Powerful generative AI models and cloud-native APIs and microservices are coming to the edge. Generative AI is bringing the power of transformer models and large language models to virtually every industry. That reach now includes areas that touch...
Est. reading time: 7 minutes
Yesterday at ROSCon 2023, we unveiled new announcements and advances for JetPack 6.0 and beyond:
We invite you to attend two upcoming webinars on the above topics:
1 Like
We published a walkthrough of the Jetson Generative AI Lab here!
1 Like
this is amazing: * bring-your-own kernel and Linux distro in JetPack 6
I hope the jetson I wish the jetson could get to use the NVIDIA Linux open GPU kernel module source. GitHub - NVIDIA/open-gpu-kernel-modules: NVIDIA Linux open GPU kernel module source
Do you already have a fixed date when Jetpack6 or beta of jetpack6 launches for jetson orin AGX devkit.
It really sounds amazing.
I can not wait for ROS2 humble native not in Docker anymore.
1 Like
@yalttlay JetPack 6 is supposed to come out later this month
1 Like
Thanks for the fast answer.
Do you know if it’s mid November or later.
We have a study project and it would be nice to start directly with jetpack6.
@yalttlay it will be late November, like 11/30
are we there yet?!! ;) roll on the 30th of november!!!
do you know if jetpack 6 will support cuda 12.1? I’m really struggling to get the orin running cuda 12.1 and guessing it’s not possible
@f.mainstone JetPack 6.0 has CUDA 12.2
3 Likes
@johnnynuca14 I have to check with the engineering teams on the status since the holiday, but baring any blockers/issues/bugs to resolved, we are tentatively targeting 11/30 for the release.
1 Like
brbl
November 30, 2023, 7:55pm
14
So make an old man happy and tell me that pytorch 2.1 on JetPack 6 will support distributed…
Thanks ( in either case),
bb
PyTorch already can support distributed on JetPack 4 and JetPack 5, it just needs to be built when you have MPI installed or with USE_DISTRIBUTED=1. Which this dockerfile for building PyTorch wheels does by default:
# sed "s|'8.6', '8.9'|'8.6', '8.7', '8.9'|" -i ${CPP_EXTENSION_PY_FILE} && \
# sed -n 1729,1746p ${CPP_EXTENSION_PY_FILE}
# Build PyTorch wheel with extra environmental variables for custom feature switch
RUN git clone --branch v${PYTORCH_BUILD_VERSION} --depth=1 --recursive https://github.com/pytorch/pytorch /tmp/pytorch && \
cd /tmp/pytorch && \
export USE_NCCL=0 && \
export USE_QNNPACK=0 && \
export USE_PYTORCH_QNNPACK=0 && \
export USE_NATIVE_ARCH=1 && \
export USE_DISTRIBUTED=1 && \
export USE_TENSORRT=0 && \
export ${PYTORCH_BUILD_EXTRA_ENV} && \
pip3 install -r requirements.txt && \
pip3 install --no-cache-dir scikit-build ninja && \
python3 setup.py bdist_wheel && \
cp dist/*.whl /opt && \
rm -rf /tmp/pytorch
# install the compiled wheel
RUN pip3 install --verbose /opt/torch*.whl
brbl
December 1, 2023, 12:39pm
16
Having an updated example of building pytorch in one place is awesome. Thanks, as always, Dusty!
Jetpack 6 today? I can’t wait for this awesome update.
1 Like
I hope so I’m also waiting
2 Likes
Also excited for Jetpack 6 release today
1 Like