fix: replace colbert-small-v1 model.onnx with 96-dim projected export (was 384-dim base encoder from upstream onnx/ subdir)

chore: sync manifest with full model catalog

docs: refresh attribution for full model catalog

Add GTE-ModernColBERT-v1 ONNX export (ModernBERT + Dense 768->128, per-token 128-d, opset 17)

Upload gte-moderncolbert-v1/1_Dense/config.json with huggingface_hub

Upload gte-moderncolbert-v1/onnx_config.json with huggingface_hub

Upload gte-moderncolbert-v1/tokenizer_config.json with huggingface_hub

Upload gte-moderncolbert-v1/special_tokens_map.json with huggingface_hub

Upload gte-moderncolbert-v1/config.json with huggingface_hub

Upload gte-moderncolbert-v1/tokenizer.json with huggingface_hub

Upload gte-moderncolbert-v1/model.onnx with huggingface_hub

Upload README.md with huggingface_hub

Upload manifest.json with huggingface_hub

Upload colbert-small-v1/tokenizer_config.json with huggingface_hub

Upload colbert-small-v1/special_tokens_map.json with huggingface_hub

Upload colbert-small-v1/config.json with huggingface_hub

Upload colbert-small-v1/tokenizer.json with huggingface_hub

Upload colbert-small-v1/model.onnx with huggingface_hub

initial commit