Downloads a given GTC 2024 video from the conference catalogue or from the on-demand sessions.

• Ensure you have Poetry and ffmpeg installed.
• Run poetry install to install the required dependencies.

• -d/--directory: The directory to save the video to. Defaults to the current directory.
• -m/--meta: Save metadata about the video to a JSON file. Stored in the same directory as the video as {AUTHORS} - {TITLE}.json.

One of the following is required:

• -o/--ondemand-session-url: The URL of the on-demand session to be downloaded.
• -c/--conf-session-id: The session ID of the session from the conference catalogue to be downloaded. This can be found in the URL of the session page. For example, https://www.nvidia.com/gtc/session-catalog/?#/session/1702594702652001JJhD has an ID of 1702594702652001JJhD.
  • -a/--rainforest-auth: Your Rainforest authentication token - required if downloading from the conference catalogue. See below for instructions.

You can download a video from an on-demand session by providing the URL of the session:

poetry run python download.py -m -o "$ONDEMAND_SESSION_URL"

This will download the video to the current directory with the filename {AUTHORS} - {TITLE}.mp4.

You will need to get an authentication token from Rainforest. You only need to do this if you don't have a token or if your token has expired; otherwise, you can keep using the same token.

1. Go to https://www.nvidia.com/gtc/session-catalog/?#/ while logged in.
2. Open the developer console (F12 / Ctrl+Shift+I) and go to the "Network" tab, and filter on XHR requests.
3. Reload the page.
4. Look for any POST requests to events.rainfocus.com after the login request. This will be the last few requests. (e.g. myData, attributes, search, etc)
5. In the headers for the request, look for rfAuthToken. Copy the value. It should be a long string of random characters.

You can now download a video given its session ID:

poetry run python download.py -a "$YOUR_AUTH_TOKEN" -m --conf-session-id $SESSION_ID

This will download the video to the current directory with the filename {AUTHORS} - {TITLE}.mp4.

I downloaded these while testing. You may find them useful as well:

• https://www.nvidia.com/en-us/on-demand/session/gtc24-s62246/: JAX Supercharged on GPUs: High Performance LLMs with JAX and OpenXLA (Chang Lan, Nitin, Qiao Zhang)

These may no longer work as the conference has ended. You can try to find the corresponding on-demand session instead.

• 1694444293481001kBAn: Better, Cheaper, Faster LLM Alignment With KTO (Amanpreet Singh)
• 1706747368510001RGVh: Fireside Chat with David Luan and Bryan Catanzaro: The Future of AI and the Path to AGI (Bryan Catanzaro, David Luan)
• 1696440824445001lFOk: Fireside Chat With Kanjun Qiu and Bryan Catanzaro: Building Practical AI Agents that Reason and Code at Scale (Bryan Catanzaro, Kanjun Qiu)
• 1694213266066001wkhe: Beyond Transformers: A New Architecture for Long Context and Linear Performance (Ce Zhang)
• 1705548169759001iR7s: Fireside Chat with Christian Szegedy and Bojan Tunguz: Automated Reasoning for More Advanced Software Synthesis and Verification (Christian Szegedy, Bojan Tunguz)
• 1706123152858001fNnT: Scaling Grok with JAX and H100 (Igor Babuschkin)
• 1702594702652001JJhD: Transforming AI (Jensen Huang, Ashish Vaswani, Noam Shazeer, Jakob Uszkoreit, Llion Jones, Aidan Gomez, Lukasz Kaiser, Illia Polosukhin)
• 1695673049743001rLIc: A Culture of Open and Reproducible Research, in the Era of Large AI Generative Models (Joelle Pineau)
• 1696292994102001l720: Diffusion Models: A Generative AI Big Bang (Karsten Kreis, Arash Vahdat)
• 1705085165951001X0W8: From Zero to Millions: Scaling Large Language Model Inference With TensorRT-LLM (Kevin Hu)
• 1696033648682001S1DC: CUDA: New Features and Beyond (Stephen Jones)