Chimera418/protein-ssp

update: added python version as a string in Readme

update: added python version in Readme

fix: better package versions

chore: enforce python 3.10 and pin exact dependency versions for pca model compatibility

fix: add robust fallback polyfill for numpy scalar in numpy 2.x

fix: UI state refresh on download & numpy scalar PCA unpickle error

fix: pin transformers<5.0.0 — 5.x broke GenerationMixin/T5EncoderModel lazy loading

ui: cleaner CSS + remove Model Comparison table from sidebar

fix: comprehensive numpy._core.* + numpy.core.* pickle shim for numpy 2.x

fix: st.image() use_column_width for older Streamlit compat

fix: numpy 2.x pickle compat + pin numpy<2 in requirements

fix: use cache-then-copy download strategy (no symlinks) + expose real exception on pickle failures

fix: use local_dir_use_symlinks=False in hf_hub_download — prevents broken symlink crash on HF Space (pickle.load fails on symlinks in Docker)

fix: download model from HF Hub before existence check — fixes 'Model not trained yet' error on cold HF Space

docs: add full pipeline from scratch section — run phase scripts to generate own models

docs: comprehensive README rewrite — correct URLs, 5-mode metrics table, pipeline phase descriptions, local setup guide

docs: remove internal misc artifact upload instructions from public README

docs: add instructions for data upload via scripts and powershell warning

chore: ignore local helper upload scripts

Delete Raw Outputs

Simplify downloading from models/ and embeddings/ folders

Update Space URL to protein-ssp

Update Hub repo name to protein-ssp-artifacts

Update Hub repo name to prot-ssp-artifacts

Update README with live deployment link and architecture details

Deploy lightweight app configured for HF Model Hub