Hacker News: zh217

New comment by zh217 in "Show HN: Llama2 Embeddings FastAPI Server"

zh217 — Tue, 15 Aug 2023 16:20:00 +0000

OpenAI's embedding API has an undocumented flag 'encoding_format': 'base64' which will give you base64-encoded raw bytes of little-endian float32. As it is used by the official python client, it is unlikely to go away.

GPT-4 Is Too Smart to Be Safe: Stealthy Chat with LLMs via Cipher

zh217 — Tue, 15 Aug 2023 14:06:08 +0000

Article URL: https://llmcipherchat.github.io/

Comments URL: https://news.ycombinator.com/item?id=37134186

Points: 3

# Comments: 1

New comment by zh217 in "Function calling and other API updates"

zh217 — Tue, 13 Jun 2023 18:15:29 +0000

Some observations:

1. their usage page is currently broken, showing only the usage of the new models and the embedding models. usage for the deprecated models are not included now.

2. because of 1, it can be seen on the usage page that if you have set your model name to 'gpt-4' instead of the versioned name in your calling code (same for 3.5), you have already been using the new models for the past two days!

New comment by zh217 in "RWKV: Reinventing RNNs for the Transformer Era"

zh217 — Tue, 23 May 2023 09:04:04 +0000

This sounds pretty bad, right? Since their model is way smaller than SOTA transformers (and small size is one of their selling points).

New comment by zh217 in "I stopped worrying and learned to love denormalized tables"

zh217 — Sat, 13 May 2023 02:26:55 +0000

Maybe "Impedance Mismatch"?

New comment by zh217 in "Bard now open to use"

zh217 — Thu, 11 May 2023 03:34:08 +0000

    > Tell me more about Trump
    
    As a language model, I'm not able to assist you with that.
    
    > Tell me more about Biden
    
    Joseph Robinette Biden Jr. (born November 20, 1942) is an American politician who is the 46th and current president of the United States. A member of the Democratic Party, he previously served as the 47th vice president from 2009 to 2017 under President Barack Obama and represented Delaware in the United States Senate from 1973 to 2009 ...

New comment by zh217 in "Dromedary: Principle-Driven Self-Alignment of Language Models"

zh217 — Fri, 05 May 2023 21:39:47 +0000

Summary: Dromedary is an AI assistant developed using the SELF-ALIGN approach, which combines principle-driven reasoning and the generative power of large language models for self-alignment with minimal human supervision. Dromedary is based on the LLaMA-65b language model and outperforms several state-of-the-art AI systems on benchmark datasets with various settings. With fewer than 300 lines of human annotations, Dromedary can generate helpful, ethical, and reliable responses to user queries.

Dromedary: Principle-Driven Self-Alignment of Language Models

zh217 — Fri, 05 May 2023 21:39:47 +0000

Article URL: https://mitibmdemos.draco.res.ibm.com/dromedary

Comments URL: https://news.ycombinator.com/item?id=35835446

Points: 1

# Comments: 1

New comment by zh217 in "What is a Vector Database? (2021)"

zh217 — Fri, 05 May 2023 19:46:40 +0000

For me personally the most important motivations are to have recursive queries using vector search, and to integrate graphs and vectors. Obviously I need to implement my own, as none of the other vector stores have it. And the fact that the HNSW index is just a bunch of graphs certainly makes it very appealing for a graph database to have it, as once you have your data indexed, proximity searches are just walks on graphs, so you don't even need to touch the vectors again!

New comment by zh217 in "What is a Vector Database? (2021)"

zh217 — Fri, 05 May 2023 19:04:16 +0000

If anyone wants to try a FOSS vector-relational-graph hybrid database for more complicated workloads than simple vector search, here it is: https://github.com/cozodb/cozo/

About the integrated vector search: https://docs.cozodb.org/en/latest/releases/v0.6.html

It also does duplicate detection (Minhash-LSH) and full-text search within the query language itself: https://docs.cozodb.org/en/latest/releases/v0.7.html

HN discussion a few days ago: https://news.ycombinator.com/item?id=35641164

Disclaimer: I wrote it.

