Menu

#Pytorch

59 posts

Feed·
20 of 59 posts
Your PyTorch Code on Trainium: What Actually Breaks
📰
0Read More

Your PyTorch Code on Trainium: What Actually Breaks

DEV Community: pytorch·RobustTrueTry·6 days ago
#p5NWJaHx
#devtopytorch#devto#dev#model#pytorch

A practitioner's porting checklist for PyTorch on AWS Trainium: compile workflow, unsupported ops, and the training-loop pitfalls that stall your step

15s
0Read More

Run Small AI Models on an Old GPU: Fix the Software Path First

DEV Community: python·Anuj Sadani·15 days ago
#eK63OnBF
#devtopython#devto#dev#driver#path

A six-year-old GTX 1650 with 4 GB of VRAM can still run small models. In my case, the obstacles were mostly software: a CPU-only PyTorch wheel, a missing PATH entry, a dtype unsupported by Turing, and a batch size that pushed data into system RAM.…

15s
How PyTorch Became the Socket Every AI Chip Plugs Into
🖼️
0Read More

How PyTorch Became the Socket Every AI Chip Plugs Into

DEV Community: pytorch·Dhruv Agrawal·19 days ago
#dPWi3GCt
#devtopytorch#devto#dev#pytorch#torchtpu

Google and AWS released PyTorch backends a day apart, both through the same small hook. Here's what that hook is, and why getting code to run on a new chip is the easy half.

15s
Gemma 4 on a 4 GB GPU in Plain PyTorch: From Out of Memory to Faster Than llama.cpp
📰
0Read More

Gemma 4 on a 4 GB GPU in Plain PyTorch: From Out of Memory to Faster Than llama.cpp

DEV Community: pytorch·xbill·22 days ago
#mal3JuAW
#devtopytorch#devto#dev#pytorch#llama

On a 4 GB GTX 1650 Ti, Gemma 4 E2B does not fit in 16-bit, and Transformers' own 4-bit path runs out of memory on its first forward pass. Kept in Google's QAT 4-bit end to end, with a few Triton kernels and a CUDA graph, plain PyTorch decodes at 88 tok/s,…

15s
Now Open Source: TorchTPU and the TPU Agent Suite for Native PyTorch on Google Cloud TPUs
📰
0Read More

Now Open Source: TorchTPU and the TPU Agent Suite for Native PyTorch on Google Cloud TPUs

DEV Community: cloud·Keelin McDonell·24 days ago
#yDJCUkvZ
#devtocloud#devto#dev#pytorch#agent

We’re excited to share that TorchTPU—our native PyTorch stack for Google Cloud TPUs—is now open...

15s
PyTorch vs. TensorFlow for Diffusion Models: A Dev Leader's Guide to Tooling and Productivity
🖼️
0Read More

PyTorch vs. TensorFlow for Diffusion Models: A Dev Leader's Guide to Tooling and Productivity

DEV Community: tensorflow·Oleg·25 days ago
#QlmbCM58
#devtotensorflow#devto#dev#state#models

The Great Debate: PyTorch vs. TensorFlow for Diffusion Models In the rapidly evolving...

15s
Amazon Opens TorchNeuron to Run PyTorch Code on Its Trainium AI Chips
🖼️
0Read More

Amazon Opens TorchNeuron to Run PyTorch Code on Its Trainium AI Chips

#webpronews#news#amazon#trainium#pytorch

Amazon has opened TorchNeuron, software that runs existing PyTorch code on its Trainium AI chips after a simple configuration change, as a public beta and is open-sourcing it at the PyTorch Conference on Oct. 20.

15s
Google Open-Sources TorchTPU and Ships llm-d 1.0 for Kubernetes AI Inference
🖼️
0Read More

Google Open-Sources TorchTPU and Ships llm-d 1.0 for Kubernetes AI Inference

#webpronews#news#google#torchtpu#pytorch

Google open-sourced TorchTPU, its native PyTorch stack for TPUs, and released llm-d 1.0, a Kubernetes-native distributed inference stack co-founded with Red Hat, IBM and others, at PyTorch Conference 2026.

15s
AWS's TorchNeuron Runs PyTorch on Trainium With .to('neuron'). Shapes Must Be Static
🖼️
0Read More

AWS's TorchNeuron Runs PyTorch on Trainium With .to('neuron'). Shapes Must Be Static

DEV Community: pytorch·Gabriel Anhaia·25 days ago
#A8nmSHMU
#devtopytorch#devto#dev#torch#neuron

AWS open-sources TorchNeuron on October 20: Trainium becomes a normal PyTorch device. The beta wants static shapes, pinned versions and PyTorch 2.14.

15s
Google and Amazon bring AI applications to their chips without rewriting code
🖼️
0Read More

Google and Amazon bring AI applications to their chips without rewriting code

DEV Community: pytorch·Hacks.gr·25 days ago
#aBt75K8n
#devtopytorch#devto#dev#applications#pytorch

TorchTPU is Google’s open-source project for running PyTorch applications on its Tensor Processing Units without rewriting application code. Announced on 2

15s
Google has joined Amazon in attacking Nvidia's stronghold with AI app developers
🖼️
0Read More

Google has joined Amazon in attacking Nvidia's stronghold with AI app developers

DEV Community: pytorch·Chandan Kumar·25 days ago
#5N1UEmIS
#devtopytorch#devto#dev#google#nvidia

Google's new PyTorch porting tool competes on migration cost rather than FLOPS, and that shifts how teams should evaluate non-Nvidia...

15s