Yesterday
-
docs: add GGUF at-a-glance (q4 593 KB, 400/400 parity; f16 1.99 MB, 399/400) above Benchmarks
-
docs: drop internal benchmark path from the GGUF verification note
-
fix: ship bhc/casual.py (translit2.py imports it; the python/ tree could not be imported)
-
fix: ship bhc/skel.py (translit2.py imports it; the python/ tree could not be imported)
-
README: reader byte size after the CLI fix
-
gguf_read.py: accept --json before the file argument
-
README: document the two GGUF files, the reader and the parity results
-
add gguf_read.py (GGUF container for the bundled reader)
-
add aksharalp-1m-v2-f16.gguf (GGUF container for the bundled reader)
-
add aksharalp-1m-v2-q4.gguf (GGUF container for the bundled reader)
4 days ago
-
Improve README: versions, metadata, wording
-
Card/Space: real-world comparison 'Boht ache h Bhai' (factual side-by-side, measured 2026-11-01)
-
Card: headline size wording "under 8 MB" (exact 7.92 MB kept in tables); banner + screenshots regenerated
5 days ago
-
AksharaLP-1M-v2 v2.0.0
-
AksharaLP-1M-v2 v2.0.0
-
initial commit
