 9B vision tower) excels particularly in OCR (Optical Character Recognition), detail detection, and minimal hallucinations. It effectively understands both handwritten and typed text, context, fine details, and background graphics. It even provides pixel coordinates for small visual targets. CovVLM surpasses other models like llava-1.5 and Qwen-VL in performance.
-
Image-to-Caption Generator
https://github.com/THUDM/CogVLM (really impressive)
-
Gemini: Google's most capable AI model yet
I'm researching using LLMs for alt-text suggestion for forum users, can you share your finding so far?
Outside of GPT-4V I had good first results with https://github.com/THUDM/CogVLM
-
Open-source LLMs with Image Interpretation
I've got some decent results with CogVLM. Resolution kinda sucks at 490x490, though.
- FLaNK Stack Weekly for 27 November 2023
vectorflow
- FLaNK Weekly 08 Jan 2024
-
Open Source Advent Fun Wraps Up!
19. VectorFlow | Github | tutorial
-
Experimenting with LLM-Based Chunk Enhancement for Better RAG Results
Hey HN! While working on VectorFlow, an open source platform for building RAG data ingestion pipelines(repo: https://github.com/dgarnitz/vectorflow), I interviewed many people who told me they had no idea how to chunk their data. When debugging their RAG system, they found that the TOP-K results often did not include the relevant chunks. To solve this, we created a tool that can enhance the quality of a chunk by extracting relevant contextual information from the whole document based on a use case specified by the user, then selectively adding relevant portions of the extracted information to each chunk. This is only a proof-of-concept but we found that it still gives us better results on our internal RAG system.
*The Problem:*
Our users tend to have many large documents, so we typically recommend either paragraph-based chunking or token-length chunking of 512 because sentence chunking causes the information to be too spread out and key pieces get missed in retrieval. But even in these larger chunks, the embedding similarity search can miss the correct chunks because they don’t contain appropriate text. For example, the HyDE research paper does not contain the phrase “top-k similarity search” but if you are performing a RAG search asking about “the latest techniques in top-k similarity search” over a collection of academic papers related to RAG, it likely won’t show up.
*Our Solution:*
To solve this problem, we used GPT-4 to extract keywords, entities, labels, and themes from the whole document. For each chunk, we then add the five most relevant items to the end of it. When you pass in a long document, this will extract too much information for the model to effectively decide what should be part of each chunk. We found that passing in a use case for the search system, generating five potential questions based on this use case, and using those to guide the information extraction yielded more relevant results. We also add a document summary chunk to the end of every list of chunks to help with high level questions.
Using our Chunk Enhancer, we can have GPT-4 add a phrase like “top-k similarity search” to the end of the relevant chunks from the HyDE research paper so that they get picked up during a search.
*The Challenges We Faced:*
Building a Chunk Enhancer is a harder problem than we originally anticipated. Just to build a proof of concept, we had to overcome several issues.
Figuring out the right prompting techniques for this specific task was by far the hardest part. The prompt should avoid asking for multiple distinct lines of reasons. If the prompt is too complicated, even more advanced techniques like Chain of Thought and Tree of Thought do not help. We found breaking things up into multiple model calls and giving very explicit instructions (i.e. choose the top 5 best matches) was most effective. The feedback loop from prompting is different than in conventional programming. You are relying a lot more on gut feel than directly actionable feedback.
Another major issue was the inconsistency of the results - we don’t get the desired outcome often enough to use this in production yet. We know prompting techniques like self-consistency can help resolve this but its expensive.
To limit costs, we tried originally to use an open source LLM, but they are slow without GPUs and the smaller ones don’t have large enough context windows. GPT 3.5 Turbo did not work well. Some other issues we ran into were high latency, the 32K context window for larger documents, and the degradation in performance as you reach the context window limit.
*How You Can Help:*
We would love to hear feedback from the community to see if a chunk enhancer is helpful to them and how to solve some of the technical problems we encountered.
To try out the chunk enhancer, check out this colab: https://colab.research.google.com/drive/1ZagHQ23ENSt0tkD1XuC...
- FLaNK Stack Weekly for 27 November 2023
- FLaNK Stack Weekly 16 October 2023
-
Multi-Modal Vector Embeddings at Scale
Check out our Open Source repo - https://github.com/dgarnitz/vectorflow
-
Good RAG implementation
Hey, you should try out VectorFlow - https://github.com/dgarnitz/vectorflow - its the only open source high volume vector embedding pipeline out there. You can embed a few thousand files in minutes if you scale up the service. We also have a discord and can help you get set up. Our product is fully compatible with Llama Index, which we recommend people use for search
-
Challenges with Image Embeddings at Scale
I have built an open source vector embedding pipeline, VectorFlow (https://github.com/dgarnitz/vectorflow) that supports image embedding for both ingestion into vector database and similarity searches.
-
Improving the performance of RAG over 10m+ documents
We are building VectorFlow an open-source vector embedding pipeline and want to know what other features we should build next after adding open-source Sentence Transformer embedding models. Check out our Github repo: https://github.com/dgarnitz/vectorflow to install VectorFlow locally or try it out in the playground (https://app.getvectorflow.com/).
What are some alternatives?
LLaVA - [NeurIPS'23 Oral] Visual Instruction Tuning (LLaVA) built towards GPT-4V level capabilities and beyond.
karapace - Karapace - Your Apache Kafka® essentials in one tool
ollama - Get up and running with Llama 3, Mistral, Gemma, and other large language models.
retake - PostgreSQL for Search [Moved to: https://github.com/paradedb/paradedb]
ComfyUI - The most powerful and modular stable diffusion GUI, api and backend with a graph/nodes interface.
llmware - Providing enterprise-grade LLM-based development framework, tools, and fine-tuned models.
Qwen-VL - The official repo of Qwen-VL (通义千问-VL) chat & pretrained large vision language model proposed by Alibaba Cloud.
kafka-manager - CMAK is a tool for managing Apache Kafka clusters
vimGPT - Browse the web with GPT-4V and Vimium
spaCy - 💫 Industrial-strength Natural Language Processing (NLP) in Python
uform - Pocket-Sized Multimodal AI for content understanding and generation across multilingual texts, images, and 🔜 video, up to 5x faster than OpenAI CLIP and LLaVA 🖼️ & 🖋️
Wails - Create beautiful applications using Go