iFANN
    Search iFANN...
    Log in
    Home
    News
    Videos
    Photos
    GIFs
    Explore
    Polls
    Awards
    iFAMOUS
    Wiki
    Anime
    Rooms
    Notifications
    Messages
    Bookmarks
    Profile
    WikiAwardsiFAMOUSRankingsIndustriesCreator RewardsUser RewardsTermsPrivacyCommunity GuidelinesTakedown / DMCAHelpDevelopers

    © 2026 iFANN

    Home
    Search
    Messages
    Alerts
    Profile

    Post

    Vojtech
    Vojtech@vojtech
    🎮Claude🏛️Stanford University📱ChatGPT

    Stanford CME 295 Transformers lecture

    The math under Claude and ChatGPT is what CME 295: Transformers & Large Language Models actually walks through, taught at Stanford by Afshine Amidi and Shervine Amidi. One lecture, 104 minutes, zero dollars. Starting from raw text, the whole climb to a working Transformer gets covered. First up: tokenization, and why splitting words into subwords happens through BPE. Then Word2Vec, plus the reason a single vector per word fell short. RNNs and LSTMs are next, with the long-sentence problem they couldn't handle. Self-attention gets its own pass, Q, K, V included, along with the formula every model you use runs on. It ends on a complete Transformer, walked through one step at a time, turning an English sentence into French. No hype. Just the math and the architecture. Bookmark it so you don't lose track of it. Then there's what that architecture can build. I put together an A-Z guide on creating 3D animations in Blender with Claude, all from one brief. Below.

    1w

    5 Likes0 Dislikes0 Reposts0 Comments
    ?

    Comments

    No comments yet. Be the first!

    Post

    Vojtech
    Vojtech@vojtech
    🎮Claude🏛️Stanford University📱ChatGPT

    Stanford CME 295 Transformers lecture

    The math under Claude and ChatGPT is what CME 295: Transformers & Large Language Models actually walks through, taught at Stanford by Afshine Amidi and Shervine Amidi. One lecture, 104 minutes, zero dollars. Starting from raw text, the whole climb to a working Transformer gets covered. First up: tokenization, and why splitting words into subwords happens through BPE. Then Word2Vec, plus the reason a single vector per word fell short. RNNs and LSTMs are next, with the long-sentence problem they couldn't handle. Self-attention gets its own pass, Q, K, V included, along with the formula every model you use runs on. It ends on a complete Transformer, walked through one step at a time, turning an English sentence into French. No hype. Just the math and the architecture. Bookmark it so you don't lose track of it. Then there's what that architecture can build. I put together an A-Z guide on creating 3D animations in Blender with Claude, all from one brief. Below.

    1w

    5 Likes0 Dislikes0 Reposts0 Comments
    ?

    Comments

    No comments yet. Be the first!