Make Tensor python bindings export bool values in protocol buffers interface

Make Tensor python bindings export bool values in protocol buffers interface

Add artifacts files to ignore list

Yolo11: add class-balanced COCO subset builder

Yolo11 add interactive loss-curve viewer for CSV logs

Yolo11: add class-weighted loss and CSV logging

Fix accumulate input gradients shared across multiple consumers

Fix vulkan align push constant packing with std430 layout rules

Fix how cpu path handle non-contiguous last dimension in row loops

Fix timeout for VkSubmit wait to avoid indefinite hang

Fix yolo11 inference: load typed checkpoints with training metadata in inference

Update yolo training info display

Refactor serialization: store checkpoints as tensor dictionaries

Update yolo11training to make possible start from saved checkpoints

Add typed checkpoints and optimizer state dictionaries

Yolo11 C++ inference application

Add Sequential and ModuleList container modules for C++

Update vulkan buffer creation to return nullopt from find_memory_type on failure

Update git ignore

Add add balanced synthetic dataset generator for yolo11 training

Improve yolo loss performance: replace pred_dist dot with broadcast sum in yolo11 loss

Update yolo11loss: cache anchors and remove host syncs in loss

Refactor Vulkan copy to host with use of dedicated fence pool for host-read submissions

Fix guard strided reductions against uneven workgroups in Vulkan shaders

Add direct grouped convolution kernels

Refactor TensorImpls: remove scatter index dimension bounds checks

Fix Tensor: ensure contiguous host tensor before GPU transfer

Add broadcasting support to eq/ne Vulkan ops

yolov11 update: scale detections back to original image coordinates in inference script

yolo11 update: drop bbox params from dataset augmentation

Copy adept soname and linker files with python binding

Add missed include

Fix wheel packaging make bundled libraries relocatable with $ORIGIN

Update yolo11: add warmup and cosine LR scheduling

Fix vulkan conv2d bias gradient sum out of batch loop

Update cmake script

Fix broadcast: allow in-place ops to resize smaller left operand

Add momentum and weight decay to SGD optimizer

Fix weights saving: propagate prefix to nested module save/load

Initialize batchnorm running variance to one

Update memory management : align allocations and enlarge memory pool cache

Make shape accept -1 dimension when element counts match

Yolo11 training pipeline fixes