CUDA error 35 at /home/runner/work/ctransformers/ctransformers/models/ggml/ggml-cuda.cu:5067: CUDA driver version is insufficient for CUDA runtime version #139

Open
opened 2023-09-18 19:41:39 -04:00 by thistleknot · 4 comments
thistleknot commented 2023-09-18 19:41:39 -04:00 (Migrated from github.com)

What cuda drivers are needed?

I'm able to run llama-cpp-pyhon and llama.cpp just fine, but not this tool

Ubuntu 22
Cuda 11.7
driver is 515.43.04

What cuda drivers are needed? I'm able to run llama-cpp-pyhon and llama.cpp just fine, but not this tool Ubuntu 22 Cuda 11.7 driver is 515.43.04
thistleknot commented 2023-09-19 20:21:47 -04:00 (Migrated from github.com)

note
works in 12.2
but not 11.7/11.8 :/
too bad soo sad for my 4GB VRAM cards

note works in 12.2 but not 11.7/11.8 :/ too bad soo sad for my 4GB VRAM cards
sujeendran commented 2023-09-25 12:59:04 -04:00 (Migrated from github.com)

@thistleknot - Here is a fix that worked for me: https://github.com/marella/ctransformers/issues/90#issuecomment-1734127938

@thistleknot - Here is a fix that worked for me: https://github.com/marella/ctransformers/issues/90#issuecomment-1734127938
thistleknot commented 2023-09-25 13:09:08 -04:00 (Migrated from github.com)

Thank you

I got llama.cpp to work on one card that Max's out on compute 5.2 by
setting the compute capability explicitly in the makelist file. I presume
this is similar. However I have yet to test similar changes on this. Will
try sometime this week hopefully

On Mon, Sep 25, 2023, 9:59 AM Sujeendran Menon @.***>
wrote:

@thistleknot https://github.com/thistleknot - Here is a fix that worked
for me: #90 (comment)
https://github.com/marella/ctransformers/issues/90#issuecomment-1734127938


Reply to this email directly, view it on GitHub
https://github.com/marella/ctransformers/issues/139#issuecomment-1734136815,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ABHKKOUZU64EOXOPDJ42VTLX4GZ6HANCNFSM6AAAAAA45OK5KU
.
You are receiving this because you were mentioned.Message ID:
@.***>

Thank you I got llama.cpp to work on one card that Max's out on compute 5.2 by setting the compute capability explicitly in the makelist file. I presume this is similar. However I have yet to test similar changes on this. Will try sometime this week hopefully On Mon, Sep 25, 2023, 9:59 AM Sujeendran Menon ***@***.***> wrote: > @thistleknot <https://github.com/thistleknot> - Here is a fix that worked > for me: #90 (comment) > <https://github.com/marella/ctransformers/issues/90#issuecomment-1734127938> > > — > Reply to this email directly, view it on GitHub > <https://github.com/marella/ctransformers/issues/139#issuecomment-1734136815>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ABHKKOUZU64EOXOPDJ42VTLX4GZ6HANCNFSM6AAAAAA45OK5KU> > . > You are receiving this because you were mentioned.Message ID: > ***@***.***> >
alifatmi commented 2023-10-12 08:37:08 -04:00 (Migrated from github.com)

i also got this error but following commands worked for me.
#Install CUDA libraries using:
pip install ctransformers[cuda]
run above command and run your project code.

i also got this error but following commands worked for me. #Install CUDA libraries using: pip install ctransformers[cuda] run above command and run your project code.
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
marella/ctransformers#139
No description provided.