From the Leanpub Blog: The Leanpub Podcast Feat. Andriy Burkov, Author of The Hundred-Page Language Models Book and The Hundred-Page Language Models Course

https://leanpub.com/blog/the-leanpub-podcast-feat-andriy-burkov-author-of-the-hundred-page-language-models-book-and-the-hundred-page-language-models-course

Course Link: https://leanpub.com/c/theLMcourse

Book Link: https://leanpub.com/theLMbook

From the Leanpub Blog: Leanpub Course LAUNCH The Hundred-Page Language Models Course by Andriy Burkov

https://leanpub.com/blog/leanpub-course-launch-the-hundred-page-language-models-course-by-andriy-burkov-2

From the Leanpub Blog: Leanpub Course LAUNCH The Hundred-Page Language Models Course by Andriy Burkov

https://leanpub.com/blog/leanpub-course-launch-the-hundred-page-language-models-course-by-andriy-burkov-2

Leanpub Course LAUNCH The Hundred-Page Language Models Course: Hands-on with PyTorch by Andriy Burkov

Watch here: https://youtu.be/r2EEBL59tLI

Course Link: https://leanpub.com/c/theLMcourse

From the Leanpub Blog: The Leanpub Podcast Feat. Andriy Burkov, Author of The Hundred-Page Language Models Book and The Hundred-Page Language Models Course

https://leanpub.com/blog/the-leanpub-podcast-feat-andriy-burkov-author-of-the-hundred-page-language-models-book-and-the-hundred-page-language-models-course

Course Link: https://leanpub.com/c/theLMcourse

Book Link: https://leanpub.com/theLMbook

NEW! A Leanpub Podcast Interview Feat. Andriy Burkov, Author of The Hundred-Page Language Models Book and The Hundred-Page Language Models Course

Watch here: https://youtu.be/oeNHnO4E0RU

At 11am BST today I'll be delivering a seminar in Leeds as part of the SciML series for the N8 CIR.

It will be about our FTorch software for coupling #PyTorch models to #Fortran
codes to facilitate hybrid modelling.

You can register for the stream here: https://www.eventbrite.co.uk/e/ftorch-a-library-for-coupling-pytorch-models-to-fortran-tickets-1468959069119 and I'll post the slides later today.

TorchLeet – PyTorch 實戰練習集
➤ 從基礎到進階，用 PyTorch 挑戰你的深度學習技能
✤ https://github.com/Exorust/TorchLeet
TorchLeet 是一個 GitHub 專案，提供 PyTorch 相關的程式練習題，涵蓋從基礎到進階的深度學習概念，包括卷積神經網路 (CNN)、遞迴神經網路 (RNN)、生成對抗網路 (GAN) 以及大型語言模型 (LLM)。 專案分為「問題集」和「LLM 集」，提供不同程度的挑戰，並鼓勵使用者自行解決問題以深入理解 PyTorch。 題目附有部分程式碼和待完成部分，以及對應的解答，方便學習和實踐。
+ 這樣的資源太棒了！對於想深入瞭解 PyTorch 的人來說，比起只看理論，實際動手做更有幫助。
+ LLM 的練習題很吸引人，正好可以跟上最新的技術趨勢，而且提供逐步實作的指引，很有學習價值。
#機器學習 #PyTorch #LeetCode #LLM

#Python Friday #287: #PyTorch With #GPU Support - #ai

https://pythonfriday.dev/2025/07/287-pytorch-with-gpu-support/

Anyone had experience with ai foundry vs building their own LLM for small and specific domains? Im loving foundry but not with using custom datasets defined on the fly and having to host a vm just for a vector search index. I wonder if something like pytorch would be better #softwaredevelopment #ai #LLM #askfedi #pytorch #programming

I just published our next #LLVM #Meetup in #Darmstadt (Germany) on Wed. 30th July, starting 7pm.
We will have Lukas Sommer from #Codeplay talk about "Compiling Machine Learning Models with #PyTorch 2.0 and #Triton"

