if you think you hate jensen, but you have never installed pytorch, you don't really hate jensen.
January 19, 2026 at 7:37 AM
if you think you hate jensen, but you have never installed pytorch, you don't really hate jensen.
Flux2.c now supports the official img2img VAE-encoded inference, and it is like having a small Nano Banana in your computer, pretty impressive! Also the MPS target is almost as fast as PyTorch. github.com/antirez/flux...
January 22, 2026 at 1:07 PM
Flux2.c now supports the official img2img VAE-encoded inference, and it is like having a small Nano Banana in your computer, pretty impressive! Also the MPS target is almost as fast as PyTorch. github.com/antirez/flux...
Olé, if not for slower loading of weights, now flux2.c is as fast, using the MPS target, as the highly optimized PyTorch MPS backend. github.com/antirez/flux...
github.com
January 21, 2026 at 11:44 PM
Olé, if not for slower loading of weights, now flux2.c is as fast, using the MPS target, as the highly optimized PyTorch MPS backend. github.com/antirez/flux...
I used Gemini-CLI to build an offline translation tool using TranslateGemma + Gemma 3 to extract & translate text from images. Excited about the African Languages support!
Framework: HF Transformers
Interface: Gradio
Runtime: Python / PyTorch
Framework: HF Transformers
Interface: Gradio
Runtime: Python / PyTorch
January 21, 2026 at 7:35 PM
I used Gemini-CLI to build an offline translation tool using TranslateGemma + Gemma 3 to extract & translate text from images. Excited about the African Languages support!
Framework: HF Transformers
Interface: Gradio
Runtime: Python / PyTorch
Framework: HF Transformers
Interface: Gradio
Runtime: Python / PyTorch
I tried using Claude Opus 4.5 on a simple PyTorch MLP Notebook with the command to "make it maximally computationally efficient" just to see what would happen and it doubled the training throughput 🙃
January 16, 2026 at 6:03 PM
I tried using Claude Opus 4.5 on a simple PyTorch MLP Notebook with the command to "make it maximally computationally efficient" just to see what would happen and it doubled the training throughput 🙃
Interested in ✨world models✨? I just open-sourced an implementation of the Dreamer 4 world model. It's in PyTorch and comes with a pretrained model + a neat little web interface that lets you interact with any of 30 DMControl tasks that I trained it on!
Link: github.com/nicklashanse...
Link: github.com/nicklashanse...
January 15, 2026 at 6:20 PM
Interested in ✨world models✨? I just open-sourced an implementation of the Dreamer 4 world model. It's in PyTorch and comes with a pretrained model + a neat little web interface that lets you interact with any of 30 DMControl tasks that I trained it on!
Link: github.com/nicklashanse...
Link: github.com/nicklashanse...
If you're willing to spend another $1k for the nvidia version pytorch Just Works out of the box and you can get started training. I have been having 'fun' with gfx1151 drivers and getting pytorch working, which is still a WIP
January 18, 2026 at 1:29 PM
If you're willing to spend another $1k for the nvidia version pytorch Just Works out of the box and you can get started training. I have been having 'fun' with gfx1151 drivers and getting pytorch working, which is still a WIP
Just to say that I wanted to run the GTE-small embedding model in pure C, without any dependency, and Claude Code wrote the 917 lines of code needed in 4/5 minutes. Perfect match with the huge codebase needed by Python to perform the inference. Same speed (!) as PyTorch.
January 10, 2026 at 10:54 AM
Just to say that I wanted to run the GTE-small embedding model in pure C, without any dependency, and Claude Code wrote the 917 lines of code needed in 4/5 minutes. Perfect match with the huge codebase needed by Python to perform the inference. Same speed (!) as PyTorch.
Claude Opus 4.5、JAXのコードをPyTorchに移植することすらできず、使えない
January 8, 2026 at 1:50 PM
Claude Opus 4.5、JAXのコードをPyTorchに移植することすらできず、使えない
From PyTorch to the browser, simplified. LiteRT.js is a new #WebAI runtime designed to make it easier to run your custom .tflite models. See the demo from Matthew Soulanille and Chintan Parikh at the Web AI Summit → goo.gle/452pWm9
January 8, 2026 at 5:53 PM
From PyTorch to the browser, simplified. LiteRT.js is a new #WebAI runtime designed to make it easier to run your custom .tflite models. See the demo from Matthew Soulanille and Chintan Parikh at the Web AI Summit → goo.gle/452pWm9
With the cluster and storage down for maintenance this week, I thought about doing some #pytorch development locally an my laptop. No dedicated GPU, but a recent #intel Core Ultra 2 CPU. So first I went for the CPU-only packages.
