Rename text_encoder.encoder.embed_tokens.weight -> text_encoder.shared.weight so AutoWeightsLoader can match the HF T5 shared-embedding alias without a remap.

Upload transformer/diffusion_pytorch_model.safetensors with huggingface_hub

Upload transformer/diffusion_pytorch_model.safetensors with huggingface_hub

Upload folder using huggingface_hub

Repack weights: transformer + vae safetensors (Diffusers layout); drop legacy conditioners shard

Remove redundant weights_format from model_index.json; re-sync safetensors bundle

Canonical weights: offline remap via vLLM-Omni audiox_weights (weights_format=canonical)

Add Diffusers vae layout: config.json + encoder/decoder safetensors for AutoencoderOobleck.from_pretrained

AudioX vLLM-Omni sharded safetensors

initial commit