Bayesian MMM as a decision engine 🧠 Causal funnels with PyMC-Marketing, optimized with PyTensor, and explored via @marimo_io with interactive counterfactuals. Full walkthrough by Teemu Säilynoja 👇 dub.link/RP5Zt8Z #PyMC #MMM #MarketingAnalytics

🚀 Big news for Mac users! PyTensor now supports the MLX backend - bringing Apple Silicon (M1/M2/M3...) hardware acceleration to PyMC. ⚡ Try it out by setting mode="MLX" and share your feedback! Huge thanks to all contributors! 🙌 #PyMC #PyTensor #MLX #AppleSilicon

This week at @_odsc AI West 2025, we're hosting a hands-on #PyTensor tutorial! Learn how the computational graph library powering #PyMC makes probabilistic models scalable & efficient, supports complex inference, and drives real-world AI decisions. #ODSC2025

Final session of our #PyTensor Workshop is live! Discover custom rewrites, speed up diagonal matrix inversions 80x, and see a mini-PyMC in action - one graph powering the full Bayesian workflow. Watch here: dub.link/MeFg68hgf #PyMC #Python #ML #DataScience

𝐎𝐩𝐞𝐧 𝐬𝐨𝐮𝐫𝐜𝐞 𝐠𝐫𝐨𝐰𝐬 𝐟𝐚𝐬𝐭𝐞𝐫 𝐰𝐡𝐞𝐧 𝐰𝐞 𝐛𝐮𝐢𝐥𝐝 𝐭𝐨𝐠𝐞𝐭𝐡𝐞𝐫. Join Rob Zinkov at @pydataamsterdam 10th Anniversary on Sep 26, 1:20 PM for a beginner-friendly #PyMC sprint. Learn, contribute, and connect with the community! #PyTensor

Session 2 of our #PyTensor workshop dives into 𝐆𝐫𝐚𝐩𝐡 𝐓𝐫𝐚𝐧𝐬𝐟𝐨𝐫𝐦𝐚𝐭𝐢𝐨𝐧𝐬! Learn eager vs. lazy transformations, autodiff under the hood, and Normalizing Flows. Watch the full session here dub.link/CR3WDRf #PyMC #ML

Behind every #PyMC model is #PyTensor — a computational graph library for speed, flexibility & autodiff. Session 1 of our PyTensor Workshop is live 🎥 👉 dub.link/pytensor-graphs #PyMC #Python #MachineLearning

PyMCが復活しました～ヽ(=´▽`=)ﾉ PC環境とPyTensorの相性が悪くて、回避策を講じるのに時間がかかりました😅 ・G のインストール ・コンパイル用パスの環境設定 ・MCMCでJaxを使わない方法 綱渡りはまだまだ続きそうです… 図はPyMCっぽいでしょｗ #のんびり統計 #PyMC #ベイズモデリング

The DeepCLIP tool takes a single or more RNA oligonucleotides (short RNA sequences) as input and predicts binding probability and calculates a binding profile. The main purpose of DeepCLIP is to identify binding sites of proteins in novel untested sequences using trained models that have extracted binding site information provided by CLIP data, to predict the effect of sequence variants on the binding, and to identify the importance of individual nucleotides for protein binding affinity. deepclip.compbio.sdu.dk/ DeepCLIP: predicting the effect of mutations on protein–RNA binding with deep learning We present DeepCLIP, a novel deep learning approach to modeling RNA-binding protein sites using a shallow neural network composed of CNN and LSTM layers to capture context-dependent binding. DeepCLIP generalizes well across a diverse set of sequences in both in vitro and in vivo settings, and produces a profile of the sequence, which indicates sequence elements important for the binding of the RNA binding protein in question. The core of DeepCLIP is a convolutional BLSTM network implemented in Theano using the Lasagne library and a few customized network layers and functions. academic.oup.com/nar/article… Theano 1.0.5 pypi.org/project/Theano/#des… Theano: a {Python} framework for fast computation of mathematical expressions arxiv.org/abs/1605.02688 Theano was a Python library that allows you to define, optimize, and evaluate mathematical expressions involving multi-dimensional arrays efficiently. It is being continued as PyTensor. PyTensor is a fork of Aesara,aesara.readthedocs.io/en/lat… which is a fork of Theano. PyTensor PyTensor is a Python library that allows one to define, optimize, and efficiently evaluate mathematical expressions involving multi-dimensional arrays. It provides the computational backend for PyMC.(pymc.io/projects/docs/learn.…) github.com/pymc-devs/pytenso… Lasagne: First release A lightweight neural network library built on top of Theano. zenodo.org/records/27878

