mids-scaling-up / v2 Goto Github PK
View Code? Open in Web Editor NEWW251 2018 reload
W251 2018 reload
Need to talk about the correct formatting of the external drive used by docker. For example, windows formatted drives may mount, but can't be used correctly.
I'd like to move HW5 to use Tensorflow 2 vs what we have now.
Thoughts?
In order to mount the s3fs storage in a docker container, the container must have had the --privileged flag upon creation, otherwise it will throws an error: fuse device not found, try 'modprobe fuse' first
The original docker file has a filepath to build the nvidia tensorflow docker container. This is a broken link. You should point directly to the .whl.
RUN pip3 install https://developer.download.nvidia.com/compute/redist/jp/v42/tensorflow-gpu/tensorflow_gpu-1.13.1+nv19.5-cp36-cp36m-linux_aarch64.whl
wondering if we should add a comment that if running in IBM Cloud, that it is best to use the COS private endpoints as they avoid going over the pubic network.
Line 19 in 9b96fac
It seems to be a typo and breaks docker. It works when I remove this line.
New CUDA drivers are required to work with the nvidia-smi at the bottom of HW3, so the Install CUDA section can be updated for now to v10.1:
wget https://developer.nvidia.com/compute/cuda/10.1/Prod/local_installers/cuda-repo-ubuntu1604-10-1-local-10.1.105-418.39_1.0-1_amd64.deb
sudo dpkg -i cuda-repo-ubuntu1604-10-1-local-10.1.105-418.39_1.0-1_amd64.deb
sudo apt-key add /var/cuda-repo-10-1-local-10.1.105-418.39/7fa2af80.pub
sudo apt-get update
sudo apt-get install cuda
Is the following a typo and a possible fix for it?
Possible Typo
Name all the layers in parameters in the network
Possible Fix
Name all the layers and parameters in the network
Hi guys, How is it going?
I am just wondering if you could share an example on how to send GPS data from Jetson with MQQT.
Thank you very much!
(I could not find your container in GitHub, so I am posting here)
Thank you for this container image. I was able to start the DIGITS server on my NVIDIA Jetson Nano. Image load works fine, but attempting to create an Image Classification model leads to these errors. Looks like a CUDA issue. TIA.
2019-05-25 20:36:31 [20190525-203625-5f68] [INFO ] Create DB (train) task started.
2019-05-25 20:36:31 [20190525-203625-5f68] [INFO ] Task subprocess args: "/usr/bin/python2 /DIGITS/digits/tools/create_db.py /DIGITS/digits/jobs/20190525-203625-5f68/train.txt /DIGITS/digits/jobs/20190525-203625-5f68/train_db 768 1272 --backend=lmdb --channels=3 --resize_mode=crop --mean_file=/DIGITS/digits/jobs/20190525-203625-5f68/mean.binaryproto --mean_file=/DIGITS/digits/jobs/20190525-203625-5f68/mean.jpg --shuffle --encoding=jpg"
2019-05-25 20:36:31 [20190525-203625-5f68] [INFO ] Create DB (val) task started.
2019-05-25 20:36:31 [20190525-203625-5f68] [INFO ] Task subprocess args: "/usr/bin/python2 /DIGITS/digits/tools/create_db.py /DIGITS/digits/jobs/20190525-203625-5f68/val.txt /DIGITS/digits/jobs/20190525-203625-5f68/val_db 768 1272 --backend=lmdb --channels=3 --resize_mode=crop --shuffle --encoding=jpg"
2019-05-25 20:36:34 [20190525-203625-5f68] [WARNING] Create DB (train) unrecognized output: cudaRuntimeGetVersion() failed with error #38
2019-05-25 20:36:34 [20190525-203625-5f68] [WARNING] Create DB (train) unrecognized output: Tensorflow support disabled.
2019-05-25 20:36:34 [20190525-203625-5f68] [WARNING] Create DB (val) unrecognized output: cudaRuntimeGetVersion() failed with error #38
2019-05-25 20:36:34 [20190525-203625-5f68] [WARNING] Create DB (val) unrecognized output: Tensorflow support disabled.
2019-05-25 20:36:39 [20190525-203625-5f68] [DEBUG] 81 images written to database
2019-05-25 20:36:39 [20190525-203625-5f68] [INFO ] Create DB (val) task completed.
2019-05-25 20:36:46 [20190525-203625-5f68] [DEBUG] 246 images written to database
2019-05-25 20:37:06 [20190525-203625-5f68] [INFO ] Create DB (train) task completed.
2019-05-25 20:37:06 [20190525-203625-5f68] [INFO ] Job complete.
2019-05-25 20:38:57 [20190525-203823-f70d] [DEBUG] Network sanity check - train
2019-05-25 20:38:57 [20190525-203823-f70d] [DEBUG] Network sanity check - val
2019-05-25 20:38:57 [20190525-203823-f70d] [DEBUG] Network sanity check - deploy
2019-05-25 20:38:57 [20190525-203823-f70d] [INFO ] Train Caffe Model task started.
2019-05-25 20:38:57 [20190525-203823-f70d] [INFO ] Task subprocess args: "/caffe/build/tools/caffe train --solver=/DIGITS/digits/jobs/20190525-203823-f70d/solver.prototxt"
2019-05-25 20:38:58 [20190525-203823-f70d] [ERROR] Train Caffe Model: Cannot create Cublas handle. Cublas won't be available.
2019-05-25 20:38:58 [20190525-203823-f70d] [ERROR] Train Caffe Model: Cannot create Curand generator. Curand won't be available.
2019-05-25 20:38:58 [20190525-203823-f70d] [ERROR] Train Caffe Model: Cannot create cuDNN handle. cuDNN won't be available.
2019-05-25 20:38:58 [20190525-203823-f70d] [ERROR] Train Caffe Model: Check failed: error == cudaSuccess (38 vs. 0) no CUDA-capable device is detected
2019-05-25 20:38:59 [20190525-203823-f70d] [ERROR] Train Caffe Model task failed with error code -6
The link to TorchVision is now: https://pytorch.org/vision/stable/
Specifically commit:
dff675c#diff-37a0f51b49e944d6095aab23fd89f5ed
Commenting out lines for data + bert_config.json files.
In week02/README.md the following links no longer point to the right place:
http://www.ibm.com/cloud-computing/us/en/what-is-cloud-computing.html
http://www.linuxplanet.com/linuxplanet/reports/6503/1
I think the first one should now be: https://www.ibm.com/cloud/learn/cloud-computing
The second one seems to have been completely removed as the link to "complete story" from here is also broken: https://www.linuxtoday.com/it_management/2008072101835OSSW
There are unused imports in the example Dockerfile in week02/labs/README.md:
# FROM nvidia/cuda
# FROM nvidia/cuda:8.0-cudnn6-devel
# FROM nvidia/cuda:8.0-cudnn5-devel
--snip--
# CMD jupyter notebook --no-browser --ip=0.0.0.0 --NotebookApp.token= --allow-root
These are unexplained and seem to be unused and can be confusing for students.
I suggest to remove the commented-out lines, or explain their raison d'être.
The version of TF in the container is GPU-specific; It dumps core when running without --runtime=nvidia
In HW1 Virtual Box was really had to work with, VMware was much better and most of us eventually used VMware. It will be helpful if the instructions were updated to use VMware instead for future student. It was much faster as well.
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.