NVIDIA NeMo Automodel now works seamlessly with Hugging Face Diffusers, enabling production-grade distributed training of video and image models without checkpoint conversion.
A focused OCR model trained on Brazilian Portuguese outperforms larger, newer competitors through targeted fine-tuning and preference optimization.
Hugging Face data shows LoRA controls 98% of PEFT implementations, but emerging techniques like DoRA and LoHa challenge its monopoly on parameter-efficient adaptation.
Hugging Face released a fine-tuned 8B model that filters LinkedIn job postings by matching them against candidate resumes, using structured reasoning from a larger teacher model.
Dharma's DharmaOCR benchmark shows task-specific fine-tuning can beat parameter scale in production AI economics.
The bitsandbytes library applies 4-bit and 8-bit quantization to PyTorch models, making 70B+ parameter LLMs runnable on consumer GPUs and underpinning the QLoRA fine-tuning wave.