M3 not working with torch

On 10 Jun 2024, at 4:47?AM, Phil McShane via R-SIG-Mac <r-sig-mac at r-project.org> wrote:

Dear Colleague			Sunday

My husband died over a year ago.  Although I still use his e-mail address I would be grateful if you would remove this e-mail address from your mailing

Regards			Joan McShane (Widow)

On 20 May 2024, at 21:48, Simon Urbanek <simon.urbanek at r-project.org> wrote:

Gilberto,

since luz is a contributed package, you should probably start first by asking the authors. Given that the torch ecosystem is quite complex and has several layers that need to work together, even if you talk to them, you probably need to add details such as exact versions used (including the torch and metal layers) and how you installed the pieces (I know you helpfully supplied sesisonInfo() but I suspect that info such as exact torch run-time is pertinent as well). Next step would be to trace the error - check the system crash reporter or run R -d lldb to find out the exact library the crash happens in which may give you more clues. I don't have any M3 machines so I can't check myself, unfortunately.

Cheers,
Simon

On 21/05/2024, at 12:48 AM, Gilberto Camara <gilberto.camara at inpe.br> wrote:

Dear R-SIG-MAC

I bought a new MacBook Air with the M3 chip, which has 8 CPUs, 10 GPUs, and 16GB of integrated memory. My R `torch` apps are crashing. I have assembled an MWE that works on other Mac architectures, including MacBook Air M1 and MacMini. The OS is the same (Sonoma 14.5). The MWE follows:
[...]
```
The error occurs in the `luz::fit` function. Inside RStudio, the code gets stuck, and then RStudio asks to restart R. When running R from the terminal, the output is:
```{r}
*** caught bus error ***
address 0x16daa0000, cause 'invalid alignment'

*** caught segfault ***
address 0x9, cause 'invalid permissions'
zsh: segmentation fault  R
```
The `sessionInfo()` output is as follows:

	[[alternative HTML version deleted]]

On 10 Jun 2024, at 4:47?AM, Phil McShane via R-SIG-Mac <r-sig-mac at r-project.org> wrote:

Dear Colleague			Sunday

My husband died over a year ago.  Although I still use his e-mail address I would be grateful if you would remove this e-mail address from your mailing

Regards			Joan McShane (Widow)

On 20 May 2024, at 21:48, Simon Urbanek <simon.urbanek at r-project.org> wrote:

Gilberto,

since luz is a contributed package, you should probably start first by asking the authors. Given that the torch ecosystem is quite complex and has several layers that need to work together, even if you talk to them, you probably need to add details such as exact versions used (including the torch and metal layers) and how you installed the pieces (I know you helpfully supplied sesisonInfo() but I suspect that info such as exact torch run-time is pertinent as well). Next step would be to trace the error - check the system crash reporter or run R -d lldb to find out the exact library the crash happens in which may give you more clues. I don't have any M3 machines so I can't check myself, unfortunately.

Cheers,
Simon

On 21/05/2024, at 12:48 AM, Gilberto Camara <gilberto.camara at inpe.br> wrote:

Dear R-SIG-MAC

I bought a new MacBook Air with the M3 chip, which has 8 CPUs, 10 GPUs, and 16GB of integrated memory. My R `torch` apps are crashing. I have assembled an MWE that works on other Mac architectures, including MacBook Air M1 and MacMini. The OS is the same (Sonoma 14.5). The MWE follows:
[...]
```
The error occurs in the `luz::fit` function. Inside RStudio, the code gets stuck, and then RStudio asks to restart R. When running R from the terminal, the output is:
```{r}
*** caught bus error ***
address 0x16daa0000, cause 'invalid alignment'

*** caught segfault ***
address 0x9, cause 'invalid permissions'
zsh: segmentation fault  R
```
The `sessionInfo()` output is as follows:

	[[alternative HTML version deleted]]

On 10 Jun 2024, at 09:37 , Ken Beath <ken at kjbeath.id.au> wrote:

I?m sorry to hear about your husband.

To unsubscribe  you need to access https://stat.ethz.ch/mailman/listinfo/r-sig-mac 

The password is contained in an e-mail which is sent regularly with the subject r-project.org mailing list memberships reminder <http://r-project.org/> or you can request the password on the webpage.

Ken

On 10 Jun 2024, at 4:47?AM, Phil McShane via R-SIG-Mac <r-sig-mac at r-project.org> wrote:

Dear Colleague			Sunday

My husband died over a year ago.  Although I still use his e-mail address I would be grateful if you would remove this e-mail address from your mailing

Regards			Joan McShane (Widow)

On 20 May 2024, at 21:48, Simon Urbanek <simon.urbanek at r-project.org> wrote:

Gilberto,

since luz is a contributed package, you should probably start first by asking the authors. Given that the torch ecosystem is quite complex and has several layers that need to work together, even if you talk to them, you probably need to add details such as exact versions used (including the torch and metal layers) and how you installed the pieces (I know you helpfully supplied sesisonInfo() but I suspect that info such as exact torch run-time is pertinent as well). Next step would be to trace the error - check the system crash reporter or run R -d lldb to find out the exact library the crash happens in which may give you more clues. I don't have any M3 machines so I can't check myself, unfortunately.

Cheers,
Simon

On 21/05/2024, at 12:48 AM, Gilberto Camara <gilberto.camara at inpe.br> wrote:

Dear R-SIG-MAC

I bought a new MacBook Air with the M3 chip, which has 8 CPUs, 10 GPUs, and 16GB of integrated memory. My R `torch` apps are crashing. I have assembled an MWE that works on other Mac architectures, including MacBook Air M1 and MacMini. The OS is the same (Sonoma 14.5). The MWE follows:
[...]
```
The error occurs in the `luz::fit` function. Inside RStudio, the code gets stuck, and then RStudio asks to restart R. When running R from the terminal, the output is:
```{r}
*** caught bus error ***
address 0x16daa0000, cause 'invalid alignment'

*** caught segfault ***
address 0x9, cause 'invalid permissions'
zsh: segmentation fault  R
```
The `sessionInfo()` output is as follows:

	[[alternative HTML version deleted]]

	[[alternative HTML version deleted]]