Closed eddelbuettel closed 9 months ago
Thanks for reporting, I have updated the docs. Unfortunatelly it will only appear in the website once we make a new release, but it can be already found in the dev version of the docs:
https://torch.mlverse.org/docs/dev/articles/installation#pre-built
I am having the same problem now with 0.12.0 following the instructions in the current vignette:
> version <- available.packages()["torch","Version"]
> options(repos = c(
torch = sprintf("https://storage.googleapis.com/torch-lantern-builds/packages/%s/%s/", kind, version),
CRAN = "https://cloud.r-project.org" # or any other from which you want to install the other R dependencies.
))
> install.packages("torch")
Installing package into ‘/usr/local/lib/R/site-library’
(as ‘lib’ is unspecified)
Warning: unable to access index for repository https://storage.googleapis.com/torch-lantern-builds/packages/cu117/0.12.0/src/contrib:
cannot open URL 'https://storage.googleapis.com/torch-lantern-builds/packages/cu117/0.12.0/src/contrib/PACKAGES'
Hah! And like before 'guessing' that cu118
may work seems to help:
> kind <- "cu118"
> version <- "0.12.0"
> options(repos = c(
torch = sprintf("https://storage.googleapis.com/torch-lantern-builds/packages/%s/%s/", kind, version),
CRAN = "https://cloud.r-project.org" # or any other from which you want to install the other R dependencies.
))
> available.packages()["torch","Version"]
[1] "0.12.0"
>
> install.packages("torch")
Installing package into ‘/usr/local/lib/R/site-library’
(as ‘lib’ is unspecified)
trying URL 'https://storage.googleapis.com/torch-lantern-builds/packages/cu118/0.12.0/src/contrib/torch_0.12.0_R_x86_64-pc-linux-gnu.tar.gz'
downloaded 2749.2 MB
[...]
I just found out that using cu118
fixed my install issues. It's worth noting that the Install Vignette on CRAN is outdated and thus, not working
Hah! And like before 'guessing' that
cu118
may work seems to help:
The instructions still show 0.10.0, but there appears to be no repo behind it. Whereas 'guessing' that 0.11.0 may work was successful. I would be lovely to make it more obvious one now needs the torch package version -- maybe the example can even farm out and use the version at CRAN to keep the docs in sync 'automagically' ?