Upload folder using huggingface_hub

Fix standalone model loading code - verified to work

Make README fully standalone - no external repo needed

Add standalone loading option using only transformers

Fix usage instructions with proper installation steps

Fix total parameter count: ~120M not ~250M

Document both model.pt and model.safetensors formats

Remove redundant pytorch_model.bin (use model.pt or model.safetensors)

Update model.safetensors with Stage 5 audio-aligned weights

Fix loading example to match checkpoint format

Fix Whisper-tiny encoder param count (8M, not 39M)

Update model with audio-text alignment (Stage 5: R@1=36.38%)

Add verified dataset sources with paper citations

Add audio training roadmap to model card

Upload multi-modal-embed-small Stage 2 model

initial commit