cstr/opus-mt-tc-big-he-en-GGUF

Opus-MT Hebrew → English — GGUF (ggml)

GGUF / ggml conversion of Helsinki-NLP/opus-mt-tc-big-he-en for use with CrispStrobe/CrispASR (--backend marian).

A larger (~245M parameters, 6+6 layers, d=1024) MarianMT model for one language pair. In CrispASR these are the fastest translators for live transcription + translation (--live-translate).

Licence and attribution

CC-BY-4.0. The weights are the opusTCv20210807+bt_transformer-big_2022-03-13 release of the OPUS-MT project (Jörg Tiedemann and Santhosh Thottingal, University of Helsinki; OPUS-MT — Building open translation services for the World, EAMT 2020), trained on OPUS data. The project distributes its pre-trained models under CC-BY 4.0 (Opus-MT README, OPUS-MT-train README); that statement, not the tag on an individual model card, is what this repository follows. This is a format conversion: the weights are unchanged at f16 and quantized at q8_0. Redistribution requires attribution to the OPUS-MT project.

Files

File	Size	Output vs. the reference implementation
opus-mt-tc-big-he-en-f16.gguf	488 MB	8/8 sentences identical greedy, 8/8 with beam 4
opus-mt-tc-big-he-en-q8_0.gguf	265 MB	8/8 sentences identical greedy — any others differ in wording. Fastest; recommended.

"Reference" is MarianMTModel.generate from Hugging Face transformers on the original checkpoint, on 8 test sentences; input token ids were identical for 8/8.

Quick start

git clone https://github.com/CrispStrobe/CrispASR && cd CrispASR
cmake -B build -DCMAKE_BUILD_TYPE=Release && cmake --build build -j

# Text → text. -bs 1 is greedy; without it the checkpoint's own beam size is used.
./build/bin/crispasr --backend marian -m opus-mt-he-en -sl he -tl en -bs 1 \
    --text "בוקר טוב וברוכים הבאים לישיבה של היום."
# Good morning and welcome to today's session.

# Live: microphone in, transcript + translation out, sentence by sentence
./build/bin/crispasr --live-translate -l he --tr-tl en \
    -m auto --backend parakeet --translate-backend marian

-m opus-mt-he-en downloads opus-mt-tc-big-he-en-q8_0.gguf on first use. The recogniser in the live example must support Hebrew.

Notes

• One model per direction; the opposite direction, where one exists, is cstr/opus-mt-en-he-GGUF.
• Live mode always decodes greedy.
• Literal </s>, <unk>, <pad> in the input are treated as ordinary text here (the reference treats them as special tokens).

Conversion

python models/convert-marian-to-gguf.py --input <dir>/opus-mt-tc-big-he-en --output opus-mt-tc-big-he-en-f16.gguf
./build/bin/crispasr-quantize opus-mt-tc-big-he-en-f16.gguf opus-mt-tc-big-he-en-q8_0.gguf q8_0
python tools/marian_parity.py --hf-dir <dir>/opus-mt-tc-big-he-en --gguf opus-mt-tc-big-he-en-f16.gguf --lang en --tgt en --sentences <file>