Only yesterday I realized that the Windows task manager shows […]
Only yesterday I realized that the Windows task manager shows […]
Original post on academiccloud.social
academiccloud.social
January 8, 2026 at 11:24 AM
New Year’s resolution: more personal coding projects this year. Just finished pretraining and instruction fine tuning my own GPT2 scale LLM from scratch in PyTorch. Was so much fun!
January 5, 2026 at 4:21 PM
New Year’s resolution: more personal coding projects this year. Just finished pretraining and instruction fine tuning my own GPT2 scale LLM from scratch in PyTorch. Was so much fun!
A reminder that we don‘t have to use US products to build AI infrastructure.
- Mistral is a leading and very powerful French AI model.
- burn.dev is a Canadian built open source ML library in Rust with a number of powerful features competitive to PyTorch and Jax.
- Mistral is a leading and very powerful French AI model.
- burn.dev is a Canadian built open source ML library in Rust with a number of powerful features competitive to PyTorch and Jax.
Burn
Next Generation AI Infrastructure
burn.dev
January 4, 2026 at 6:50 AM
A reminder that we don‘t have to use US products to build AI infrastructure.
- Mistral is a leading and very powerful French AI model.
- burn.dev is a Canadian built open source ML library in Rust with a number of powerful features competitive to PyTorch and Jax.
- Mistral is a leading and very powerful French AI model.
- burn.dev is a Canadian built open source ML library in Rust with a number of powerful features competitive to PyTorch and Jax.
Pytorch使う際にM4GPUでの処理を指定してみたけど、速いのか遅いのか分からなかった
January 4, 2026 at 11:40 PM
Pytorch使う際にM4GPUでの処理を指定してみたけど、速いのか遅いのか分からなかった
I might have let a few long dt slip through recently to make it easier to train with pytorch. 😳
January 2, 2026 at 5:46 PM
I might have let a few long dt slip through recently to make it easier to train with pytorch. 😳
January 1, 2026 at 9:18 PM
For sure-that's critical. But 99% of PyTorch users couldn't derive backprop on a white board, much less program from scratch. So the entire field is based on codified primitives that are increasingly abstracted away from how the brain works. And that's fine for today's AI, but likely not tomorrow's.
January 2, 2026 at 3:40 AM
For sure-that's critical. But 99% of PyTorch users couldn't derive backprop on a white board, much less program from scratch. So the entire field is based on codified primitives that are increasingly abstracted away from how the brain works. And that's fine for today's AI, but likely not tomorrow's.
Deep Learning with PyTorch and Python : Neural Networks, Computer Vision, and NLP Applications
www.clcoding.com/2025/12/deep...
www.clcoding.com/2025/12/deep...
December 30, 2025 at 4:57 PM
Deep Learning with PyTorch and Python : Neural Networks, Computer Vision, and NLP Applications
www.clcoding.com/2025/12/deep...
www.clcoding.com/2025/12/deep...
