Non-Pythonic code requirements and boilerplate overhead

TensorFlow forces non-idiomatic Python patterns, requiring session handlers and TensorFlow-specific equivalents for basic operations like loops. This creates verbose, un-Pythonic code and makes the framework feel like a language within a language.

Sources

• Pros And Cons Of Using TensorFlow In The Production Environment
• The Decline of a Framework - The Probability and the Word

Collection History