jsonformer
zod
jsonformer | zod | |
---|---|---|
25 | 292 | |
3,868 | 30,630 | |
- | - | |
5.4 | 9.1 | |
3 months ago | 6 days ago | |
Jupyter Notebook | TypeScript | |
MIT License | MIT License |
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.
jsonformer
- Forcing AI to Follow a Specific Answer Pattern Using GBNF Grammar
-
Refact LLM: New 1.6B code model reaches 32% HumanEval and is SOTA for the size
- Tools like jsonformer https://github.com/1rgs/jsonformer are not possible with OpenAIs API.
-
Show HN: LLMs can generate valid JSON 100% of the time
How does this compare in terms of latency, cost, and effectiveness to jsonformer? https://github.com/1rgs/jsonformer
-
Ask HN: Explain how size of input changes ChatGPT performance
You're correct with interpreting how the model works wrt it returning tokens one at a time. The model returns one token, and the entire context window gets shifted right by one to for account it when generating the next one.
As for model performance at different context sizes, it's seems a bit complicated. From what I understand, even if models are tweaked (for example using the superHOT RoPE hack or sparse attention) to be able to use longer contexts, they still have to be fined tuned on input of this increased context to actually utilize it, but performance seems to degrade regardless as input length increases.
For your question about fine tuning models to respond with only "yes" or "no", I recommend looking into how the jsonformers library works: https://github.com/1rgs/jsonformer . Essentially, you still let the model generate many tokens for the next position, and only accept the ones that satisfy certain criteria (such as the token for "yes" and the token for "no".
You can do this with openAI API too, using tiktoken https://twitter.com/AAAzzam/status/1669753722828730378?t=d_W... . Be careful though as results will be different on different selections of tokens, as "YES", "Yes", "yes", etc are all different tokens to the best of my knowledge
- A framework to securely use LLMs in companies – Part 1: Overview of Risks
-
LLMs for Schema Augmentation
From here, we just need to continue generating tokens until we get to a closing quote. This approach was borrowed from Jsonformer which uses a similar approach to induce LLMs to generate structured output. Continuing to do so for each property using Replit's code LLM gives the following output:
-
Doesn't a 4090 massively overpower a 3090 for running local LLMs?
https://github.com/1rgs/jsonformer or https://github.com/microsoft/guidance may help get better results, but I ended up with a bit more of a custom solution.
-
“Sam altman won't tell you that GPT-4 has 220B parameters and is 16-way mixture model with 8 sets of weights”
I think function calling is just JSONformer idk: https://github.com/1rgs/jsonformer
- Inference Speed vs. Quality Hacks?
-
Best bet for parseable output?
jsonformer: https://github.com/1rgs/jsonformer
zod
-
Simplifying Form Validation with Zod and React Hook Form
[Zod Documentation](https://zod.dev/) [Zod Error Handling](https://zod.dev/ERROR_HANDLING?id=error-handling-in-zod) [React-Hook-Form Documentation](https://react-hook-form.com/get-started) [Hookform Resolvers](https://www.npmjs.com/package/@hookform/resolvers)
-
Figma's Journey to TypeScript
This is a very fair comment, and you seem open to understanding why types are useful.
"problems that are due to typing" is a very difficult thing to unpack because types can mean _so_ many things.
Static types are absolutely useless (and, really, a net negative) if you're not using them well.
Types don't help if you don't spend the time modeling with the type system. You can use the type system to your advantage to prevent invalid states from being represented _at all_.
As an example, consider a music player that keeps track of the current song and the current position in the song.
If you model this naively you might do something like: https://gist.github.com/shepherdjerred/d0f57c99bfd69cf9eada4...
In the example above you _are_ using types. It might not be obvious that some of these issues can be solved with stronger types, that is, you might say that "You rarely see problems that are due to typing".
Here's an example where the type system can give you a lot more safety: https://gist.github.com/shepherdjerred/0976bc9d86f0a19a75757...
You'll notice that this kind of safety is pretty limited. If you're going to write a music app, you'll probably need API calls, local storage, URL routes, etc.
TypeScript's typechecking ends at the "boundaries" of the type system, e.g. it cannot automatically typecheck your fetch or localStorage calls return the correct types. If you're casting, you're bypassing the type systems and making it worthless. Runtime type checking libraries like Zod [0] can take care of this for you and are able to typecheck at the boundaries of your app so that the type system can work _extremely_ well.
[0]: https://zod.dev/ note: I mentioned Zod because I like it. There are _many_ similar libraries.
-
From Flaky to Flawless: Angular API Response Management with Zod
Zod is an open-source schema declaration and validation library that emphasizes TypeScript. It can refer to any data type, from simple to complex. Zod eliminates duplicative type declarations by inferring static TypeScript types and allows easy composition of complex data structures from simpler ones. It has no dependencies, is compatible with Node.js and modern browsers, and has a concise, chainable interface. Zod is lightweight (8kb when zipped), immutable, with methods returning new instances. It encourages parsing over validation and is not limited to TypeScript but works well with JavaScript as well.
- TypeScript Essentials: Distinguishing Types with Branding
-
You can’t run away from runtime errors using TypeScript
Zod is a TypeScript-first schema declaration and validation library. It helps create schemas for any data type and is very developer-friendly. Zod has the functional approach of "parse, don't validate." It supports coercion in all primitive types.
-
Best Next.js Libraries and Tools in 2024
Link: https://zod.dev/
-
Popular Libraries For Building Type-safe Web Application APIs
You can check out their documentation here.
-
Epic Next JS 14 Tutorial Part 4: How To Handle Login And Authentication in Next.js
You can learn more about Zod on their website here.
-
What even is a JSON number?
In JS, it's a good idea anyway to use some JSON parsing library instead of JSON.parse.
With Zod, you can use z.bigint() parser. If you take the "parse any JSON" snippet https://zod.dev/?id=json-type and change z.number() to z.bigint(), it should do what you are looking for.
-
Error handling in our form component for the NextAuth CredentialsProvider
We will validate our input using client-side zod. Zod handles TypeScript-first schema validation with static type inference. This means that it will not only validate your fields, it will also set types on validated fields.
What are some alternatives?
mlc-llm - Enable everyone to develop, optimize and deploy AI models natively on everyone's devices.
class-validator - Decorator-based property validation for classes.
aider - aider is AI pair programming in your terminal
joi - The most powerful data validation library for JS [Moved to: https://github.com/sideway/joi]
clownfish - Constrained Decoding for LLMs against JSON Schema
typebox - Json Schema Type Builder with Static Type Resolution for TypeScript
outlines - Structured Text Generation
Yup - Dead simple Object schema validation
gpt-json - Structured and typehinted GPT responses in Python
ajv - The fastest JSON schema Validator. Supports JSON Schema draft-04/06/07/2019-09/2020-12 and JSON Type Definition (RFC8927)
jikkou - The Open source Resource as Code framework for Apache Kafka
io-ts - Runtime type system for IO decoding/encoding