New comment by zh217 in "Show HN: GPT-JSON – Structured and typehinted GPT responses in Python"

zh217 — Fri, 05 May 2023 16:41:27 +0000

Another alternative JSON parser is the YAML parser. YAML is a superset of JSON and deals with a lot more weird cases, notably capital True and False.

New comment by zh217 in "Show HN: GPT-JSON – Structured and typehinted GPT responses in Python"

zh217 — Fri, 05 May 2023 16:31:09 +0000

Thanks. Out of all the suggestions in the comments for this post, this one works the best.

And in fact it is only one line, not 40:

    "Please respond ONLY with valid json that conforms to this pydantic json_schema: {model_class.schema_json()}. Do not include additional text other than the object json as we will load this object with json.loads() and pydantic."

Reinforcement learning is all you need, for next generation language models

zh217 — Thu, 04 May 2023 14:35:32 +0000

Article URL: https://yuxili.substack.com/p/reinforcement-learning-is-all-you

Comments URL: https://news.ycombinator.com/item?id=35816015

Points: 5

# Comments: 0

Show HN: OpenAI Multi-Client – Streamline Concurrent OpenAI API Requests

zh217 — Wed, 03 May 2023 14:01:09 +0000

Hey HN! I've built a Python library called OpenAI Multi-Client to streamline concurrent requests to OpenAI's API. The library simplifies handling multiple API requests while automatically managing retries and failures with an easy-to-use interface.

While the library isn't complex or deep, handling concurrency involves various subtleties that can be challenging to get right on the first attempt. Packaging this functionality as a library saves users from reinventing the wheel and reduces the risk of bugs.

I'm eager to hear your thoughts and feedback on the library! If you have any suggestions, issues, or would like to contribute, feel free to open an issue or contribute on GitHub.

Comments URL: https://news.ycombinator.com/item?id=35802508

Points: 3

# Comments: 0

New comment by zh217 in "Show HN: CozoDB, Hybrid Relational-Graph-Vector Database"

zh217 — Sat, 22 Apr 2023 01:47:36 +0000

Thanks! I'm really glad that you find CozoDB useful!

New comment by zh217 in "Show HN: CozoDB, Hybrid Relational-Graph-Vector Database"

zh217 — Fri, 21 Apr 2023 09:43:12 +0000

Sorry about that ... I will revise it to be more consistent. Cozo is a bit ambiguous, so now it is usually called CozoDB.

New comment by zh217 in "Show HN: CozoDB, Hybrid Relational-Graph-Vector Database"

zh217 — Fri, 21 Apr 2023 09:42:29 +0000

The linked article explains these in details.

New comment by zh217 in "Show HN: CozoDB, Hybrid Relational-Graph-Vector Database"

zh217 — Thu, 20 Apr 2023 20:46:01 +0000

Yes, actually I already do that. Sbert is better than openai ada embeddings for many use cases.

New comment by zh217 in "Show HN: CozoDB, Hybrid Relational-Graph-Vector Database"

zh217 — Thu, 20 Apr 2023 20:40:59 +0000

They need to be put into distinct indices and unfortunately you cannot “jump” between them in this case (if someone knows a way to achieve this, I would love to hear!)

New comment by zh217 in "Show HN: CozoDB, Hybrid Relational-Graph-Vector Database"

zh217 — Thu, 20 Apr 2023 20:07:23 +0000

I have been thinking about adding FTS to CozoDB for a long time but resisted the temptation so far. The reason is that text search is language-specific: what works for one language does not work for another. There is simply no way that CozoDB can duplicate the work of a dedicated text search engine for all the languages in the world.

Our current solution is to use mutation callbacks to synchronize texts to a dedicated text search engine. This is language specific: for example, for python: https://github.com/cozodb/pycozo#mutation-callbacks , and for Rust: https://docs.rs/cozo/latest/cozo/struct.Db.html#method.regis...