Hacker News: z4y5f3

New comment by z4y5f3 in "Kimi K3: Open Frontier Intelligence"

z4y5f3 — Thu, 16 Jul 2026 18:26:55 +0000

They will release the weights by 7/27 along with support in vLLM. Stop second guessing. Source: their blog post https://mp.weixin.qq.com/s/V4xhEIy8xDXSMDPrPkmUAQ

New comment by z4y5f3 in "Kimi K3: Open Frontier Intelligence"

z4y5f3 — Thu, 16 Jul 2026 18:25:06 +0000

They will release the full weights by 7/27 along with support in vLLM.

Source: their release blog on WeChat. https://mp.weixin.qq.com/s/V4xhEIy8xDXSMDPrPkmUAQ

New comment by z4y5f3 in "Nvidia's Project Digits is a 'personal AI supercomputer'"

z4y5f3 — Tue, 07 Jan 2025 05:49:46 +0000

NVIDIA is likely citing 1 PFlops at FP 4 sparse (they did this for GB200), so that is 128 TFlops BF16 dense, or 2/3 of what RTX 4090 is capable of. I would put the memory bandwidth at 546 GBps, using the same 512 bit LPDDR5X 8533 Mbps as Apple M4 max.

New comment by z4y5f3 in "IsoFLOP curves of large language models are flat"

z4y5f3 — Fri, 02 Aug 2024 18:20:17 +0000

Yep I have seen this paper before, and thank you for linking it here for reference. My personal opinion is that compared to single epoch scaling laws, we still need more evidence and literature on effects of multiple epochs, but this paper is one of the best results we have so far on using multiple epochs.

New comment by z4y5f3 in "IsoFLOP curves of large language models are flat"

z4y5f3 — Fri, 02 Aug 2024 17:16:26 +0000

What they missed is that current scaling laws (OpenAI, Deepmind Chinchilla) are based on the assumption that the model is trained for one epoch. This essentially means that in order to scale compute, you will have to scale the model size and/or the size of the dataset. So Meta cannot simply spend 3.8e25 FLOPs on a 70B model - to do this they must find 86T pretraining tokens which they do not have.

Of course, ultimately we will figure out scaling laws for LLMs trained on multiple epochs of data, but not today.

New comment by z4y5f3 in "Qwen2 LLM Released"

z4y5f3 — Fri, 07 Jun 2024 20:54:22 +0000

My experience is that < 500M models are pretty useful when fine-tuned on traditional NLP tasks, such as text classification and sentence/token level labeling. A modern LM with a 32K context window size could be a nice replacement for BERT, RoBERTa, BART.

New comment by z4y5f3 in "PCIe 7.0 Draft 0.5 Spec: 512 GB/s over PCIe x16 On Track For 2025"

z4y5f3 — Sun, 07 Apr 2024 01:49:48 +0000

Checked your numbers in another thread - excellent breakdown. Thanks for the clarification.

I did not read the original anandtech post so I did not realize the 512GBps already refers to the full-duplex bandwidth. You are right that PCIe 7.0 x16 sits between V100 and A100 NVLink.

New comment by z4y5f3 in "PCIe 7.0 Draft 0.5 Spec: 512 GB/s over PCIe x16 On Track For 2025"

z4y5f3 — Thu, 04 Apr 2024 18:27:47 +0000

NVLink advertises combined bandwidth in both direction, so the 1800 GBps NVLink on Blackwell is actually 900 GBps for everyone else. PCIe can also do multi-node direct transfer via PCIe switches and has been already widely adopted. NVLink still has the power and chip arena advantage even if the bandwidth is similar.

New comment by z4y5f3 in "PCIe 7.0 Draft 0.5 Spec: 512 GB/s over PCIe x16 On Track For 2025"

z4y5f3 — Thu, 04 Apr 2024 18:23:31 +0000

Depends. NVLink advertises bidirectional bandwidth, whereas PCIe and standard networking calculate bandwidth in a single direction. So a 1800 GBps NVLink is actually 900 GBps in PCIe and standarding networking terms.

Therefore, a 512 GBps PCIe would sit between the current H100 NVLink (450 GBps) and next generation B200 NVLink (900 GBps). With that being said, NVLink still has lower power draw and smaller chip area, so it would still have a competitive advantage even if the bandwidth is similar.

New comment by z4y5f3 in "NY Times copyright suit wants OpenAI to delete all GPT instances"

z4y5f3 — Fri, 29 Dec 2023 08:36:44 +0000

Unfortunately GZIP won't beat LLMs for text classification. The research you cited is just poorly done science that has been widely debunked. The original paper compared top-2 accuracy of GZIP with top-1 accuracy with BERT. The dataset also contains a lot of train/test data leakage. See this article for the rebuttal: https://kenschutte.com/gzip-knn-paper/ and this thread for a previous discussion on hackernews: https://news.ycombinator.com/item?id=36758433.

Further, the evidence presented by NYT in the lawsuit could be hard to reproduce. I tried multiple prompts on multiple versions of GPT-4 APIs but still could not get GPT-4 to reproduce NYT articles exactly. NYT might as well tried to let GPT-4 reproduce 100,000 articles and only found a few cases where GPT-4 actually recited the whole article. In that case OpenAI might as well be arguing that this is only a rare bug and avoid losing the lawsuit in a massive way.