kevinlu1248

7 ms·

31.
Linear-time tokenizer crate by GitHub(github.blog)
2 points by kevinlu1248 1y ago|0 comments
32.
Patience Diff: Diff algorithm using patience sorting(blog.jcoglan.com)
2 points by kevinlu1248 3y ago|0 comments
33.
Firecracker for Leetcode-style Code Grading(stanislas.blog)
1 points by kevinlu1248 3y ago|0 comments
34.
by kevinlu1248 3y ago
We also found a lot of cases where caching ends up being actually slower than doing the operation. The 100% solution would probably be to use a SQL db the way diskcache does it, but this is easier to use for us.
35.
by kevinlu1248 3y ago
We're building a webhook services on FastAPI + Celery + Redis + Grafana + Loki and the experience with setting up every service incrementally was miserable, and even then it feels like logs are being dropped and we run into reliability
36.
Python `Raise from` Causal Exceptions Syntax(stackoverflow.com)
1 points by kevinlu1248 3y ago|0 comments
37.
by kevinlu1248 3y ago
The new Sweep assistant is also interactive! You can request for a plan, approve the plan to have Sweep run through each of the proposed files to edit. We found that it was really difficult to iterate with an agent that only responds every
38.
by kevinlu1248 3y ago
It probably makes it easier for many companies to move off of OpenAI since they won’t need to drastically alter their codebase.
39.
by kevinlu1248 3y ago
The other problem is with latency: Deepseek 34B on A100s seem slower than GPT-4 but perhaps it will be better on H100s.
40.
by kevinlu1248 3y ago
We have tried many of the current open source models but unfortunately the only model whose capability is close to GPT-4 is Deepseek and unfortunately Deepseek can’t follow our specified format and is very sensitive to prompt changes.
41.
From __future__ Import Braces(stackoverflow.com)
1 points by kevinlu1248 3y ago|0 comments
42.
by kevinlu1248 3y ago
I feel like at least for me, when I want to build fast, writing the implementation and having AI generate the unit tests to double check and catch edge cases is easier. I also don't use the unit tests as a verification but just to ensu
43.
Getting GPT-4 to modify files is harder than getting it to write them(docs.sweep.dev)
7 points by kevinlu1248 3y ago|0 comments
44.
by kevinlu1248 3y ago
They say it's semantic search but it looks more like lexical search
45.
by kevinlu1248 3y ago
How has your experience with function calls been? We tried doing code generation and making ChatGPT generate diffs and it seems to perform worse than the March edition.
46.
by kevinlu1248 3y ago
This makes me believe simulation theory even more tbh. Quantum mechanics exist to fuse operations, altogether making simulating our universe more computationally inexpensive.
47.
Chroma DB Random Seg Faults(github.com)
1 points by kevinlu1248 3y ago|0 comments
48.
Setting Up Python LRU Cache for Methods(medium.vaningelgem.be)
3 points by kevinlu1248 3y ago|0 comments
49.
by kevinlu1248 3y ago
Might be wrong but I think it started with Jupyter/MatplotLib + Python's permissive license. Heard it started all with a physics professor unable to use Matlab for teaching.
50.
Codeblocks: Library for merging incomplete LLM-generated code(github.com)
1 points by kevinlu1248 3y ago|3 comments
51.
by kevinlu1248 3y ago
This is amazing! I think for production and rapid development use-cases though we just use XML for information extraction. It's extremely easy to parse with regex and rarely do the models make mistakes since the start and end tokens ar
52.
by kevinlu1248 3y ago
There's a few tools out there like AgentGPT ( https://github.com/reworkd/AgentGPT , although it's a more conversational interface), and ( https://github.com/logspace-ai/langflow ) and others
53.
by kevinlu1248 3y ago
Aider's awesome and we're using a few of the ideas from the aider blogs like ctags and the search-and-replace editing prompt. I'm wondering, is anyone experimenting with fine-tuning Llama 2 to get it reply in this search-and-
54.
by kevinlu1248 3y ago
Thanks for the suggestion! CLI tools can be simple to deploy and setup but are limited in the long run. Running the logic server side allows us to perform way more validation, such as GitHub Actions, linting, and pre-commit hooks. You can d
55.
Using Retrieval Augmented Generation (RAG) to clear our GitHub backlog(docs.sweep.dev)
52 points by kevinlu1248 3y ago|7 comments
56.
by kevinlu1248 3y ago
Are you referring to the installation.md? It's a bit of a misnomer, we intended for it to be for post-installation instructions for redirect after you install.
57.
by kevinlu1248 3y ago
I think for teams that want to move fast in a non-critical environment (health, finance etc.) something that works 90% of the time is fine. Getting to 95% takes twice the amount of time but does not provide twice the value. When the 5% diff
58.
by kevinlu1248 3y ago
Yup, it's a bit frustrating since it's a problem with LLM, RLHF and fine-tuning for chat. In fact, we also added in the prompt to not say that it did testing. I find that in general it seems really difficult to tell a language mod
59.
by kevinlu1248 3y ago
Just changed the color to royal blue via https://github.com/sweepai/sweep/pull/932
60.
by kevinlu1248 3y ago
I think they meant that now that an AI can write the tests they can bring themselves to write enough to hit the 100% coverage. And I think importance of coverage just depends on if you want to build fast or have better maintenance, but I co

More ›