Model Explorer Files

What's Changed

🚀 Highlights & Features

• Hugging Face Transformers Adapter (model-explorer-hfgraph): Added the Hugging Face Transformers adapter as a built-in adapter, including branch status colors (groupNodeConfigs) and support for keepLayersWithASingleChild (#752, #762, #763, #776).
• Quantization Support: Added blockwise and multi-axis quantization support in the LiteRT FlatBuffer converter (#748).
• Graph Picker UX: Users can now expand and collapse subgraphs directly in the graph picker (#739).
• Diagnostics & Progress Reporting: Introduced backend diagnostic collection, structured logging, JSON serialization, and ABI-safe C-API progress callbacks for graph conversions (#718, #726, #728).

🐛 Bug Fixes

• Fixed SIGSEGV crash when printing 0-D non-splat tensors (#740).
• Secured file handling by keeping uploaded model files confined to their temporary upload directory (#765).
• Corrected typos across adapter and metadata modules in the server (#764).

🔧 Build, CI & Infrastructure

• Workflow Improvements: Hardened GitHub Actions workflows against zizmor security findings (#692, #727).
• Adapter CI/CD: Configured self-hosted Linux ARC job containers, resolved build failures, and improved Bazel caching / C-ABI packaging (#689, #720, #721).
• Dialect & Dependencies: Synchronized OpDefs via TableGen AST with StableHLO dialect tags (#725), updated TensorFlow pinned dependencies (#770), and removed resolve_symbols_in_exec (#724).
• Version Bumps: Bumped Model Explorer backend version to 0.2.1 (#769).

📖 Documentation

• Updated Model Explorer backend documentation to adopt LiteRT naming conventions (#719).

Full Changelog: model-explorer-v0.1.33...model-explorer-v0.1.34