Відео

interesting. Using AI in football.

Dude why suggesting Discord community?

Love these gnn talks!

nice talk !

I can imagine this being used for war

I like how it has become a regular ritual for these talks that the speaker tells any questions are welcome in between the speech and then Aleksa reminding everyone that they should still raise hands XD

great talk so far. can you maybe create a in depth tutorial for cloud compute? how to get the dataset to the cloud and so on.

We talked about Petar's latest work on Tactic AI! Let me know how you like this one!

Is DSA really important for a Research Engineer/Scientist Role- ML field? Can it be ignored? Isn't ML foundations, Statistics and Math enough?

Great job, way to go :)

Are these chips compatible with an ordinary computer like incase if i want to use the chip in an offline model. and is it on sale

1:20:00 is there a proper explanation of why there is a logit scale factor when calculating similarity? Thanks.

why not i9 14900k?

man I normally love your videos but this one is a bit unclear

So how exactly is this making money?

Hi, i am always confused about the forward process equation defined in (2). we say the our images x come from an unknown distribution q(.). but in equation (2) we are saying that this distribution is normal ? we are sampling from a normal distribution to get next forward step. sorry I am not that good when it comes to probability theory.

Gabarjobat?

Thank u so much

Simple eh?

loving this series, will be sad after i watch the fourth and final video :(

Loved this, thanks!

Hi, I love your videos, especially the one about MuZero and VQ-VAEs. I recently stumped upon a paper about a Stochastic Muzero, which combines the two. Also, I found an improved version of MuZero called Efficient Muzero. It would be awesome if you would cover one of the two as your explanation really helps to build the intuition about models.

This video help me understand the architecture of the model. Great job!!! What changes do we have to do in the code in order to take images 512 or 1024 ? Which part of the code should we change?

does it work when I install it with deb (Network ) ?

So the parallelism you demo'd with pmap, that was data parallelism correct? replicating the whole model across all the devices, sending different batches to each device, and then collecting the mean model back on the host device after forward and backwards pass? am i understanding that correctly?

ty ty ty ty ty for this video

how to did you purge the incorrect cudnn that was installed first?

awesome video

Does this still exist?

How does the fine-tuning actually work here? They don't talk about it in the paper, right?

Yes, it is a kind of parameterized classification though embedding vector from text input as the last Fully connected layers in the classification network of image classification network.

if you drop the attention for the more past tokens, then they begin to specialise themsselves for the local computation, and their computation will not be lost the further you go into the futur because it will be more referenced in the higher layers and will come back as a general past.

question regarding DETR bipartite Matching loss If I got it right than the loss match for every target only one query expert Therefor if there are 10 targets than the loss guide the model to match 10 query to the target classes and 290 (out of 300) to the no-object class is this correct? If so, than it might happen that a query that was not match to a target, but also was very close to match the target will be panelize by the loss model (panelize for not showing the no-obj class). This might confuse the model, especially if in the next epoch the the matched query and the unmatched query will be flipped. If I got it correct than the penalty for not showing the no-obj class is low, but it still shouldn't be panelize. I think it would be much better to include in the loss all the query that matched the target, so that they could be use for by the loss to improve the model This will cause the model to produce multiple prediction per target, but during prediction we can use only the prediction with the best score and ignore the rest (or use NMS)

Wow! I watch to the end, and learned a lot. Few remarks * I think it would be better to break it into a serie * The first hour was very clear, but till the end it became a bit vague * I download the mini coco and modified the download script as u suggested, but needed chatgpt help to generate the appropriate annotation file Thanks 💫

Please help, when using expressivity_predict I get the error expressivity_predict is gated. How to fix it?

Is this method lossy or lossless?

3:51 PackagesNotFoundError: The following packages are not available from current channels: - torchvission torchvission 0.5.0 is not available to install , can someone please help me out?

the video doesnt demonstrate the actual steps to actually deploy large models using deepspeed? which is much more important than understanding all the theory stuff

get prices, and while Groq seems awesome, no one asked how much a groq system that compares with a H100 ( for LLM ).

great video thanks for explaining so nicely. Can you pls share the code as well?

Your haiku video never came out sir 😅, thanks for this series btw

Has anyone run a raytracer on this yet? could be interesting.

Jako dobar interviju. Samo tako nastavi!

Thanks for your effort. The video is perfect. I have a question. Can I detect objects in video with this?

Very interesting.

Great presentation!

You mentioned posterior and prior, can you provide some reference, why they model it in this way ?

Not a genius, just a massive data thief, nothing more.

Does CPU need large cache size?

This is such a great presentation!