A Bit About Byte Magazine: The Premier Technical Journal of Its Time
11 by zdw | 8 comments on Hacker News.
Saturday, September 30, 2023
Friday, September 29, 2023
Thursday, September 28, 2023
Wednesday, September 27, 2023
New top story on Hacker News: Show HN: Using LLMs and Embeddings to classify application errors
Show HN: Using LLMs and Embeddings to classify application errors
19 by vadman97 | 0 comments on Hacker News.
Hi Hacker News! We’re Vadim and Chris from Highlight.io [1]. We do web app monitoring and are working on using LLMs/embeddings to add new functionality to our error monitoring product. Given that there’s a lot of founders/engineers using LLMs in their products, we figured we’d share how we built the new functionality, their impact on our workflows, and how you can try it out. Our goal was to build two features: (1) tagging errors (e.g. deeming an error as “authentication error” or a “database error”); and (2) grouping similar errors together (e.g. two errors that have a different stacktrace and body, but are semantically not very different). Each of these rely heavily on comparing text across our application. After some experimentation with the OpenAI embeddings API [3], we went ahead and hosted a private model instance of thenlper/gte-large (an open-source MIT licensed model), which is a 1024-dimension model running on an Intel Ice Lake 2 vCPU machine on Hugging face [4]. Our general approach for classifying/comparing text is as follows. As each set of tokens (i.e a string) comes in, our backend makes a request to an inference endpoint and receives a 1024-dimension float vector as a response (see the code here [5]). We then store that vector using pgvector [6]. To compare any two sets for similarity, we simply look at the Euclidian distance between their respective embeddings using the ivfflat index implemented by pgvector (example code here [7]). To tag errors, we assign an error its most relevant tag from a predetermined set decided by us. For example, if we tag an error as an "authentication error" or a "database error", we can allow developers to have a starting point before inspecting an issue.(see the logic here [8]). Anecdotally, this approach seems to work very well. For example, here are two authentication errors that got tagged as “Authentication Error”: * Firebase: A network AuthError has occurred * Error retrieving user from firebase api for email verification: cannot find user from uid. We also use these error embeddings to group similar errors. To decide whether an error joins a group or starts a new one, we decide on a distance threshold (using the euclidean distance) ahead of time. An interesting thing about this approach, compared to using a text-based heuristic, is that two errors with different stack traces can still be grouped together. Here’s an example: * github.com/highlight-run/highlight/backend/worker.(*Worker).ReportStripeUsage * github.com/highlight-run/highlight/backend/private-graph/graph.(*Resolver).GetSlackChannelsFromSlack.func1 Both reported as `integration api error` as they involve the Stripe and Slack integrations respectively. The neat thing is that the LLM can use the full context of an error and match based on the most relevant details about the error. We have rolled out a first version of the error grouping logic to our cloud product [9], and there’s a demo of all the functionality at [2]. Long-term, if the HN community has other ideas of what we could build with LLM tooling in observability, we’re all ears. Let us know what you think! Links [1] https://ift.tt/UiZVr1T [2] https://ift.tt/X7Ty2nH [3] https://ift.tt/gETKHUo [4] https://ift.tt/IZQHRbK [5] https://ift.tt/9FgBQrc... [6] https://ift.tt/sdTRMVe... [7] https://ift.tt/NMLS537... [8] https://ift.tt/nWhJDF7... [9] https://ift.tt/xzqk2sW
19 by vadman97 | 0 comments on Hacker News.
