Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
mathstodon.xyz is one of the many independent Mastodon servers you can use to participate in the fediverse.
A Mastodon instance for maths people. We have LaTeX rendering in the web interface!

Administered by:

Server stats:

2.8K
active users

mathstodon.xyz: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.7

#pytorch

10 posts10 participants0 posts today
Public
Hacker News

PyGraph: Robust Compiler Support for CUDA Graphs in PyTorch

arxiv.org/abs/2503.19779

arXiv logo
arXiv.orgPyGraph: Robust Compiler Support for CUDA Graphs in PyTorchCUDA Graphs -- a recent hardware feature introduced for NVIDIA GPUs -- aim to reduce CPU launch overhead by capturing and launching a series of GPU tasks (kernels) as a DAG. However, deploying CUDA Graphs faces several challenges today due to the static structure of a graph. It also incurs performance overhead due to data copy. In fact, we show a counter-intuitive result -- deploying CUDA Graphs hurts performance in many cases. We introduce PyGraph, a novel approach to automatically harness the power of CUDA Graphs within PyTorch2. Driven by three key observations, PyGraph embodies three novel optimizations: it enables wider deployment of CUDA Graphs, reduces GPU kernel parameter copy overheads, and selectively deploys CUDA Graphs based on a cost-benefit analysis. PyGraph seamlessly integrates with PyTorch2's compilation toolchain, enabling efficient use of CUDA Graphs without manual modifications to the code. We evaluate PyGraph across various machine learning benchmarks, demonstrating substantial performance improvements over PyTorch2.
#HackerNews#PyGraph#CUDA
Public *
Denzil Ferreira :fedora:

So, good news. ROCm 6.3.4 and PyTorch 2.4.0 seems stable enough with gfx1103 if I use HSA override for 11.0.0, using latest firmware blobs and kernel 6.13.10 on Fedora 41.

In your Dockerfile, build your AI app from:
```
FROM rocm/pytorch:rocm6.3.4_ubuntu24.04_py3.12_pytorch_release_2.4.0
```

#rocm#pytorch#amdgpu
Public
Amartya

My brain is absolutely fried.
Today is the last day of coursework submissions for this semester. What a hectic month.
DNN with PyTorch, Brain model parallelisation with MPI, SYCL and OpenMP offloading of percolation models,hand optimizing serial codes for performance.
Two submissions due today. Submitted one and finalising my report for the second one.
Definitely having a pint after this

#sycl#hpc#msc
Public
sc0v0ne

MCP, Agentic Knowledge Graphs & AI Models: Solving Conversational Analytics

eventbrite.com/e/mcp-agentic-k

In this free webinar led by ex-Snowflake, Cloudera, and Amazon leaders, we'll unveil how cutting-edge LLMs (GPT 4.5, Sonnet 3.7, Deepseek V3/R1, Gemini 2.5, etc.) are revolutionizing data products.
#python #machinelearning #deeplearning #ai #developer #dev #devsecops #devops #mlops #learn #learning #study #git #github #codeberg #tensorflow #pytorch #jax #huggingface #linux #ubuntu #popos #llm

EventbriteMCP, Agentic Knowledge Graphs & AI Models: Solving Conversational AnalyticsCo-hosted by Gunther Hagleitner and Wangda Tan, former engineering and product leaders at Snowflake, Cloudera, & Cribl. Founders of Waii.ai.
SearchLive feeds
About