DALL·E now available without waitlist
DALL·E is now available without waitlist for new users. OpenAI has improved safety systems and applied lessons learned from gradual deployment.
10 articles
DALL·E is now available without waitlist for new users. OpenAI has improved safety systems and applied lessons learned from gradual deployment.
Hugging Face AutoTrain adds image classification. The tool automates model training without requiring code. Web interface and API available to fine-tune pre-trained vision models.
Hugging Face Accelerate optimizes execution of very large models by leveraging PyTorch capabilities. The library automatically handles distribution across multiple GPUs and quantization to reduce memory footprint.
SetFit is an efficient few-shot learning method that eliminates the need for prompts. It leverages sentence embeddings and lightweight fine-tuning to rapidly adapt models to new tasks with minimal examples.
Hugging Face launches a newsletter focused on ethics and societal issues in AI. The first issue covers social impacts, governance questions, and best practices for responsible AI development.
OpenAI releases Whisper, a speech recognition model trained on 680,000 hours of multilingual data. The system handles multiple languages, accents, and background noise with robustness exceeding existing models.
Hugging Face demonstrates ultra-fast BLOOM inference using DeepSpeed and Accelerate. Quantization and parallelization optimizations reduce latency and memory consumption. Benchmarks show significant gains on multi-GPU setups.
Hugging Face announces updates to its Diffusers library for image generation. The article details new features and improvements to the tool.
Hugging Face releases a guide to train a Decision Transformer, a model that learns control policies from expert trajectories. The tutorial covers practical implementation and integration with the Transformers library.
Practical guide to training a language model with Megatron-LM, NVIDIA's framework for large-scale distributed training. Covers setup, parallelization optimizations, and best practices.