Remove final model artifact from repo

Merge remote-tracking branch 'origin/main'

Add timestep-aware sparse KD weighting

Upload final Stage 3B trained ViL-DLM model

Filter unusable Cauldron image payloads

Force cached KD masks during Stage 3b

Guarantee masked assistant tokens in diffusion training

Fix HF Sequence type alias in Stage 2 loader

Make Stage 2 Cauldron loading cost-safe

Fix The Cauldron aokvqa config name

Filter invalid LLaVA image records for real runs

Fix LLaVA fallback image loading in Stage 1

Implement stage-aware real-run training pipeline

Fix VRAM property access in training script to use total_memory

Update dependencies in pyproject.toml to specify versions for torch and torchvision, and add source configuration for UV tool

Add Trackio integration for enhanced logging in training process

Enhance LLaVAPretrainDataset to support fallback loading from JSON and improve image path resolution

Add dropout parameter to ProjConfig for improved model regularization

Add Vision Transformer and utility functions for sequence processing

Update model configuration and training scripts with new vision backbone support and dependencies

Add comprehensive README with architecture details and training recipe

Upload model_config.py

Upload vil_dlm_model.py

Upload vision_xlstm.py

Upload train_production.py

Upload production training script

initial commit