Hi Hacker News! We’re Vadim and Chris from Highlight.io [1]. We do web app monitoring and are working on using LLMs/embeddings to add new functionality to our error monitoring product. Given that there’s a lot of founders/engineers using LLMs in their products, we figured we’d share how we built the new functionality, their impact on our workflows, and how you can try it out. Our goal was to build two features: (1) tagging errors (e.g. deeming an error as “authentication error” or a “database error”); and (2) grouping similar errors together (e.g. two errors that have a different stacktrace and body, but are semantically not very different). Each of these rely heavily on comparing text across our application. After some experimentation with the OpenAI embeddings API [3], we went ahead and hosted a private model instance of thenlper/gte-large (an open-source MIT licensed model), which is a 1024-dimension model running on an Intel Ice Lake 2 vCPU machine on Hugging face [4]. Our general approach for classifying/comparing text is as follows. As each set of tokens (i.e a string) comes in, our backend makes a request to an inference endpoint and receives a 1024-dimension float vector as a response (see the code here [5]). We then store that vector using pgvector [6]. To compare any two sets for similarity, we simply look at the Euclidian distance between their respective embeddings using the ivfflat index implemented by pgvector (example code here [7]). To tag errors, we assign an error its most relevant tag from a predetermined set decided by us. For example, if we tag an error as an "authentication error" or a "database error", we can allow developers to have a starting point before inspecting an issue.(see the logic here [8]). Anecdotally, this approach seems to work very well. For example, here are two authentication errors that got tagged as “Authentication Error”: * Firebase: A network AuthError has occurred * Error retrieving user from firebase api for email verification: cannot find user from uid. We also use these error embeddings to group similar errors. To decide whether an error joins a group or starts a new one, we decide on a distance threshold (using the euclidean distance) ahead of time. An interesting thing about this approach, compared to using a text-based heuristic, is that two errors with different stack traces can still be grouped together. Here’s an example: * github.com/highlight-run/highlight/backend/worker.(*Worker).ReportStripeUsage * github.com/highlight-run/highlight/backend/private-graph/graph.(*Resolver).GetSlackChannelsFromSlack.func1 Both reported as `integration api error` as they involve the Stripe and Slack integrations respectively. The neat thing is that the LLM can use the full context of an error and match based on the most relevant details about the error. We have rolled out a first version of the error grouping logic to our cloud product [9], and there’s a demo of all the functionality at [2]. Long-term, if the HN community has other ideas of what we could build with LLM tooling in observability, we’re all ears. Let us know what you think! Links [1] https://ift.tt/UiZVr1T [2] https://ift.tt/X7Ty2nH [3] https://ift.tt/gETKHUo [4] https://ift.tt/IZQHRbK [5] https://ift.tt/9FgBQrc... [6] https://ift.tt/sdTRMVe... [7] https://ift.tt/NMLS537... [8] https://ift.tt/nWhJDF7... [9] https://ift.tt/xzqk2sW
Tuesday, September 26, 2023
Monday, September 25, 2023
Sunday, September 24, 2023
Saturday, September 23, 2023
Friday, September 22, 2023
Thursday, September 21, 2023
Wednesday, September 20, 2023
Tuesday, September 19, 2023
New top story on Hacker News: Show HN: Hydra - Open-Source Columnar Postgres
Show HN: Hydra - Open-Source Columnar Postgres
38 by coatue | 3 comments on Hacker News.
hi hn, hydra ceo here hydra is an open-source extension that adds columnar tables to Postgres for efficient analytical reporting. With Hydra, you can analyze billions of rows instantly without changing code. demo video (5 min): https://youtu.be/1yzxgb0Oyrw github repo: https://ift.tt/Fni1dxh For 1.0 GA release, aggregate queries are over *60% faster* than Hydra beta due to aggregate vectorization. Spatial indexes (gin, gist, spgist, and rum indexes) and pg_hint_plan are now enabled for performance optimization. postgres is great, but aggregates can take minutes to hours to return results on large data sets. long-running analytical queries hog database resources and degrade performance. use hydra to run much faster analytics on postgres without changing code. for testing, try the hydra free tier to create a column postgres instance on the cloud. https://ift.tt/MP10D8B
38 by coatue | 3 comments on Hacker News.
hi hn, hydra ceo here hydra is an open-source extension that adds columnar tables to Postgres for efficient analytical reporting. With Hydra, you can analyze billions of rows instantly without changing code. demo video (5 min): https://youtu.be/1yzxgb0Oyrw github repo: https://ift.tt/Fni1dxh For 1.0 GA release, aggregate queries are over *60% faster* than Hydra beta due to aggregate vectorization. Spatial indexes (gin, gist, spgist, and rum indexes) and pg_hint_plan are now enabled for performance optimization. postgres is great, but aggregates can take minutes to hours to return results on large data sets. long-running analytical queries hog database resources and degrade performance. use hydra to run much faster analytics on postgres without changing code. for testing, try the hydra free tier to create a column postgres instance on the cloud. https://ift.tt/MP10D8B
Monday, September 18, 2023
New top story on Hacker News: Tell HN: Goodbye HTML Gmail
Tell HN: Goodbye HTML Gmail
37 by howmayiannoyyou | 11 comments on Hacker News.
To be (dis)continued: "We’re writing to let you know that the Gmail Basic HTML view for desktop web and mobile web will be disabled starting early January 2024. The Gmail Basic HTML views are previous versions of Gmail that were replaced by their modern successors 10+ years ago and do not include full Gmail feature functionality."
37 by howmayiannoyyou | 11 comments on Hacker News.
