4 ms·

I am pretty good at devices like this. Have done a lot of raspberry pi development, arduino, nvidia jetson etc. for real ML production application. I ordered a

I am pretty good at devices like this. Have done a lot of raspberry pi development, arduino, nvidia jetson etc. for real ML production application. I ordered a Coral before and just had no luck. Worked for some contrived applications, but had nothing but trouble with the drivers and anything remotely custom just failed. It's cool tech, but I'll pass on this vanity project that will undoubtedly get cancelled right as people start to rely on it.

joshvm 5y ago
I'm on the fence. It's a very nice device if you can get your models working on it - basically untouched at the price/power point. Drivers for me have been OK. I have an M.2 card connected to a Jetson devkit (makes for a nice embedded test bench) and it runs fine, no worse than the NCS for setup anyway^. There were a couple of PCI settings to tweak but I documented the setup here [0]. For common use cases it's a decent option, I think. For custom models you really need to know what you're doing. The main issue I've had is that the compiler behaviour differs between versions (and it's very difficult to find older releases), so where previously you could run a big model and delegate things to the CPU, now it sometimes won't compile at all. There were also problems where we trained a model in AutoML - using free credits but the real cost would have been over $100 - and the edgetpu compiled model lost a lot of performance for no apparent reason. The developers have been very helpful when I've contacted them, and generally you can get through to real devs (not generic support) who can look at your model for you. Mostly I think you need to take care when training models for these devices, but quantisation-aware training is not trivial to use in Tensorflow and there are only a few off-the-shelf models which are supported in the various toolkits. Model maker looks promising, but it's also finnicky in my experience [1]. I guess that's my main caution: it's often very difficult to figure out why your model performance goes to crap when you compile it, especially if the input TFLite model is fine. I'm not super worried about hardware availability. They're suffering from the chip shortage like everyone else, so it's not surprising that lead times are long. I was able to buy my device in late 2020 without any trouble. They also make a nice little environmental sensor + OLED board that's compatible with Pi headers and is the same form factor as the Pi Zero. A bit odd, because on the Coral dev board the heatsink is squarely in the way and you have to buy an extension header for it. [0] https://github.com/jveitchmichaelis/edgetpu-yolo/blob/main/hardware.md https://github.com/jveitchmichaelis/edgetpu-yolo/blob/main/h... [1] https://www.tensorflow.org/lite/guide/model_maker https://www.tensorflow.org/lite/guide/model_maker ^Remember that time when Intel released an OpenVINO installer that clobbered any existing OpenCV build on your system?

joshvm 5y ago

I'm on the fence. It's a very nice device if you can get your models working on it - basically untouched at the price/power point. Drivers for me have been OK. I have an M.2 card connected to a Jetson devkit (makes for a nice embedded test bench) and it runs fine, no worse than the NCS for setup anyway^. There were a couple of PCI settings to tweak but I documented the setup here [0]. For common use cases it's a decent option, I think. For custom models you really need to know what you're doing. The main issue I've had is that the compiler behaviour differs between versions (and it's very difficult to find older releases), so where previously you could run a big model and delegate things to the CPU, now it sometimes won't compile at all. There were also problems where we trained a model in AutoML - using free credits but the real cost would have been over $100 - and the edgetpu compiled model lost a lot of performance for no apparent reason. The developers have been very helpful when I've contacted them, and generally you can get through to real devs (not generic support) who can look at your model for you. Mostly I think you need to take care when training models for these devices, but quantisation-aware training is not trivial to use in Tensorflow and there are only a few off-the-shelf models which are supported in the various toolkits. Model maker looks promising, but it's also finnicky in my experience [1]. I guess that's my main caution: it's often very difficult to figure out why your model performance goes to crap when you compile it, especially if the input TFLite model is fine. I'm not super worried about hardware availability. They're suffering from the chip shortage like everyone else, so it's not surprising that lead times are long. I was able to buy my device in late 2020 without any trouble. They also make a nice little environmental sensor + OLED board that's compatible with Pi headers and is the same form factor as the Pi Zero. A bit odd, because on the Coral dev board the heatsink is squarely in the way and you have to buy an extension header for it. [0] https://github.com/jveitchmichaelis/edgetpu-yolo/blob/main/hardware.md https://github.com/jveitchmichaelis/edgetpu-yolo/blob/main/h... [1] https://www.tensorflow.org/lite/guide/model_maker https://www.tensorflow.org/lite/guide/model_maker ^Remember that time when Intel released an OpenVINO installer that clobbered any existing OpenCV build on your system?