Viewing profile — serialx

Look into all-smi https://github.com/lablup/all-smi It supports all GPUs thinkable including Apple Silicon and many AI accelerator cards.

Yeah, attention sinks were applied to gpt-oss

Ah sorry, you might be right. I meant "sparse reward" as a reward system that is mostly 0 but occasionally 1. Your "sparse reward" means only providing reward at the end of each ou…

I don't think it's only using sparse rewards because of the format rewards. The training recipe is pretty comprehensive and involves multiple stages.[1] The paper mentions that whe…

So to my understanding, this work reproduces DeepSeek R1's reinforcement learning mechanism in a very small language model. The AI gets "rewards" (like points) for doing two things…

Change the currency to USD

What are the compute requirements for solving efficient energy grid transmission? Is there a efficient algorithm that is able to solve this?

Is there any plans to add Well-known URI[1] as a standard? It would be awesome if we can add services just by inputting domain names of the services. [1]: https://en.wikipedia.org/…

PSA: You can also use singleflight[1] to solve the problem. This prevents the thundering herd problem. Pocache is an interesting/alternative way to solve thundering herd indeed! [1…

Actually, llama.cpp running on Apple silicon uses GPU(Metal Compute Shader) to inference LLM models. Token generation is also very memory bandwidth bottlenecked. On high end Apple …

Well, iPhone already does that with photos. :)

Now we just need VkDNN