pilooch

19 ms·

151.
by pilooch 9y ago
Research and experimentation with novel architectures has switched to pytorch everywhere I look. All colleagues and myself outside production pipelines have switched to pytorch. TF is good but too messy and hard for R&D, that's the
152.
by pilooch 9y ago
With all due respect you don't appear to know much about what you are commenting about. There is a huge community of research and applications around rnn and all other architectures. Pytorch has an extremely vivid and fast growing code
153.
by pilooch 9y ago
Rasp3 running openelec / kodi, never looked back. All Open Source, not snooping on you :)
154.
by pilooch 9y ago
Ads are targeted with machine learning, the only possible long term defence is fighting back with machine learning. It's already been done and both sides will still improve over the next few years. But in the end, much like spam, onli
155.
by pilooch 10y ago
because maybe some people now have theories that would become mainstream in 300 years?
156.
by pilooch 10y ago
I'm interested in knowing more about your chips. If smart sensors are one of the next steps for DL, as it seems from the high demand I get for embedded DL into cameras, etc... the existing offer by Nvidia (TK and TX) is not enough stil
157.
by pilooch 10y ago
Deepdetect for both dev and prod, with the minimum code in front of it. This setup is definitely not able to accommodate all modern ML needs, but the fast and secure model update from dev to prod is the easiest for us. Disclaimer: DD autho
158.
by pilooch 10y ago
An NVIDIA representative a week ago told me 2x but that might be a bit high, I'd take it as a cap, and maybe it is the case using tensorRT.
159.
by pilooch 10y ago
Yes, as a plugin, and on its way to make it to the trunk AFAIK. I've integrated it into deepdetect recently because even as a beta it works well and complements GPU based DNNs fairly naturally. Deep learning practitioners are already w
160.
by pilooch 10y ago
https://aws.amazon.com/marketplace/search/results?searchTerm... Some more AMIs for deep learning.
161.
by pilooch 10y ago
Try deepdetect, it s easier than serving, though probably still less flexible when it comes to the TF backend.
162.
by pilooch 10y ago
Any link to that talk, or was it internal to openai ? Thanks :)
163.
by pilooch 10y ago
I believe what general_ai means is that running models is not the issue here, it's about the FPS. NVidia has special GPUs on the TX1 and TK1 for this. Ability to run a model is about having enough memory for it. Ability to apply a mode
164.
by pilooch 10y ago
Look https://deepdetect.com for an Open Source serve that covers most functionalities for building a personal deep learning as a service and API platform. It's not perfect but does a good job at images classification, objec
165.
by pilooch 10y ago
Chromebags, they've been on my back for over ten years now, and I still own and use the three I've bought over the years. The only drawback is that they never wear out, so you have no real excuse to change!!
166.
by pilooch 10y ago
It bears some similarities with with Nvidia's tensorRT that is closed source.
167.
by pilooch 10y ago
opencl appears to be supported by TF now. Regarding FPGA, some startups such as teradeep are on it, whether it is using Caffe or TF.
168.
by pilooch 10y ago
It varies :) Mores especially it depends on your final platform and programming language. Android is fine, pure C++ without Python is more difficult though doable (eg deepdetect server has support for it), many other wrappers exist around T
169.
by pilooch 10y ago
UNK usually comes from the final sampling step: the distribution of words contains a special token UNK to represent all words with insufficient statistics in the training corpus.
170.
by pilooch 10y ago
In reality, most production code is C++ for machine learning backends.
171.
by pilooch 10y ago
tesseract 4 comes with LSTM and pre trained models.
172.
by pilooch 10y ago
I ve switched from air running Ubuntu to XPS 13 dev edition with Ubuntu. Not everything beautiful but a decent switch, great power, autonomy and screen.
173.
by pilooch 10y ago
FTR you can just build your own very easily out of any of the open source DL frameworks out there and their derivatives. E.g. digits and deepdetect will give you an API along with the pre-trained models.
174.
by pilooch 10y ago
Bernard?
175.
by pilooch 10y ago
Comparison is wrong between tfidf on words and CNN char. You should use char ngrams along with LR and this will beat all your classifiers with high probability. This is because your CNN char does not have enough data to draw all the useful
176.
by pilooch 10y ago
Very nice :) Thanks.
177.
by pilooch 10y ago
It depends whether the rented ML is key to your product / startup. Alternative is running your own mini service with API based on open source software and public models or slightly modified models. Look at digits, deepdetect, bit fusio
178.
by pilooch 10y ago
you may want a branch per mailbox so that you can merge them, fork etc
179.
by pilooch 10y ago
Beautiful! If you have pictures and one day blog a few lines about this encounter, I'd be delighted to read it!
180.
by pilooch 10y ago
For those interested in these matters, the books by Masanobu Fukuoka (1) are a must. They may challenge some of the HN crowd views on progress, technology and religion, as they did definitely challenge mine, but they carry significative sci

More ›