8 days ago
-
Remove backend experiment notes from the published model repository
-
Document standard Vulkan-first Android model selection
-
Keep only the Vulkan FP16 variant
-
Group backend precision notes under a shared heading
-
Add device-validated Vulkan FP16 and FP32 variants
-
Add validated selective XNNPACK INT8 variant
-
Remove verification checksum manifest from the published model repository
-
Document rejected XNNPACK quantization candidates and profiler results
-
Align model card with the established ExecuTorch model-card format
-
Add U2NetP foreground segmentation exports for XNNPACK and CoreML
-
initial commit
