PyTorch

1. Adoption: PyTorch has become a dominant framework, evidenced by Shopify joining the PyTorch Foundation, its inclusion in Microsoft’s AI curriculum, and active community events. 2. Core function: Provides tensor operations and automatic differentiation for tasks such as computer vision and NLP, scaling to large models via distributed training. 3. Barrier and cost: Completely open-source and free, requiring Python and ML expertise; not suitable for no-code users. Costs are limited to compute resources. 4. Community consensus and evidence quality: Evidence leans heavily on Zhihu tutorials (beginner guides, training system breakdowns), official announcements (Shopify partnership, Miles framework), and high-engagement X posts. Useful proof primarily comes from official sources and practical tutorials, but in-depth comparative reviews remain limited. Not a model library or API service; it is comparable to TensorFlow as a foundational framework.