marconolimits/NMT

feat: handle HTTP requests tolerantly on TCP port & optimize threads for all cores

perf: translate multi-line input line-by-line; env beam/max-decode; robust env ints

fix: parse NMT_* int env vars with multiline/whitespace (HF Space UI)

fix: download CT2 model from HF Hub instead of COPY (avoids Space LFS quota)

fix: rewrite .dockerignore as UTF-8 (was UTF-16 LE, broke Docker BuildKit)

chore: sync binary gitignore from hf-space

deploy: clean orphan branch for HF Spaces - CPU threading optimisation