List Transcripts

curl --request GET \
  --url https://transcript.lol/api/listUserTranscripts \
  --header 'X-API-KEY: <api-key>'

[
  {
    "transcriptId": "652bdbdaf970bb1ffa500916",
    "sourceUrl": "https://www.youtube.com/watch?v=kCc8FmEb1nY",
    "platform": "youtube",
    "title": "Let's build GPT: from scratch, in code, spelled out.",
    "description": "We build a Generatively Pretrained Transformer (GPT), following the paper \"Attention is All You Need\" and OpenAI's GPT-2 / GPT-3. We talk about connections to ChatGPT, which has taken the world by storm. We watch GitHub Copilot, itself a GPT, help us write a GPT (meta :D!) . I recommend people watch the earlier makemore videos to get comfortable with the autoregressive language modeling framework and basics of tensors and PyTorch nn, which we take for granted in this video.\n\nLinks:\n- Google colab for the video: https://colab.research.google.com/drive/1JMLa53HDuA-i7ZBmqV7ZnA3c_fvtXnx-?usp=sharing\n- GitHub repo for the video: https://github.com/karpathy/ng-video-lecture\n- Playlist of the whole Zero to Hero series so far: https://www.youtube.com/watch?v=VMj-3S1tku0&list=PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ\n- nanoGPT repo: https://github.com/karpathy/nanoGPT\n- my website: https://karpathy.ai\n- my twitter: https://twitter.com/karpathy\n- our Discord channel: https://discord.gg/3zy8kqD9Cp\n\nSupplementary links:\n- Attention is All You Need paper: https://arxiv.org/abs/1706.03762\n- OpenAI GPT-3 paper: https://arxiv.org/abs/2005.14165 \n- OpenAI ChatGPT blog post: https://openai.com/blog/chatgpt/\n- The GPU I'm training the model on is from Lambda GPU Cloud, I think the best and easiest way to spin up an on-demand GPU instance in the cloud that you can ssh to: https://lambdalabs.com . If you prefer to work in notebooks, I think the easiest path today is Google Colab.\n\nSuggested exercises:\n- EX1: The n-dimensional tensor mastery challenge: Combine the `Head` and `MultiHeadAttention` into one class that processes all the heads in parallel, treating the heads as another batch dimension (answer is in nanoGPT).\n- EX2: Train the GPT on your own dataset of choice! What other data could be fun to blabber on about? (A fun advanced suggestion if you like: train a GPT to do addition of two numbers, i.e. a+b=c. You may find it helpful to predict the digits of c in reverse order, as the typical addition algorithm (that you're hoping it learns) would proceed right to left too. You may want to modify the data loader to simply serve random problems and skip the generation of train.bin, val.bin. You may want to mask out the loss at the input positions of a+b that just specify the problem using y=-1 in the targets (see CrossEntropyLoss ignore_index). Does your Transformer learn to add? Once you have this, swole doge project: build a calculator clone in GPT, for all of +-*/. Not an easy problem. You may need Chain of Thought traces.)\n- EX3: Find a dataset that is very large, so large that you can't see a gap between train and val loss. Pretrain the transformer on this data, then initialize with that model and finetune it on tiny shakespeare with a smaller number of steps and lower learning rate. Can you obtain a lower validation loss by the use of pretraining?\n- EX4: Read some transformer papers and implement one additional feature or change that people seem to use. Does it improve the performance of your GPT?\n\nChapters:\n00:00:00 intro: ChatGPT, Transformers, nanoGPT, Shakespeare\nbaseline language modeling, code setup\n00:07:52 reading and exploring the data\n00:09:28 tokenization, train/val split\n00:14:27 data loader: batches of chunks of data\n00:22:11 simplest baseline: bigram language model, loss, generation\n00:34:53 training the bigram model\n00:38:00 port our code to a script\nBuilding the \"self-attention\"\n00:42:13 version 1: averaging past context with for loops, the weakest form of aggregation\n00:47:11 the trick in self-attention: matrix multiply as weighted aggregation\n00:51:54 version 2: using matrix multiply\n00:54:42 version 3: adding softmax\n00:58:26 minor code cleanup\n01:00:18 positional encoding\n01:02:00 THE CRUX OF THE VIDEO: version 4: self-attention\n01:11:38 note 1: attention as communication\n01:12:46 note 2: attention has no notion of space, operates over sets\n01:13:40 note 3: there is no communication across batch dimension\n01:14:14 note 4: encoder blocks vs. decoder blocks\n01:15:39 note 5: attention vs. self-attention vs. cross-attention\n01:16:56 note 6: \"scaled\" self-attention. why divide by sqrt(head_size)\nBuilding the Transformer\n01:19:11 inserting a single self-attention block to our network\n01:21:59 multi-headed self-attention\n01:24:25 feedforward layers of transformer block\n01:26:48 residual connections\n01:32:51 layernorm (and its relationship to our previous batchnorm)\n01:37:49 scaling up the model! creating a few variables. adding dropout\nNotes on Transformer\n01:42:39 encoder vs. decoder vs. both (?) Transformers\n01:46:22 super quick walkthrough of nanoGPT, batched multi-headed self-attention\n01:48:53 back to ChatGPT, GPT-3, pretraining vs. finetuning, RLHF\n01:54:32 conclusions\n\nCorrections: \n00:57:00 Oops \"tokens from the _future_ cannot communicate\", not \"past\". Sorry! :)\n01:20:05 Oops I should be using the head_size for the normalization, not C",
    "thumbnailUrl": "https://res.cloudinary.com/dnvn5ub8z/image/upload/v1697373145/transcript-lol-prod/5tod6bah20.webp",
    "duration": 6979,
    "currentStatus": "COMPLETED",
    "failureReason": "UNKNOWN",
    "uploaderId": "@andrejkarpathy",
    "uploaderName": "Andrej Karpathy",
    "uploadedAt": "2023-01-17T05:42:21.000Z"
  }
]