briggers

8 ms·

1.
PyTorch Memory Tuning(paulbridger.com)
1 points by briggers 3y ago|0 comments
2.
by briggers 3y ago
Machine learning productionization: https://paulbridger.com
3.
PyTorch Performance Features and How They Interact(paulbridger.com)
3 points by briggers 3y ago|0 comments
4.
by briggers 4y ago
Absolutely yes. I write fairly deep ML performance tuning articles at https://paulbridger.com and the (many) hours I've spent on each article have been hugely worth it. Many people reach out to me via this work, and when we
5.
by briggers 4y ago
Before you start optimizing runtime performance: measure, trace, inspect, or whatever is appropriate to understand current performance.
6.
by briggers 5y ago
Perhaps you are looking for something like this? docker run -d -p 5000:5000 --name registry registry:2 https://docs.docker.com/registry/#:~:text=The%20Registry%20i... .
7.
Wombo Dream: prompt-guided generative art(app.wombo.art)
2 points by briggers 5y ago|0 comments
8.
Solving Machine Learning Performance Anti-Patterns: A Systematic Approach(paulbridger.com)
54 points by briggers 5y ago|1 comments
9.
Solving Machine Learning Performance Anti-Patterns: A Systematic Approach(paulbridger.com)
10 points by briggers 5y ago|2 comments
10.
by briggers 6y ago
I used to love doing this with Clojure, it’s an awesome way to increase productivity by a good chunk. It’s less about saving the time to re-run something, and more about removing conceptual overhead (I think).
11.
Analyzing the performance of Tensorflow training on M1 Mac Mini and Nvidia V100(wandb.ai)
226 points by briggers 6y ago|91 comments
12.
Object Detection at 2530 FPS with TensorRT and 8-Bit Quantization(paulbridger.com)
9 points by briggers 6y ago|0 comments
13.
Optimizing models using the PyTorch JIT(lernapparat.de)
1 points by briggers 6y ago|0 comments
14.
by briggers 6y ago
Yes! I do this and love it. Works locally or via SSH with no difference at all.
15.
Mastering TorchScript: Tracing vs. Scripting, Device Pinning, Graph Modification(paulbridger.com)
7 points by briggers 6y ago|0 comments
16.
Mastering TorchScript: Tracing vs. Scripting, Device Pinning, Graph Modification(paulbridger.com)
7 points by briggers 6y ago|0 comments
17.
by briggers 6y ago
Nice one! I've long been interested in the ONNX serving path.
18.
by briggers 6y ago
Yeah. A 2080Ti doesn't fit in your pocket or in your AR glasses but the same techniques and tools scale down.
19.
by briggers 6y ago
BTW, this is pumping the same video file through the network - not just a single file. I don't measure latency, but this is not a deep pipeline so it's easy to calculate.
20.
by briggers 6y ago
Mobile phones definitely since these days most of them have pretty powerful GPUs.
21.
by briggers 6y ago
Great question, now I wish I'd recorded power consumption for all these experiments. Judging from cumulative hours of watching the output of nvidia-smi I've definitely seen a linearish relationship between utilization and power dr
22.
by briggers 6y ago
Very practical question :) Exactly as you say, multi-stream throughput. Also for faster than realtime offline processing of video. Check the caveats section at the end of the post - DeepStream is probably not well suited to high throughput
23.
Object Detection at 1840 FPS with TorchScript, TensorRT and DeepStream(paulbridger.com)
162 points by briggers 6y ago|55 comments
24.
by briggers 6y ago
I looked into the GIL saturation as measured with gil_load ( https://github.com/chrisjbillington/gil_load ), but perhaps I should have included more numbers here. To me, seeing the GIL held for 40% of time and significan
25.
by briggers 6y ago
Great point - dependencies between frames are inherently problematic for many of these techniques. Everything lostdog says. I've had experience speeding up tracking immensely using the same big hammer I talk about in the article - movi
26.
by briggers 6y ago
Author here. As other commenters are saying, the Pytorch JIT and torchscript might be your friend here. Alternatively, there are some quite fast OSS libraries for object detection. Nvidia's retinanet will export to a TensorRT engine wh
27.
by briggers 6y ago
Author here. I really appreciate your feedback. Completely agree that almost none of the SoTA github repos are really ready for production and making this stuff work can be pretty hard. Getting this done on C++ and moving up to the next lev
28.
Object Detection from 9 FPS to 650 FPS(paulbridger.com)
165 points by briggers 6y ago|38 comments
29.
by briggers 6y ago
Nice one. More useful to me would be something similar that operates on tensors on the GPU. Doing image annotations on host/CPU often becomes a bottleneck.
30.
by briggers 6y ago
It’s important to note this is just within visual range dogfighting/BFM. Also just guns, and also perfect enemy state information. Important progress but it’s super early. Beyond visual range/BVR, coordination with wingmen and man

More ›