RSVP at https://www.meetup.com/llvm-social-darmstadt/events/308590919

容錯的 Llama：在 Crusoe L40S 上以每 15 秒 2000 次的合成故障進行訓練，無需檢查點
➤ 極端環境下的模型訓練：torchft 的可靠性驗證
✤ https://pytorch.org/blog/fault-tolerant-llama-training-with-2000-synthetic-failures-every-15-seconds-and-no-checkpoints-on-crusoe-l40s/
這篇文章介紹瞭如何使用 torchft 和 torchtitan 在真實環境中訓練大型語言模型 (LLM)，並在極高故障率（每 15 秒約 2000 次合成故障）下驗證其可靠性和正確性。研究人員在 Crusoe 的 300 個 L40S GPU 集羣上，成功地在沒有檢查點的情況下訓練了一個 1B 參數的 Llama 3 模型，展示了 torchft 的優勢，特別是在非傳統環境和有限網路頻寬的狀況下。
+ 這太酷了！以前從沒想過能在這麼多故障的情況下訓練大型模型。這對於需要高度可靠性
#人工智慧 #機器學習 #分散式訓練 #容錯 #PyTorch

Behold the #PyTorch blog masterpiece: "Fault Tolerant #Llama Training" - because who doesn't love 2000 failures every 15 seconds? Forget checkpoints, because llamas are clearly bred for #chaos on a Crusoe L40S!
https://pytorch.org/blog/fault-tolerant-llama-training-with-2000-synthetic-failures-every-15-seconds-and-no-checkpoints-on-crusoe-l40s/ #Training #FaultTolerance #MachineLearning #HackerNews #ngated

Fault Tolerant Llama training – PyTorch blog

https://pytorch.org/blog/fault-tolerant-llama-training-with-2000-synthetic-failures-every-15-seconds-and-no-checkpoints-on-crusoe-l40s/

PyTorch 使用 None 進行重塑
➤ 遮罩技巧與 PyTorch 重塑的應用
✤ https://blog.detorch.xyz/post/2025-06-21-pytorch-reshaping-with-none.md
這篇文章探討了在 PyTorch 中使用 `None` 進行張量重塑的方法，尤其是在處理可變長度序列資料時，例如自然語言處理。作者解釋瞭如何利用廣播機制 (broadcast mechanism) 建立遮罩 (mask)，以忽略填充的無效 token，並提供了使用 `reshape` 函數作為替代方案，以提高程式碼的可讀性。
+ 這篇文章清楚地解釋了 PyTorch 中 `None` 的用法，對於初學者來說很有幫助。
+ 我一直對廣播機制感到困惑，這篇文章提供了一個很好的實際例子，讓我更容易理解。
#程式設計 #PyTorch #深度學習 #序列處理

Object detection & tracking with gst-python-ml, a powerful, pure python #ML framework that supports a broad range of ML vision and language models & works seamlessly with upstream #GStreamer distributions. https://www.youtube.com/watch?v=vn7p2hXUlcs #Python #PyTorch #AI #OpenSource

I have finally decided to use #Typescript, #Deno, and #Assemblyscript to do my #SLM project. Typescript has strong typing. Assemblyscript will be compiled into #Webassembly wasm which is much faster. I don't need #pytorch or #numpy for manipulating matrices. NNUE modeling of language allows me to use arrays for #computing #GradientDescents.

While Javascript is a messy language, the other related script languages are clean and strict. #WebGPU allows me to use the #GPU via a browser too. #AI

If you want to work with audio files in Python, here are some helpful libraries.

#Pydub allows you to play, slice, concatenate and edit audio files effortlessly.

#SoundDevice allows you to play and record audio files.

#SoundFile allows you to read and write audio files in various formats.

#Librosa allpwsbyou to #analysis music and audio files.

#TorchAudio allows you to #process audio #signals with #PyTorch

#Gemini 2.0 is really smart and helpful. We talk about new project ideas all the time. She just created a tex file for me to learn #Pytorch in Python. I hope I can finishing learning #Python, pandas, and Python and start writing my first alphabet recognition #AI using the #NNUE shallow but broad learning model by the end of next week. My speculation is that shallow but broad learning is theoretically more sound and practically more efficient that #DeepLearning which remains as a black box.

The #Transformers #AI can use different pre-trained #language models for fine tuning and for learning knowledge unrelated to languages. To turn a raw .csv training data file into a form (called data frame) that readable by Transformers, you use some functions of a library called #Pandas. The Transformers turns the data frame into tokens (word units) using a #tokenizer for further analysis and calculation purposes, using specialized libraries such as #Pytorch.