Google Coral TPU USB-Accelarator (Featuring QNAP): Unboxing & Testing

In this video we take a closer look at the AI accelerator TPU from Coral/Google. We will unbox, and try it out using QNAP server with QuMagie and AI Core, to see if things speed up in terms of face, objects proccesing.
Buy Here: www.conrad.com...
---
Website: www.whcvip.com
---
#coral #usbaccelerator #tpu

[@nadiaezzarhouni300]

Try stable diffusion on it

[@Justdigit2]

But does it regonize better?? And more options?? Like cars bikes etc??

[@the009]

yes they do. You can process about 140fps with one TPU

[@rajesh1980m]

thanks for the Video ... was looking for 1 such stuff

[@teabagNBG]

im not getting what these things are used for ??

[@jtn191]

Machine learning. Alternately, expensive gpus handle these tasks

[@talleycollins9353]

tpu is a special matrix processor for processing llm tensor flow... it makes making massive machine learning computations SUPER FAST... Your ml platform will recognize it and utilize it! Without it you would need multi gpu's to do the same thing! You definitely wouldnt be able to run anything efficiently locally without it!

[@henrikcarlsen1881]

@@talleycollins9353 So, if I want to do some AI I use a web-based machine learning (ML) platforms (like Datarobot, I guess) to crunch my data? If I plug in a Coral TPU they will allocate some tasks to it, perhaps all? He's feeding the Qnap-data to the ML-platform but are getting it processed at the Coral TPU. To detect load, he watched the Qnap-joblist. That's it, isn't it (I was really distracted by his thick Danish accent and understod zilch)?

[@Chris-gh3om]

Useful information, thank you for your test, but I have a question that I would like you to test. That is, can other containers, such as jellyfin and plex, use this TPU for encoding and decoding?

[@subhangmokkarala]

No we can't use it for media encoding

[@Chris-gh3om]

In addition to image recognition and thumbnail generation, is there any other use for TPU on QNAP?@@subhangmokkarala

[@tohur]

this could possibly make it more viable to run our AI models on local hardware.. I about to pick one up and see if I can get it to work work KoboldAI and few other AI back ends and front ends

[@pn4960]

Hello, any results ?

[@treudden]

Unfortunately I don't think running bigger models is possible since the TPU only has 8MB of RAM

[@subhangmokkarala]

We cant run bigger models on it we can run only tf lite models

[@SeifEddineB]

for 100 Euro, I think I could get a better NAS with a faster CPU and get these tasks done faster? I also own a TS 264, and honestly, I am not very impressed by the speed offered by this TPU, as i have synchronized some 20.000 photos into the NAS, it took it two days to process using its tiny N5105 but kept working fine, generating thumbnails, streaming to Plex, recording a single IP camera.., this tpu feels meh, unless you have more results to show us?

[@abhishektidke7303]

Can we work on OpenVino using this?

[@kklol07]

where did u get this

[@whcvip]

Link in description.

[@lancelotxavier9084]

Hard disk bottleneck.

[@abzs5811]

I can be the only one that doesn’t understand a single goddamn word