Finally, with models like OPT and BLOOM at 175B params, I'm hopeful that better open-source models will be soon be trained, given the pace of innovation in the space.
OPEN SOURCE
-
Flan-T5 Public Release with Instruction Tuning Data Ablation
By
–
(2 cont.) Flan-t5 is publicly available along with the dataset (thanks to @ShayneRedford
), so you can even ablate parts of the instruction tuning data to see what unlocks performance. -
LangChain Adds Data Loaders for Cloud Storage and Popular Platforms
By
–
We added a lot of data loaders over the past few days: Cloud Storage: s3, GCS
Prosumer: Roam, Obsidian
Fun Websites: YouTube, AZLyrics, College Confidential https://
langchain.readthedocs.io/en/latest/modu
les/document_loaders/how_to_guides.html
… Which ones are you favorites? -
Magic Sandbox Project Reaches Hacker News Front Page
By
–
my "magic sandbox" makes the frontpage of hackernews never too old! https://
github.com/thomwolf/Magic
-Sand
… -
Open Assistant: Community-Driven Open Source ChatGPT Alternative
By
–
NEW VIDEO! Can you imagine an Open Source ChatGPT? Created by and for the community, available to create, research, and use freely. That's the idea behind Open Assistant! And we need your help to make it speak good SPANISH well https://
youtu.be/XYT1TxINbuM
— -
Open Source Package Preview Event Tonight
By
–
Excited to see everyone tonight! There will be drinks, snacks, good conversations, and maybe a sneak preview of some features coming to the open source package in the near future…
-
Day 1 Submissions OpenCV AI Art Contest Kickstarter
By
–
Interesting entries from Day 1 Submission 1: Raghavan
Submission 2: Fatima
Submission 3: @jnirsch Submission 4: Malviya Participate now: https://
opencv.org/winiPadAir #OpenCVKickstarter4StableDiffusion #aiartcontest #aiart #AI -
Open-source ML Tools Content and Newsletter Subscription
By
–
That's it. Thanks for reading. Follow @Sumanth_077 as I will be sharing content on more open-source ML Tools in the coming threads. If you like this content you will definitely love my newsletter as well. Subscribe here: http://
mlnewsletter.substack.com -
Training Predictive Models with Lightwood Mixers
By
–
Once you get Python Code for the Problem you can then train the model on the data and run predictions. There are many default available predictive models in Lightwood called "Mixers" but it also supports you to create your custom mixers. Check this: https://
lightwood.io/mixer.html -
Pandas Data Loading and JSON-AI Code Generation Workflow
By
–
Let's see how this works: Load data in Pandas and define a "ProblemDefinition" via dictionary specifying the name of the target column. Now Create a JSON-AI syntax using "json_ai_from_problem" function From Json AI generate a python code using "code_from_json_ai"