chai_py
H2O
chai_py | H2O | |
---|---|---|
3 | 10 | |
60 | 6,737 | |
- | 0.8% | |
0.0 | 9.7 | |
about 1 year ago | 2 days ago | |
Python | Jupyter Notebook | |
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.
chai_py
-
WHAAATTT! Since when the bots can do this? How to use it?
The package is here: https://github.com/chai-research/chai_py Quoted directly from the documentation: "The bot response accepts markdown and so you can include an image like ![image_name](http://image-url/file.jpg)"
-
Developer Docs
Worth noting the code is not open source. As you can see here, the inside of the functions are empty : https://github.com/chai-nexus/chai_py/blob/main/chai_py/chai_bot.py But it may be possible to get its real content with the inspect module
-
Is there an updated Python API in the works by chance?
The latest release is 0.4.0 from about a year ago. This looks abandoned ( https://github.com/chai-nexus/chai_py and https://pypi.org/project/chaipy/ ) and doens't work for even simply operations like retrieving the bot list. (yes with authentication of course; I have 1 bot deployed and that call bombs with a 500 status claiming "Payload is too large")
H2O
-
Really struggling with open source models
I would use H20 if I were you. You can try out LLMs with a nice GUI. Unless you have some familiarity with the tools needed to run these projects, it can be frustrating. https://h2o.ai/
- Democratizing Large Language Models
-
Interview AI Coach - by email
Here is the transcribed portion of what you sent: Within this project, or another example, for some examples of maybe encountering resistance or someone who's just like a specific person who seemed really opposed to your ideas that you had to influence or win over, and how you approach that sort of personality-based problem. Yeah, great question. So, at Lineate, I mentioned earlier that I helped to kind of upscale the entire workforce. We're talking 200 engineers, marketing folks, sales folks, account managers. And I had just, so in an effort to kind of upscale this and identify opportunities for machine learning, I followed Andrew Ng's framework for approaching ML in the enterprise. Basically, it's like one-pagers, where I define the problem statement. Do we have access to the data? Do we have data privacy or regulation concerns? What are some risk assumptions, success criteria, all that stuff. So, I put together like 20 plus one-pagers across all the different opportunities, and I generated a successful proof of concept with the team after it was a failure, of course, at first, but we turned it into a success. And part of this Andrew Ng framework in AI in the enterprise, it's like you want to generate a center of AI excellence, where it's like you share best practices with the rest of the organization. So, nobody told me that I had to do this, but this is kind of like something that I aspire towards. And in the process of trying to be inclusive with the 200 engineers, there was one engineer who was unwilling to participate. There was a phase two of his project that had an AI component that used the same tool that we used in Google Cloud. And I opened a Slack channel with our team and himself to try to get him to share what he's working with so that my team can also share what learnings we had with that tool. He just wasn't willing to participate. I just couldn't understand. It's like, how can you not? I mean, this isn't your benefit. This is a team. You got to be a team player. So, my first reaction was like, seek to understand, what's the context here? What's the background? I asked around. I talked to engineers who worked with him. I talked to higher ups without kind of like mentioning that this person was problematic, but just to understand what the nature is. And it turns out he doesn't report to the director of Solution Architecture Engineering. Instead, he reports directly to the CEO. I was like, oh, that's interesting. It turns out he came into the company through an acquisition. He was like a startup founder. So, he's used to running the show. So, when it comes to working with a team of 200 engineers, he's a superstar in terms of performance, but maybe team play-wise, not so much. So, understanding that context really helped me understand where he's coming from. And the next thing I did was I tried to anticipate, what are some of his needs? What can I do to help him reach his goals? And he wanted to, of course, do well on his project because he's a high performer. He wants to be aware of any risks early on. So, what I did was I got a hold of a sample dataset from the work that he was doing. And since I had access to some tools that he did not have, like h2o.ai, DataRobot, I took some of his data samples, put it into these tools, ran different algorithms on them, like GBM, different neural networks, to get a