I cannot evaluate how valuable this model is based on the performance table alone because the comparison is kind of narrow. Waiting for third-party benchmarks to provide more information.
I do not have the hardware to run it properly without super-heavy quantization (about q2). Besides, the architecture is not optimized for local usage (sliding window attention also requires KV cache). I think I will just move on.
At least student movement succeeded in an authoritarian / totalitarian country. The movement was a dangerous move and can be fatal. Please remember that in 1989, Chinese students and workers protested in Tienanmen Square, and got massacred and failed.
At least this parliament is elected by people, not manipulated by a corrupted authoritarian / totalitarian party or such.
At least Serbia society can get a chance to rest or recovery after a lot of destruction happened.
As a person who uses consumer level hardware to do neural network research, this unlock is truly INVALUABLE.
This unlock is super useful as it improves the GPU performance on computation-bounded neural network workloads like LLM prefilling, image/video generation, and general neural network training, in a way of disabling deliberate software segmentation on the same hardware imposed by Nvidia.
I'm on rtx 50 series, and no free performance uplifting for now. Waiting for more reverse engineering efforts to reveal whether the unlock is possible for 50 series.
I'm aware that long-lasting reverse engineering efforts on nvidia gpus, like cmp gpus, will lead to something impactful. I just didn't see it clearly until your project came out. Good job on that.
For now, I think reverse engineering efforts on 50 series gpus are just beginning. There are projects like reading sensor temperatures directly from the gpu itself instead of using deliberately DOWNGRADED software like nvidia-smi. However, I don't really see these efforts spreading, and people behind them really needs more recognition. Props to these people, too.
In China, at least 100 million retired people with rural Hukou (户口) can only get 100-300 CNY (or ~15-45 USD) pension per month from the government. Note that they just pay their taxes like everyone else. That's why raising children so they can help you when you're old (养儿防老) is so wide spreading in rural China.
Ask me how I know this? Most (99.999999%) people in China do not have the freedom to express themselves online safely. You meet them and ask them personally to know this kind of thing exists. Then you check the statistics from Chinese government about pension distribution, retired people with rural Hukou get the most discrimination from the government.
Hilarious. My first thought on these fonts is that similar clip art style text is frequently used in Persona 5. Now I know this is an active design choice.
P.S. I saw keming and my mind went blank for a little while.
The 996 working culture in this page is actually *996.ICU*. "Work by '996', sick in ICU", an ironic saying among Chinese developers, which means that by following the "996" work schedule (9 am to 9 pm, six days a week), you are risking yourself getting into the ICU (Intensive Care Unit). [1]
Take a step further, do you want to overwork to death if you have other choices? One of the reasons behind 277k stars in 996.icu repo is that 996 working culture literally takes over every single corporations in China. With the downturn of China economy and the selective execution of Chinese labor law, in recent years, 996 becomes even harsher to labor force in China because there are more than overwork now, e.g. low salary and low human rights. When you say 'make the average consumer better off', you conveniently ignore the tears, sweets, and blood behind the facade of neijuan.
> I was shown a prompt and told the internal research model had simply been given the problem statement. Levent had been told by Sebastien “very little human input” had been used. This turned out not to be true. Over the course of the call, as members of their team sent Sebastien corrections and details over their internal chat, it emerged that an entire team had been working on the problem, that this was one of a number of things that was tried, that work had started on the unforced problem, that the team first set the model on easier problems, including Euler, that even the prompt that had been shown to me had been written by prompting Codex, and that an insane amount of compute had been used.
> I asked when the first prompt had been sent by them. This question was not answered directly by OpenAI for some time. Eventually it was agreed that it had been sent in the past few days, after information about our work had reached OpenAI.
> I asked whether the model had been trained on, or had access to, our sessions in Codex, into which we had been putting all our drafts for the whole of this project. I was told the model did not look up user data. I asked again, about training, and I did not get an answer.
> Two proposals were offered to me. The first was that we post our Euler result, and that OpenAI post its Navier-Stokes result the next day. The second was that, after posting Euler, I alone write a paper presenting the Navier-Stokes result, acknowledging that an internal OpenAI model had resolved it. Sebastien twice asserted that he wanted Levent removed from authorship, and said it would all be simple if only it were not the case that, and it was so annoying that, Levent works at Anthropic. It was also said that if OpenAI posted after us, they would say that we deserved the Clay Prize, and that we were the “closest humans to the problem”. I declined both offers.
> I said that if OpenAI released its result in the way proposed I would go public with what happened. The reply was, “Why would you ruin your career?” I replied that I am an academic, and asked why he thought going public would ruin my career. The reply was, “If you don’t want me to be nice, then I don’t have to be nice.”
Wow, that's some VERY friendly communication. Besides, will the career of the person be ruined because of “Why would you ruin your career?” came out of his or her own mouth?
In contrast to memory-bounded autoregressive decoding, diffusion decoding can be computation-bounded. Devices with a decent amount of compute but very small memory, e.g. nvidia consumer-grade GPUs, could benefit from diffusion decoding massively.
Performance-wise, I don't think diffusion decoding should be much worse than autoregressive decoding, see https://arxiv.org/html/2604.11035 .
As a local LLM user, I really want to see more small diffusion LLM/VLMs with decent performance coming.
I cannot evaluate how valuable this model is based on the performance table alone because the comparison is kind of narrow. Waiting for third-party benchmarks to provide more information.
I do not have the hardware to run it properly without super-heavy quantization (about q2). Besides, the architecture is not optimized for local usage (sliding window attention also requires KV cache). I think I will just move on.
reply