Search Results for "command line dynamic"

...Trained on a large bilingual dataset — including 30 million high-quality Chinese image-text pairs from CogView and 300 million English pairs — VisualGLM-6B is designed for image understanding, description, and question answering. Fine-tuning on long visual QA datasets further aligns the model’s responses with human preferences. The repository provides inference APIs, command-line demos, web demos, and efficient fine-tuning options like LoRA, QLoRA, and P-tuning. It also supports quantization down to INT4, enabling local deployment on consumer GPUs with as little as 6.3 GB VRAM.