You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
* Test with GPU and m–t4 machines
* Restyled by whitespace
* Use `Standard_NC4as_T4_v3` on AKS
* Bump default disk size on iterative_task resource
* Ditto for tests
* Fix `az` NGC machine image version
* Migrate Google Cloud GPU images to official
* Remove redundant `grep`
* Bump default disk to 50 GB
* Ditto
* Try Azure DSVM GPU images
* Try which providers support unset disk size
* Restyled by gofmt
* Make disk size optional
* Whoops!
* Restyled by gofmt
* Use disk_size > 0 everywhere
* Fix `gcp` derp
* Fix GCP GPU machines
* Use `yes`because... why not?
* Indent back script
* Simplify test script error handling
* Remove redundant `Storage` requirement
* Avoid `mkdir` error if directory exists
* Improve test verbosity & fail fast
* Upgrade AWS DLAMI to CUDA 11.3
* Keep it simple
* Test `m+k80` to see if `k8s` breaks
* Fix `k8s` storage size
* Restyled by gofmt
* Ahem, ahem
* Use `t4` again
* Remove `k8s` granular GPU selectors
* Fix last `k8s` issues 🤞
* Revert cluster instance change
* Delete linux_amd64
* Avoid mkdir errors
* Update task/k8s/resources/resource_job.go
Co-authored-by: Casper da Costa-Luis <casper.dcl@physics.org>
* docs: auto-disk_size
* docs: nvidia images
* nvidia descrtiption
Co-authored-by: Restyled.io <commits@restyled.io>
Co-authored-by: Casper da Costa-Luis <casper.dcl@physics.org>
-`region` - (Optional) [Cloud region/zone](#cloud-region) to run the task on.
58
58
-`machine` - (Optional) See [Machine Types](#machine-type) below.
59
-
-`disk_size` - (Optional) Size of the ephemeral machine storage in GB.
59
+
-`disk_size` - (Optional) Size of the ephemeral machine storage in GB.`-1`: automatic based on `image`.
60
60
-`spot` - (Optional) Spot instance price. `-1`: disabled, `0`: automatic price, any other positive number: maximum bidding price in USD per hour (above which the instance is terminated until the price drops).
61
61
-`image` - (Optional) [Machine image](#machine-image) to run the task with.
62
62
-`parallelism` - (Optional) Number of machines to be launched in parallel.
@@ -169,7 +169,7 @@ In addition to generic types, it's possible to specify any machine type supporte
169
169
The Iterative Provider offers some common machine images which are roughly the same for all supported clouds.
170
170
171
171
-`ubuntu` - Official [Ubuntu LTS](https://wiki.ubuntu.com/LTS) image (currently 20.04).
172
-
-`nvidia` - Official [NVIDIA NGC](https://docs.nvidia.com/ngc/ngc-deploy-public-cloud)-based images, typically needing `disk_size = 32` GB or more.
172
+
-`nvidia` - Official Ubuntu LTS with NVIDIA GPU drivers and CUDA toolkit (currently 11.3).
0 commit comments