jinaai/jina-clip-implementation

fix for latest hf trasformer pretrained model initializer (#11)

fix: encode_text latency (#9)

Fix docs (#8)

feat: disable flash attn if not supported CUDA version or device capability

fix: throw warnings if xformers or flash-attn cant be used

fix: throw warnings if xformers or flash-attn cant be used

fix: throw warnings if xformers or flash-attn cant be used

fix: throw warnings if xformers or flash-attn cant be used

fix: throw warnings if xformers or flash-attn cant be used

fix: truncate embeddings

fix: suppress timm import warning

feat: remove adapter_mask from interface

docs: update README

fix: various fixes

docs: update README

feat: jina clip v2 implementation

feat-matryoshka-embeddings (#6)

feat: add autocasting in vision.patch_embed

feat: customize image processor in __call__

fix: processor class types

fix: processor class types

Update modeling_clip.py (#5)

Update configuration_clip.py (#4)

Update hf_model.py (#3)

support use_flash_attn in from_pretrained (#2)

fix: do all relative imports at the top level

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

fix: encode_image and encode_text

style: apply ruff & isort

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

Update modeling_clip.py

feat-inference-mode (#1)

feat: initial commit

initial commit