Downloads · 30 days
0
IconicAI/sherpa-ncnn-zipformer-libri-giga-70M-2023-04-04
sherpa-ncnn-zipformer-libri-giga-70M-2023-04-04 is a automatic speech recognition model from IconicAI. Use it when you need speech turned into text. The card lists the license as apache-2.0.
This repo provides an NCNN export of the icefall streaming Zipformer model trained on LibriSpeech + GigaSpeech (2023-04-04). It is ready for sherpa-ncnn (Android and desktop).
Downloads · 30 days
0
Access
Public
Updated Aug 25, 2025
Repo size
141 MB
Likes
0
Public
Click a slice to open those files.
.bin141 MB · 100%
From the Hugging Face model README
This repo provides an NCNN export of the icefall streaming Zipformer
model trained on LibriSpeech + GigaSpeech (2023-04-04).
It is ready for sherpa-ncnn (Android and desktop).
Files included
encoder_jit_trace-pnnx.ncnn.paramencoder_jit_trace-pnnx.ncnn.bindecoder_jit_trace-pnnx.ncnn.paramdecoder_jit_trace-pnnx.ncnn.binjoiner_jit_trace-pnnx.ncnn.paramjoiner_jit_trace-pnnx.ncnn.bintokens.txt (500 tokens; blank id = 0)The encoder
.paramalready contains the required SherpaMetaData layer. Files are UTF-8 without BOM to match NCNN’s plain-text parser.
2,4,3,2,43841921,2,4,8,231512 / 512decode_chunk_len=32, num_left_chunks=4, pad_len=7csukuangfj/ncnn fork, MSVC build on Windows)SherpaMetaData into the encoder .paramCopy the six NCNN files and tokens.txt into your app’s assets/ (or load from files),
then construct the recognizer with those paths. Ensure your NCNN build includes ops
used by modern Zipformers (e.g., CumulativeSum). If Vulkan isn’t present, useGPU
falls back to CPU automatically.
If you use the sherpa-ncnn CLI, point each component explicitly (paths adjusted):
sherpa-ncnn \
--encoder-param encoder_jit_trace-pnnx.ncnn.param \
--encoder-bin encoder_jit_trace-pnnx.ncnn.bin \
--decoder-param decoder_jit_trace-pnnx.ncnn.param \
--decoder-bin decoder_jit_trace-pnnx.ncnn.bin \
--joiner-param joiner_jit_trace-pnnx.ncnn.param \
--joiner-bin joiner_jit_trace-pnnx.ncnn.bin \
--tokens tokens.txt \
--decoding-method greedy_search \
--num-threads 4