As the Gentleman says, work with train, eval and test sets. PyTorch (as Sklearn) provides a good overview of regression results. Use regularization to optimize the model, then final test on unseen test set. May I ask why you use PyTorch? Not a natural choice unless you do deep learning as well as LR
December 30, 2025 at 2:10 PM
As the Gentleman says, work with train, eval and test sets. PyTorch (as Sklearn) provides a good overview of regression results. Use regularization to optimize the model, then final test on unseen test set. May I ask why you use PyTorch? Not a natural choice unless you do deep learning as well as LR
Python Libraries for ML Engineers
- Data Handling: Pandas, Dask
- Feature Engineering: Featuretools, Scikit‑learn
- Model Training: TensorFlow, PyTorch
- Experiment Tracking: MLflow
- Model Deployment: ONNX, TorchServe
- Data Handling: Pandas, Dask
- Feature Engineering: Featuretools, Scikit‑learn
- Model Training: TensorFlow, PyTorch
- Experiment Tracking: MLflow
- Model Deployment: ONNX, TorchServe
December 29, 2025 at 4:23 PM
Python Libraries for ML Engineers
- Data Handling: Pandas, Dask
- Feature Engineering: Featuretools, Scikit‑learn
- Model Training: TensorFlow, PyTorch
- Experiment Tracking: MLflow
- Model Deployment: ONNX, TorchServe
- Data Handling: Pandas, Dask
- Feature Engineering: Featuretools, Scikit‑learn
- Model Training: TensorFlow, PyTorch
- Experiment Tracking: MLflow
- Model Deployment: ONNX, TorchServe
🚀 Datalab seeks a Research Engineer in NYC! Build tiny LLMs that outperform big models for OCR & document AI, ship open‑source tools. 3+ yrs LLM training, PyTorch, Python, H100/L40/CPU required. #MachineLearning #AI #OpenSource #DocAI #NYC aihackerjobs.com/company/data...
Research Engineer - Datalab · AI Hacker Jobs
Research Engineer at Datalab - Python, PyTorch, LLM, H100, L40, CPU | Apply now on AIHackerJobs.com
aihackerjobs.com
December 26, 2025 at 3:37 PM
🚀 Datalab seeks a Research Engineer in NYC! Build tiny LLMs that outperform big models for OCR & document AI, ship open‑source tools. 3+ yrs LLM training, PyTorch, Python, H100/L40/CPU required. #MachineLearning #AI #OpenSource #DocAI #NYC aihackerjobs.com/company/data...
2004: הקומפיילר של C++ אומר לי אם הקוד שכתבתי לא מתקמפל
2009: מגלה בזמן ריצה אם קוד הפייתון שכתבתי תקין או לא
2020: תגיד, mypy, עשיתי שגיאות typing? ואם אפשר תרוץ מהר הפעם
2022: מגלה בזמן ריצה אם עשיתי שגיאות מימדים ב-pytorch
2025: קלוד אחי תעיף מבט ב-git diff. עשיתי שגיאות איפשהו?
2009: מגלה בזמן ריצה אם קוד הפייתון שכתבתי תקין או לא
2020: תגיד, mypy, עשיתי שגיאות typing? ואם אפשר תרוץ מהר הפעם
2022: מגלה בזמן ריצה אם עשיתי שגיאות מימדים ב-pytorch
2025: קלוד אחי תעיף מבט ב-git diff. עשיתי שגיאות איפשהו?
December 24, 2025 at 4:49 AM
2004: הקומפיילר של C++ אומר לי אם הקוד שכתבתי לא מתקמפל
2009: מגלה בזמן ריצה אם קוד הפייתון שכתבתי תקין או לא
2020: תגיד, mypy, עשיתי שגיאות typing? ואם אפשר תרוץ מהר הפעם
2022: מגלה בזמן ריצה אם עשיתי שגיאות מימדים ב-pytorch
2025: קלוד אחי תעיף מבט ב-git diff. עשיתי שגיאות איפשהו?
2009: מגלה בזמן ריצה אם קוד הפייתון שכתבתי תקין או לא
2020: תגיד, mypy, עשיתי שגיאות typing? ואם אפשר תרוץ מהר הפעם
2022: מגלה בזמן ריצה אם עשיתי שגיאות מימדים ב-pytorch
2025: קלוד אחי תעיף מבט ב-git diff. עשיתי שגיאות איפשהו?
Physicists probably laugh when they look at tensors in pytorch and at the losers who live only on Euclidean metric space.
December 24, 2025 at 6:02 AM
Physicists probably laugh when they look at tensors in pytorch and at the losers who live only on Euclidean metric space.
Good thread. I've been using this slide for those who insist on a technical definition for AI/ML in science. Coming from the perspective of scientific/numerical computing, the colloquial meaning does a lot to exclude (good, "classical") methods. It's deeply frustrating and constraining.
December 22, 2025 at 7:59 PM
Good thread. I've been using this slide for those who insist on a technical definition for AI/ML in science. Coming from the perspective of scientific/numerical computing, the colloquial meaning does a lot to exclude (good, "classical") methods. It's deeply frustrating and constraining.