To be (dis)continued: "We’re writing to let you know that the Gmail Basic HTML view for desktop web and mobile web will be disabled starting early January 2024. The Gmail Basic HTML views are previous versions of Gmail that were replaced by their modern successors 10+ years ago and do not include full Gmail feature functionality."
Sunday, September 17, 2023
Saturday, September 16, 2023
Friday, September 15, 2023
Thursday, September 14, 2023
Wednesday, September 13, 2023
Tuesday, September 12, 2023
Monday, September 11, 2023
Sunday, September 10, 2023
New top story on Hacker News: Show HN: Erlmacs – a script to update your .emacs file for Erlang development
Show HN: Erlmacs – a script to update your .emacs file for Erlang development
9 by dlachausse | 0 comments on Hacker News.
erlmacs automatically configures and updates your .emacs file with support for the emacs mode that is included with Erlang/OTP. It frees you from having to locate the installation directory of Erlang/OTP and its bundled emacs mode. It is an escript that only depends upon Erlang/OTP and Emacs. Note: There is not much in the way of error checking at this moment, but it does make a backup of your .emacs files before any destructive operations.
9 by dlachausse | 0 comments on Hacker News.
erlmacs automatically configures and updates your .emacs file with support for the emacs mode that is included with Erlang/OTP. It frees you from having to locate the installation directory of Erlang/OTP and its bundled emacs mode. It is an escript that only depends upon Erlang/OTP and Emacs. Note: There is not much in the way of error checking at this moment, but it does make a backup of your .emacs files before any destructive operations.
Saturday, September 9, 2023
New top story on Hacker News: Show HN: WhatsApp-Llama: A clone of yourself from your WhatsApp conversations
Show HN: WhatsApp-Llama: A clone of yourself from your WhatsApp conversations
26 by advaith08 | 4 comments on Hacker News.
Hello HN! I've been thinking about the idea of a LLM thats a clone of me - instead of generating replies to be a helpful assistant, it generates replies that are exactly like mine. The concept's appeared in fiction numerous times (the talking paintings in Harry Potter that mimic the person painted, the clones in The Prestige), and I think with LLMs, there might actually be a possibility of us doing something like this! I've just released a fork of the facebookresearch/llama-recipes which allows you to fine-tune a Llama model on your personal WhatsApp conversations. This adaptation can train the model (using QLoRA) to respond in a way that's eerily similar to your own texting style. What I've figured out so far: Quick Learning: The model quickly adapts to personal nuances, emoji usage, and phrases that you use. I've trained just 1 epoch on a P100 GPU using QLoRA and 4 bit quantization, and its already captured my mannerisms Turing Tests: As an experiment, I asked my friends to ask me 3 questions, and responded with 2 candidate responses (one from me and one from llama). My friends then had to guess which candidate response was mine and which one was Llama's. Llama managed to fool 10% of my friends, but with more compute, I think it can do way better. Here's the GitHub repository: https://ift.tt/75og901 Would love to hear feedback, suggestions, and any cool experiences if you decide to give it a try! I'd love to see how far we can push this by training bigger models for more epochs (I ran out of compute credits)
26 by advaith08 | 4 comments on Hacker News.
Hello HN! I've been thinking about the idea of a LLM thats a clone of me - instead of generating replies to be a helpful assistant, it generates replies that are exactly like mine. The concept's appeared in fiction numerous times (the talking paintings in Harry Potter that mimic the person painted, the clones in The Prestige), and I think with LLMs, there might actually be a possibility of us doing something like this! I've just released a fork of the facebookresearch/llama-recipes which allows you to fine-tune a Llama model on your personal WhatsApp conversations. This adaptation can train the model (using QLoRA) to respond in a way that's eerily similar to your own texting style. What I've figured out so far: Quick Learning: The model quickly adapts to personal nuances, emoji usage, and phrases that you use. I've trained just 1 epoch on a P100 GPU using QLoRA and 4 bit quantization, and its already captured my mannerisms Turing Tests: As an experiment, I asked my friends to ask me 3 questions, and responded with 2 candidate responses (one from me and one from llama). My friends then had to guess which candidate response was mine and which one was Llama's. Llama managed to fool 10% of my friends, but with more compute, I think it can do way better. Here's the GitHub repository: https://ift.tt/75og901 Would love to hear feedback, suggestions, and any cool experiences if you decide to give it a try! I'd love to see how far we can push this by training bigger models for more epochs (I ran out of compute credits)
Friday, September 8, 2023
Thursday, September 7, 2023
Wednesday, September 6, 2023
Tuesday, September 5, 2023
Monday, September 4, 2023
Sunday, September 3, 2023
Saturday, September 2, 2023
New top story on Hacker News: Show HN: Modular Diffusion – A modular Python library for diffusion models
Show HN: Modular Diffusion – A modular Python library for diffusion models
4 by secularchapel | 0 comments on Hacker News.