参考になるかわからんですが、真っ新な状態から普通に poetry add で入りました。poetry の pyproject.toml が以下で動いてますよー Python は 3.12.7 です。 [tool.poetry.dependencies] python = ">=3.10,<3.13" pytensor = "^2.26.3" pymc = "^5.18.2" notebook = "^7.2.2" tensorflow = "^2.18.0" torch = "^2.5.1" torchvision = "^0.20.1" graphviz = "^0.20.3" ipywidgets = "^8.1.5"

Spacy便利だったのになぁ・・・🤔 Python=3.11だったら、自作Webアプリでも、TensorflowとSpacy, PyMC5(Pytensor)が共存できたのに😩 Python=3.12にしたら、共存できなくなったのは少し残念😢 #Python

無事に完了したと思ったら・・・ spacy==3.7.6が削除された・・・と😩 最新の3.8.2にしたら、numpyを2.0以上にしないとダメで・・・ そうすると、pytensorの最新版でも、numpy<2.0 なので、最終的に、streamlit上でPymc5のMCMC激遅い問題が再発してしまう😩 最終的に、Spacyを諦める事に😅

Pythonのバージョン 3.12 に対応させる問題・・・ 結構、大変なんすよね😭 PyMCのMCMC激遅い問題を解決するため、pytensorをこのバージョンにしたんですけど、Python 3.12にするとエラーになるんよね😭 PyMCがマジでネックなんだよなぁ😭 暫くは、3.11で、問題を先送りするか・・・😩

移動中、電車の中で、PyMC5のMCMCが激遅い問題を考えているのですが… 仮想環境で、Pythonのバージョンを下げて… NumPy<=2.20、pytensor==2.18、pandas, scikit-learn, scipy, matplotlib の衝突を回避できても、streamlitだとPythonのバージョンを指定できないと思うけど、違うかな🤔

タクシーの中で、 PyMC5のMCMCが激遅い問題を調べてたのですが… もしかしたら、NumPyのバージョンが高すぎるのかも🤔 Pytensorも、2.18くらいじゃないとアカンかも🤔 でも、Pythonが、3.11だから… なんかもう、考えるだけで、ウンザリしてくるわ😂

酔ってるけど・・・ 対して難しくないので、さっさとPyMC3で実装してみました🤔 PyMC5の時は別の警告(mkl-serviceをcondaでinstallしたぜ)が出て・・・ 結論、PyMC3もPyMC5も、MCMCが激遅い😂 pytensorのバージョンもあげたのに。。。😩😩

遅れ分布カーネルを使った一次元ベクトルPyTensorオブジェクトとの内積は取れる、ように見える、スカラ扱いだからかな

PyTensorの型にすれば普通に独自関数でできるんじゃろうか、grad要素がなくなるというのがいまいちよくわからん、 ただこれを使えるようになればかなり自由度が上がりそう

i was just confused why they're comparing it to pytensor lol "hmm this data vector sure looks like a data vector"

Major new BEAT release 2.0.0. is out! Finally, supporting python3.9 . #pymc #pyrocko #pytensor installation. New features: Multi-Sourcetype inference (e.g.Rectangular & MTSource) and boundary element modeling mode using GPU #cutde @tbenthompson. New tutorials coming. Stay tuned!