Add psutil dependency required by fast_pytorch_kmeans

Load encoder directly from its own repo

Revert "Fix extract_features_value_attention and add _get_all_blocks"

Fix extract_features_value_attention and add _get_all_blocks

Update app.py

Fix indentation in load_variant (token + from_pretrained) (#1)

Update app.py

Fix: Use HF_TOKEN to access private models

TIPSv2 Feature Explorer

initial commit