Viewing profile — skyde

Actually not surprised. I guess this is for the same reason “say it twice” [1] is working. Because LLm are trained as causal language model, past token cannot attend to future toke…

this is CUDA backend to MLX not MLX backend for CUDA!

Can you give more detail on what you mean by it can be a valuable experience with the right people around to help. My son (7 years old) is gifted in Math and as a parent I find it …

is it only me or this completely miss all the recent research on causal inference using causal graphical model ?

Des os is the best graphic calculator ever built. And its amazing it has un directly in your browser or without internet on your phone. Just wish it was open source :-) Anyone know…

how does this compare to ESP32-S3-BOX-3B ?

Thanks a lot for writing that. I agree 100% with you. But I always wondered how polymath like Leonard davinci and Isaac newton that are excellent in many area are possible.

Is MIT class also taught by outsourced instructor instead of MIT instructor?

Redis Sentinel provides high availability and monitoring for Redis, but it does not guarantee strong consistency. Linearizability requires that once a write is acknowledged, all su…

Paxos and Raft are consensus algorithms that provide certain guarantees and capabilities that a master-slave system with synchronous replication, such as PostgreSQL, cannot offer. …

Redis is a very bad store for a distributed lock but Postgres is only slightly better. What you truly need is something like ZooKeeper and etcd that are designed to achieve distrib…

But inside on epoch there is a lot of duplication already. By duplication I mean if context length is N there is many sequence of N word that are not unique.

Could not try it. Saying valid institutional or company email address. It doesn’t recognize my university.

It “work” but the LLM having to use the calculator mean the LLM doesn’t understand arithmetic enough and doesn’t know how to use an follow a set of step (algorithm ) natively to fi…

Just discovered e-graph recently and I have a good understanding of compiler from taking compiler class at university. I would like to understand why you say e-graph would need con…

https://github.com/uwplse/tensat

What do you mean by close to CNN? What is your architecture? Is it just a fully connected layer of chebyshev?

Given that Alice has 13 brothers and 31 sisters, we can update the Prolog program with this information. We need to adjust the fact about Alice's siblings and then use the rule to …

Asking gpt to first output prolog program seem to 100% fix it! Given that Alice has 13 brothers and 31 sisters, we can update the Prolog program with this information. We need to a…

What do you mean by simplest in term of optimization? I get it find solution that are easy for SGD or Adam optimizer to find. But why would such solution be less simple than other …

Where is the code for it ?

It seems it has been done before: "Syntax-Aware Transformer Models for Neural Machine Translation" by Yang et al. (2019). This model enhances the transformer architecture with synt…

Why not apply same concept every time a word is split into more than one token? Basically if a word contain a Prefix, suffix or root word. We could have a token position relative t…

Forest are good at classification but they cannot leverage pre-training on unclassified data.

As a parent myself I see this happening daily. Teacher public shaming my kids for having pretzel in his lunch saying it’s not healthy. Then later the same day the school give all t…