Hacker News new | past | comments | ask | show | jobs | submit | lovelearning's comments login

I'm not sure how SOTA it is but the sentence about connecting DRAM differently reminded me of Cerebras' scalable MemoryX and its "weight streaming" architecture to their custom ASIC. You may find it interesting.

[1]: https://cerebras.ai/press-release/cerebras-systems-announces...

[2]: https://cerebras.ai/chip/announcing-the-cerebras-architectur...


Yeah, Cerebras seems to be the SOTA. I suspect we need something more radically different for truly memory-centric computing that will be significantly more efficient.

Those medical personnel may also be members of Hezbollah. Doctors, nurses, and emergency responders can have ideological beliefs. For example, in my country, there are medical personnel with far-right and extremist views, some even holding leadership positions in such ideological organizations.


It's surely possible and or even plausible ideology has infiltrated civic institutions. Still it's quite a stretch to assume any victim of what looks like a supply chain attack is guilty, simply because of the scale of the attack.


It is evil to go around blowing up people for what they "might believe".


This visual IDE for LLM pipelines was posted recently: https://github.com/langgenius/dify

See if it helps.


There's an implicit assumption here that the ban enforcement will involve/require adversarial technical approaches.

But I think it'll be similar to India's 4-year-old ban on TikTok where the local corporate entity itself has cooperated with the ban to avoid legal repercussions. TikTok has been detecting IP addresses from India and showing a legal notice that it's banned in the country due to a government/court order.


Ah, I wasn't aware of the India ban or how it was handled.

I guess I assumed TikTok might try to keep operating in the US out of defiance.


From the PDF:

> Requirements for 6G include extremely high data rates (>100 Gbit/s), ultra-low latency (<0.1 ms)....

> For the wireless communication of the 6G era, new radio frequency bands in sub-THz, ranging from 100 GHz to 300 GHz, have been identified as one of the most promising bands.

> There is also a technical challenge for seamlessly connecting wireless communication systems with fiber-optic communication networks


This is the most awesome thing I've come across on HN in recent months if not years!


I started with "The Story of Philosophy" by Will Durant. It introduced me to the high-level strands and terms in philosophy by studying the works of eminent philosophers.


I find its documentation quite poor though: "If specified, adds a new batch of zeros to the key and value sequences at dim=1."

Doesn't describe the implications even briefly. If they add just your second sentence to that description, it'll immediately become so much more useful.


Spending actual money on GCP. I don't have any GCP credits.

I asked for just 1 GPU and explained that I need to run 20GB+ models. Didn't realize they expect a doctorate application in that 3-line textbox.

All of it is so bureaucratic. I have to look up their zone-GPU mapping. I have to test availability indirectly via the quota workflow route. It reminds me of my country's shitty bureaucracy!

I tried 3 zones. In one zone, a GPU is granted in the quota but still shows quota error on the main UI. The other 2 zones were denied.

If they don't have any to rent, I don't know why they don't just say that. Instead, they expect an onerous workflow of requesting quota increases with pathetic user experience, only to deny them and notify the denials via email.

Any suggestions for Colab alternatives?


I have had no issues with GCP quotas, you are exaggerating the situation out of frustration, understandable. The turn around is typically a few minutes, and if you haven't tried this on other clouds, you don't know how good that is.

There is a single page list of GPU availability by zone: https://cloud.google.com/compute/docs/gpus/gpu-regions-zones

You do not have to write a "doctorate" but you do need to explain your need, not the what will you do.

again, A100s are in very high demand, so everyone is preferring the highest value customers first.

Did you get a A100 in one zone? The error messages for "you are out of quota" vs "(google) resource not available" look similar. We have devs getting them confused all the time


I don't think I'm exaggerating - I wasted hours today on their terrible workflow and UI. All of that can be simplified and automated by Google via their frontend instead of a workflow that expects users to visit multiple pages.

I have a quota of 1 GPU in one zone and zero allocated GPUs. I keep seeing only "exceeded quota by 1 GPU" error on their UI when trying to create a VM for that single GPU in the correct zone.

Thank you for the replies. Looks like most users are fine with all this. I'll probably explore non-GCP alternatives.


you're welcome

For GPUs (among others), there are two quotas, regional and zonal. We have N regional T4s and unlimited zonal. Maybe you have 1 zonal and 0 regional?

If you just need a GPU, there may indeed be better or cheaper providers


It's still "in-context learning" as per the GPT-3 definition because they are supplying some demonstrations of the task in the prompt.

Only thing special is that the input for each demonstration is obtained through embedding-based retrieval.


Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: