/AI Weekly/Issue 5
How do GPT models work?
Published Monday, May 1, 2023
During training, the model is exposed to a lot of text, and acquires the ability to predict good probability distributions, given a sequence of input tokens. That’s because the model doesn’t actually produce a single predicted token; instead it returns a probability distribution over all the possible tokens.
ChatGPT users can now turn off chat history, allowing you to choose which conversations can be used to train our models.
Jingya Huang joins Cassie Breviu to talk about how to use Optimum + ONNX Runtime to accelerate the training of Hugging Face models. In the demo, we will fine-tune the DeBERTa model with Trainer and show the gain by leveraging the ONNX Runtime backend.
Silicon Valley’s quest to automate everything is unceasing, which explains its latest obsession: Auto-GPT. In essence, Auto-GPT uses the versatility of OpenAI’s latest AI models to interact with software and services online, allowing it to “autonomously” perform tasks like X and Y. But as we are learning with large language models, this capability seems to […] What is Auto-GPT and why does it matter? by Kyle Wiggers originally published on TechCrunch
Hidet is a powerful deep learning compiler that simplifies the process of implementing high-performing deep learning operators on modern accelerators (e.g., NVIDIA GPUs). With the new feature of torch.compile(...) in PyTorch 2.0, integrating a novel compiler into PyTorch is easier than ever - Hidet now can be used as a torch.compile(...) backend to accelerate PyTorch models, making it an attractive option for PyTorch users who want to improve the inference performance of their models, especially for those who also need to implement extremely optimized custom operators.
The AI links worth your time, curated by the community and always free.