stt-ka-fastconformer-hybrid-ctc-large-GGUF

GGUF conversions of the CTC branch of nvidia/stt_ka_fastconformer_hybrid_large_pc for CrispASR. The upstream model is a hybrid transducer+CTC Georgian ASR release; the shared FastConformer encoder plus the auxiliary CTC head are extracted here as a standalone CTC model (the RNNT prediction network and joint are dropped), giving a compact Georgian ASR and forced-alignment model with punctuation + capitalisation.

Quant	Size	Description
F16	219 MB	Full precision
Q8_0	130 MB	8-bit
Q4_K	82 MB	4-bit K-quant (recommended)

Architecture

17-layer NeMo FastConformer encoder + Conv1d CTC head. d_model=512, 8 heads, SentencePiece vocab, 80 log-mel features, ~115M params.

Usage

# Georgian ASR:
crispasr --backend fastconformer-ctc -m stt-ka-fastconformer-hybrid-ctc-large-q4_k.gguf -f audio.wav

# Forced alignment (word timestamps for known text, or re-timing an .srt):
crispasr --align-only -am stt-ka-fastconformer-hybrid-ctc-large-q4_k.gguf \
    -f audio.wav --text-file subtitles.srt --align-output retimed.srt

Attribution

All credit for the model goes to NVIDIA's NeMo team; this repository only repackages the CTC branch in GGUF form under the same CC-BY-4.0 license. Conversion: models/convert-stt-fastconformer-ctc-to-gguf.py in CrispASR.