LECTURE: An Introduction to JAX (#237)
* Added jax lecture
* disable build cache, enable execution reports upon failure
* enable upload of build failure reports for all outputs
* catch error using raises-exception tag
* Fix missing variables for solution
* enable html build only
* fix comment typo
* Add jax to environment
* add jaxlib and jax[cuda]
* Update lectures/jax_intro.md
Co-authored-by: Smit Lunagariya <smitlunagariya.mat18@itbhu.ac.in>
* Install jax outside of environment creation
* update bash and remove pip upgrade
* move to install jax from conda-forge
* Remove sandpit packages and revert to pip
* Allow 4 minute timeout for cell execution
* enable 10min timeout
* Use jax install from Docs
* Install nvidia drivers to check if driver issue
* use nvidia cuda docker container
* remove nvidia driver install step
* disable latex support for now
* try install cudatoolkit
* give container access to gpu
* remove locate and keep whereis
* Upgrade CUDA to latest
* upgrade cuda before installing jax
* Max HDD bigger for new cuda drivers
* use latest nvidia docker container release using 11.8
* match cuda toolkit from conda
* ensure cuda installed as reported not available
* enable nvidia-smi
* add diagnostics
* add nvidia runtime to docker launch
* enable nvidia-smi on jb run
* revert to cml base image
* try installing within the container
* Test quantecon local runner
* simply and see if can run on local conda env
* fix syntax for conda
* run containerised version
* enable full preview build including pdf, download notebook + build cache
* remove install of latex as installed locally on runner
* add explanations for block_until_ready()
* change wording
* improve based on comments
* switch back to using ec2
* remove cache
* make sure nvcc binaries are available
* check solution from jax issues
* install cuda and cuda-toolkit
* try using python in build to container
* add python packages for lectures
* use nvidia as base docker
* move back to conda
* rely on nvidia docker for drivers
* ensure jax is uninstalled
* try nvidia cuda=11.2
* Check tensorflow docker container
* check tensorflow==2.9.2
* change hardware to V100
* remove tensorflow, use nvidia docker (quicker), check opt_savings
* remove tensorflow docker, update title for test of opt_savings
* tidy up, enable pdf and download nb builds
* remove sudo
* setup timezone data
* remove ENV
* remove infrastructure testing and move to new PR
* remove test file from toc
* check conda environment
* more conda path debug
* remove debu
* add hardware details
* add more useful information in lecture
* improve note in lecture
Co-authored-by: mmcky <mmcky@users.noreply.github.com>
Co-authored-by: mmcky <mamckay@gmail.com>
Co-authored-by: Smit Lunagariya <smitlunagariya.mat18@itbhu.ac.in>
Co-authored-by: Humphrey Yang <u6474961@anu.edu.au>
4 files changed
Lines changed: 629 additions & 4 deletions