Hello everyone! I've been working on this project for a few months as part of my thesis in Machine Learning. It's meant to be a library that provides an easy-to-use but flexible API to design and train Diffusion Models. I decided to make it because I wanted to quickly prototype a Diffusion Model but there were no good tools to do it with. I think it really can help people prototype their own Diffusion Models a lot faster and only in a few lines of code. The base idea is to have a Model class that takes different modules corresponding to the different aspects of the Diffusion Model process (noise schedule, noise type, denoising network, loss function, guidance, etc.) and allow the user to mix and match different modules to achieve different results. The library ships with a bunch of prebuilt modules and the plan is to add many more. I also made it super easy to implement your own modules, you just need to extend from one of the base classes available. Contrary to HuggingFace Diffusers, this library is focused on designing and training your own Diffusion Models rather than finetuning pretrained ones (although this is possible). I would really appreciate your feedback.
4 by secularchapel | 0 comments on Hacker News.
Hello everyone! I've been working on this project for a few months as part of my thesis in Machine Learning. It's meant to be a library that provides an easy-to-use but flexible API to design and train Diffusion Models. I decided to make it because I wanted to quickly prototype a Diffusion Model but there were no good tools to do it with. I think it really can help people prototype their own Diffusion Models a lot faster and only in a few lines of code. The base idea is to have a Model class that takes different modules corresponding to the different aspects of the Diffusion Model process (noise schedule, noise type, denoising network, loss function, guidance, etc.) and allow the user to mix and match different modules to achieve different results. The library ships with a bunch of prebuilt modules and the plan is to add many more. I also made it super easy to implement your own modules, you just need to extend from one of the base classes available. Contrary to HuggingFace Diffusers, this library is focused on designing and training your own Diffusion Models rather than finetuning pretrained ones (although this is possible). I would really appreciate your feedback.
Friday, September 1, 2023
New top story on Hacker News: Show HN: An Immersive Game of Thrones Multiverse Experience
Show HN: An Immersive Game of Thrones Multiverse Experience
1 by thronesMultiV | 0 comments on Hacker News.
Alpha Version Demo: https://ift.tt/81VHCTN Twitter: https://twitter.com/ThronesMultiV/status/1697440568874348953 We're here to present an experimental product empowered by the blend of Stable Diffusion and ChatGPT! Dive into Westeros like never before. Our experimental product offers an immersive storytelling experience where you play a pivotal role in shaping the narrative. Ever wondered if the ending of the final seasons of Game of Thrones could've been different? Now's your chance to twist the tale. Current Features : - AI-driven alternative endings starting from the end of S7. - Real-time story interventions, allowing you to change the plotline as you read. What's Next : - Continuous enhancements to refine and polish the storytelling experience. - And yes, we're contemplating open-sourcing the project – giving back to this amazing community and encouraging further innovation. We truly believe in the power of collaboration. If you have feedback, suggestions, or just want to geek out about Westeros, shoot us an email at ready2play.contact@gmail.com ! Additionally, if you're as passionate about AI and storytelling as we are, we'd love for you to collaborate with us on this exciting project. Remember, winter is coming, but with AI, the possibilities are endless. Stay excited and stay kind! Valar Morghulis!
1 by thronesMultiV | 0 comments on Hacker News.
Alpha Version Demo: https://ift.tt/81VHCTN Twitter: https://twitter.com/ThronesMultiV/status/1697440568874348953 We're here to present an experimental product empowered by the blend of Stable Diffusion and ChatGPT! Dive into Westeros like never before. Our experimental product offers an immersive storytelling experience where you play a pivotal role in shaping the narrative. Ever wondered if the ending of the final seasons of Game of Thrones could've been different? Now's your chance to twist the tale. Current Features : - AI-driven alternative endings starting from the end of S7. - Real-time story interventions, allowing you to change the plotline as you read. What's Next : - Continuous enhancements to refine and polish the storytelling experience. - And yes, we're contemplating open-sourcing the project – giving back to this amazing community and encouraging further innovation. We truly believe in the power of collaboration. If you have feedback, suggestions, or just want to geek out about Westeros, shoot us an email at ready2play.contact@gmail.com ! Additionally, if you're as passionate about AI and storytelling as we are, we'd love for you to collaborate with us on this exciting project. Remember, winter is coming, but with AI, the possibilities are endless. Stay excited and stay kind! Valar Morghulis!
Subscribe to:
Posts (Atom)