Excited to share our work on automating crevasse detection on continental glaciers at the SAGES ASM at Heriot-Watt University last week.
#computervision #deeplearning #semanticsegmentation #glaciology #climatetech #SAGES #HeriotWatt
28
Steven Wallace retweeted
I am happy to share the work of our team. The outcome of a collaborative effort, by a joyful group of skilled and determined scientists and engineers! Congrats to the team on this amazing milestone!
Introducing DINOv3: a state-of-the-art computer vision model trained with self-supervised learning (SSL) that produces powerful, high-resolution image features. For the first time, a single frozen vision backbone outperforms specialized solutions on multiple long-standing dense prediction tasks.
Learn more about DINOv3 here: ai.meta.com/blog/dinov3-self…
Thank you to the University of Aberdeen for inviting me to conduct a guest lecture during the Data Mining with Deep Learning module on the MSc AI yesterday. The lecture covered segmenting crevasses on continental glaciers in the Arctic Circle. #AI #segmentation
4
159
Steven Wallace retweeted
The proceedings for NLDL 2025 is now available at PMLR: proceedings.mlr.press/v265/ 🤖❄️
Thanks to NLDL 2025 for inviting me to present our work with SAM and SAM 2 published in Proceedings of Machine Learning Research. #nldl #ai #machinelearning #deeplearning
Steven Wallace retweeted
NLDL 2025 Day 1 Highlights: Exciting research showcases, discussions on diversity in AI, and Northern Lights! 🤖 🧠 🌍
A huge thanks to Day 1's invited keynote speakers, scientific presenters, and "Diversity in AI" speakers and panelists for a great first day here in Tromsø!
It was great yesterday to be given the chance to present our conference paper publication at the Nothern Lights Deep Learning Conference 2025 in Tromsø, Norway. #deeplearning #AI #nldl2025
1
117
It's time for the final poster presentation of 2024 at PhD Connect, Horizon Leeds. #AI #DeepLearning #MachineLearning #segmentation
1
1
72
There are some very interesting insights in this I have to say. #deeplearning #MachineLearning #LLMs #ai
I gave a talk at GPU MODE workshop last week on llm.c
- the origin story of llm.c
- being naked in the world without PyTorch and having to re-invent Array, Autograd, Device, Dtype, Compile, Distributed
- how to port a PyTorch layer to 1) explicit PyTorch
- and then to 2) write the backward pass
- 3) port forward & backward pass to C
- 4) string all the layers together
- achieving one file of C with no dependencies that compiles and runs ~instantly, where all memory is pre-planned and allocated a single time, fully deterministic, portable code that can run on a potato or a von Neumann probe
- how most of llm.c was built at 1am-7am in a water villa porch in Maldives and why this is the recommended way to develop software
- convert all of it to run in CUDA on GPU in fp32
- port matmul to cuBLAS
- port attention to cuDNN flash-attention
- introduce bfloat16 mixed precision
- introduce many more optimizations and features like kernel fusions, Packed128, stochastic rounding, full determinism
- add multi-GPU training, NCCL, sharded optimizer
- add multi-node with MPI or file system or socket
- reproduce GPT-2 (1.6B) on one 8XH100 node in 24 hours for $672 in llm.c, achieving (at the time) 29% less memory, 19% faster training that PyTorch nightly, and much faster compile & run
- how open source development attracts Avengers from the internet
- port to training Llama 3 imminent (branch exists)
- many other notable forks
- last thought: how software abstractions like Python/PyTorch and everything else really exist only because humans are finite in knowledge, IQ and attention, and how with increasing AI capability LLMs may export custom binaries like llm.c for any application directly, tearing apart and refactoring all abstractions as needed.
<|endoftext|>
More links in reply
1
107
Thanks to the Interdisciplinary Research and Innovation Symposium 2024 for having me present a poster at the University of Aberdeen. Thanks to all the organisers for a well-run event. #DeepLearning #MachineLearning #ClimateCrisis
1
3
123
Steven Wallace retweeted
It has been a very successful BMVA Computer Vision Summer School this year. Thanks to all volunteers and staff from @comp_sci_durham and my co-Chairs @TobyBreckon @UlrikBeierholm @jjdengjj @sttujames
Steven Wallace retweeted
Dr Oscar Mendez @OscarMendezM delivered a talk on "Deep Learning for Computer Vision and Robotics" in the 27th BMVA Computer Vision Summer School in Durham University.
Steven Wallace retweeted
Dr Gwangbin Bae @BaeGwangbin delivered on a talk on "Always-on Perception for 3D Computer Vision" in the 27th BMVA Computer Vision Summer School in Durham University.
I had a great experience at the BMVA Summer School 2024 at Durham University. Thanks to the organisers and all the speakers who participated in the event. #AI #deeplearning #MachineLearning #computersciencestudents #BMVC2024
1
271
Steven Wallace retweeted
Prof. Victor Sanchez delivered a talk on "Contrastive Language-Image Pretraining (CLIP) for Video Analytics" in the 27th BMVA Computer Vision Summer School in Durham University.
Steven Wallace retweeted
Dr Andrew Gilbert delivered a talk on "Video Understanding" in the 27th BMVA Summer School in Durham University.
Steven Wallace retweeted
Dr. Paul Henderson @pmh47_ml delivered a talk on "Structured Generative Models for Computer Vision" in the 27th BMVA Summer School in Durham University.
It's been good to attend the BMVA 2024 Summer School. Great presentations and ideas were brought to the table this week. #AI #DeepLearning #MachineLearning #BMVA #BMVC
3
542
Steven Wallace retweeted
Dr. Oya Celiktutan @oyaceliktutan delivered a talk on "Multimodal Behaviour Understanding and Generation for Human-Robot Interaction" in the 27th BMVA Computer Vision Summer School in Durham.









