sense of what does a confusion matrix look like? What is this two by two matrix of true positive, false positive, and stuff like that. So, I was able to deliver some of these confusion matrices to him so that he's aware of it. And another thing is, I said, the tool that you're using is the same tool that we used. Well, guess what? It doesn't do so well in a sub-10 millisecond environment, which is one of the needs of your project. You might want to consider SageMaker endpoint where you can deploy artifact there so that this latency requirement is not a problem. So, I kind of anticipated where his needs are, being proactive to help him, offer advice where I anticipated that he needed help and extra guidance. He started kind of like more open up. And guess where I shared some of these insights? I shared it in the channel that he originally did not want to participate in. And I said, I'm going to share it in this channel. So, then he takes a look there and he starts replying to that. So, now I kind of like, kind of guided him to take one step into like this channel. So, now whatever reply he says, then my engineers can see that reply. And now it's like we have a team spirit going on now. So, that's like how I kind of got him from not wanting to participate to now participating. And on top of that, I also did like these company wide webinars where I showcase our teams. I put their profile pictures on the front slide. So, when everybody dials in, then they could see like, these are the people on my team. Here's what we're working on. And I asked him, you're really good at what you do. I would love to include you in this team in the next meeting. Are you okay with it if I put your profile picture on the front page? And he said, yes, right away. So, like helping to kind of like, because it's not like I need the credit. I just distribute some of the visibility to some of these star engineers and kind of like in exchange, you get like better collaboration. And that goal of the AI Center for Excellence for better kind of sharing best practices and learnings. So, I think by doing that, I was able to kind of like turn an icky situation into something that became a team effort. That's awesome. Love it.
-
Top 10+ OpenAI Alternatives
H2O.ai
-
Best machine learning framework(s) for production
Thanks for the input. To clarify, I am more focused on choosing the modeling framework(s) that makes the most sense to use for future production. For example, is h2o.ai a good framework for training models for later deployment (through something like elastic beanstalk, Flask API's etc.)? I came across a number of mentions of Tensorflow, however it is focused on neural nets while I also want to use classic models such as random forests, etc.
-
Time Series Analysis - Too Narrow a Dataset / Feature Set?
I've also initialised an instance of H2O.ai, so I can parse into the server each product, by store, segmented. It can then train the models, determine which model is the most performant, and then save it. Because the variability of different product SKU, at different hospitals, is substantial.
- A Tiny Grammar of Graphics
-
20+ Free Tools & Resources for Machine Learning
H2O.ai H2O is a deep learning tool built in Java. It supports most widely used machine learning algorithms and is a fast, scalable machine learning application interface used for deep learning, elastic net, logistic regression, and gradient boosting.
-
Data Science Competition
H20
-
[PAID] Looking for Phaser.js game developer
Built and founded various web3 projects for last 2 years such as OpenArt and 8RealmDojo for last 2 years as well as being high performing student in CTU in Prague and SeoulTech. Was offered internships in Amazon and H2O.ai. Created robots assistants using robots from SoftBank.
What are some alternatives?
scikit-learn - scikit-learn: machine learning in Python
MLflow - Open source platform for the machine learning lifecycle
tensorflow - An Open Source Machine Learning Framework for Everyone
Prophet - Tool for producing high quality forecasts for time series data that has multiple seasonality with linear or non-linear growth.
pycaret - An open-source, low-code machine learning library in Python
MLP Classifier - A handwritten multilayer perceptron classifer using numpy.
LightGBM - A fast, distributed, high performance gradient boosting (GBT, GBDT, GBRT, GBM or MART) framework based on decision tree algorithms, used for ranking, classification and many other machine learning tasks.
PaddlePaddle - PArallel Distributed Deep LEarning: Machine Learning Framework from Industrial Practice (『飞桨』核心框架,深度学习&机器学习高性能单机、分布式训练和跨平台部署)
adaptive - :chart_with_upwards_trend: Adaptive: parallel active learning of mathematical functions
FLAML - A fast library for AutoML and tuning. Join our Discord: https://discord.gg/Cppx2vSPVP.