We are using Buildroot and parts of JetPack 6.1 is installed inside and trying to run the base image in NVIDIA L4T PyTorch | NVIDIA NGC,
Thank you so much!!
Hi,
Jetson’s container is now available in the below link, together with the dGPU image:
Please use the image with igpu tag for the Jetson.
Thanks.
Thanks for replying!
So in comparison to the NVIDIA L4T PyTorch | NVIDIA NGC, the PyTorch | NVIDIA NGC does not need Jetpack at all?
And I therefore only need to use this image as base instead of the L4T one? Thanks!
Hi,
The required dependencies are all included in the container for both location.
L4T location is not updated anymore.
If you need to use it on JetPack 6, please try the PyTorch location with iGPU tag instead.
Thanks.
Hi!
I have been trying to build the base image for my Docker image and have been failing because it cannot find the manifest. Do you have a documentation which repositories should be registered in my docker or the requirements to be able to use the base image properly?
[internal] load metadata for nvcr.io/nvidia/pytorch:25.03-py3-igpu:
ERROR: failed to solve: nvcr.io/nvidia/pytorch:25.03-py3-igpu: failed to resolve source metadata for nvcr.io/nvidia/pytorch:25.03-py3-igpu: no match for platform in manifest: not found
in comparison to that, the manifest of nvcr.io/nvidia/pytorch:25.03-py3 seems to be available
Hi,
Just test it and it works well in our environment:
Dockerfile
FROM nvcr.io/nvidia/pytorch:25.03-py3-igpu
Test
$ docker build .
[+] Building 208.4s (6/6) FINISHED docker:default
=> [internal] load build definition from Dockerfile 0.1s
=> => transferring dockerfile: 81B 0.0s
=> [internal] load metadata for nvcr.io/nvidia/pytorch:25.03-py3-igpu 4.7s
=> [auth] nvidia/pytorch:pull token for nvcr.io 0.0s
=> [internal] load .dockerignore 0.1s
=> => transferring context: 2B 0.0s
=> [1/1] FROM nvcr.io/nvidia/pytorch:25.03-py3-igpu@sha256:f398a0955ec5fcf9e3bbf77610225ff4e953e137423ab248e2bf32cd4971a1dc 202.7s
=> => resolve nvcr.io/nvidia/pytorch:25.03-py3-igpu@sha256:f398a0955ec5fcf9e3bbf77610225ff4e953e137423ab248e2bf32cd4971a1dc 0.0s
=> => sha256:3318952e176bb82db993dec95624f35e505a8abd380e30c1b7f895e38947d658 114.78MB / 114.78MB 8.4s
...
=> => writing image sha256:6b22318cec7e872d35c09116dea4ba8921d39387a3adbada10e14933f919b07d 0.0s
Thanks.
I just checked again, it still does not work, do I need to do something beforehand?
nvcr.io/nvidia/pytorch:25.03-py3-igpu: failed to resolve source metadata for nvcr.io/nvidia/pytorch:25.03-py3-igpu: no match for platform in manifest: not found
-igpu
ngc registry image info nvcr.io/nvidia/pytorch:25.03-py3-igpu
Error: Image ‘nvcr.io/nvidia/pytorch:25.03-py3-igpu’ could not be found.
If I drop the -igpu that image is found.
ngc registry image info nvcr.io/nvidia/pytorch:25.03-py3
Image Information
Name: nvidia/pytorch:25.03-py3
Architecture: amd64
Image Size: 10.88 GB
Digest: sha256:54fd644e268346657619a987596e522e813d7abaaf92faafbe815b2b43cff16d
Architecture: arm64
Image Size: 10.31 GB
Digest: sha256:fd85ecca9dc831abce5fec544bd861509fc900c4863562fc998bb93da52ccc3e
Schema Version: 1
Signed?: True
Last Updated: Mar 25, 2025
exactly, I have that problem
Hi, both
Are you able to pull the image?
$ sudo docker pull nvcr.io/nvidia/pytorch:25.03-py3-igpu
25.03-py3-igpu: Pulling from nvidia/pytorch
5b17151e9710: Pulling fs layer
6055d2b1b566: Pulling fs layer
3318952e176b: Pulling fs layer
5cc84ac41104: Waiting
38a09ab6290f: Waiting
...
Thanks.
I cannot pull it,
docker pull nvcr.io/nvidia/pytorch:25.03-py3-igpu
Error response from daemon: no matching manifest for linux/amd64 in the manifest list entries: no match for platform in manifest: not found
ngc registry image info --debug nvcr.io/nvidia/pytorch:25.03-py3-igpu
Requesting URL (GET): https://api.ngc.nvidia.com/v2/org/nvidia/repos/pytorch/images/25.03-py3-igpu
payload: None
params: None
Retrieving API Key Token from cache…
Headers:
{‘Content-Type’: ‘application/json’, ‘User-Agent’: ‘ngccli/3.64.2 linux-aarch64’, ‘Authorization’: '***************************************************************************************************** ‘nv-ngc-context’: ‘{“orgName”: “bai5suy3beon”}’}
Response status: 404 - Reason: Not Found
Retries remaining: 4
Response is: {“requestStatus”:{“statusCode”:“NOT_FOUND”,“statusDescription”:"404 NOT_FOUND,
*ProblemDetail[type=‘urn:kaizen:problem-details:not-found’, title=‘Not Found’, status=404, *
detail=‘Manifest not found for nvidia/pytorch:25.03-py3-igpu not found’, instance=‘null’, properties=‘null’]",“requestId”:“ec4ba112-f679-43b2-8d8a-d8485c44c449”}}
Error: Image ‘nvcr.io/nvidia/pytorch:25.03-py3-igpu’ could not be found.
Error: Image ‘nvcr.io/nvidia/pytorch:25.03-py3-igpu’ could not be found.
Traceback (most recent call last):
File “ngcbase/command/basecli.py”, line 158, in run
File “ngcbase/tracing.py”, line 229, in wrapper_span
File “ngcbase/command/clicommand.py”, line 247, in traced_func
File “ngcbase/command/clicommand.py”, line 304, in f_mutex
File “registry/command/image.py”, line 231, in info
File “ngcbase/util/utils.py”, line 538, in wrapper
File “registry/api/image.py”, line 196, in info
ngcbase.errors.ResourceNotFoundException: Error: Image ‘nvcr.io/nvidia/pytorch:25.03-py3-igpu’ could not be found.
Command: ngc registry image info --debug nvcr.io/nvidia/pytorch:25.03-py3-igpu
os: Linux-5.15.148-tegra
ngc-cli version: NGC CLI 3.64.2
Hi,
Error response from daemon: no matching manifest for linux/amd64 in the manifest list entries: no match for platform in manifest: not found
Do you try to pull the container on the desktop?
The container is for Jetson so please pull it on the AGX Orin directly.
It can work for the system of LINUX/ARM64.
Thanks.
On agx orin dev kit.
Now it has been fixed. and am pulling it now. Thank you.
It works for me in Linux as well. Thank you so much!