Downloads · 30 days
74
19% of all-time downloads
soundsgoodai/Zipformer-cr-ctc-transducer-XL-290M
Zipformer-cr-ctc-transducer-XL-290M is a automatic speech recognition model from soundsgoodai. Use it when you need speech turned into text. It is set up for k2. The card lists the license as cc-by-nc-4.0.
Offline English ASR model based on the Icefall/K2 pruned transducer Zipformer-CR-CTC recipe. The model was trained with a transducer objective and CR-CTC (Consistency-Regularized CTC) auxiliary loss, and supports Icef…
Downloads · 30 days
74
19% of all-time downloads
All-time downloads
387
Public
Repo size
1.2 GB
Likes
5
Public
Click a slice to open those files.
.pt1.2 GB · 100%
From the Hugging Face model README
Offline English ASR model based on the Icefall/K2 pruned transducer Zipformer-CR-CTC recipe. The model was trained with a transducer objective and CR-CTC (Consistency-Regularized CTC) auxiliary loss, and supports Icefall transducer and CTC decoding.
model.pt: Zipformer CR-CTC transducer checkpointbpe.model: SentencePiece BPE modeltokens.txt: Icefall token table exported from bpe.modelconfig.yaml: model architecture, feature extraction, tokenizer, Zipformer
model, decoding settings, and Hugging Face Hub download-metrics query fileOpen ASR Leaderboard English short-form result, decoded with
modified_beam_search and beam size 6:
| Metric | Value |
|---|---|
| Average WER | 5.56 |
| Parameters | 288M |
Dataset WERs:
| Dataset | WER |
|---|---|
| AMI | 11.30 |
| Earnings22 | 7.64 |
| GigaSpeech | 8.41 |
| LibriSpeech Clean | 1.31 |
| LibriSpeech Other | 3.01 |
| SPGISpeech | 1.64 |
| VoxPopuli | 5.60 |
| Average | 5.56 |
The model was trained on a combined English training mixture built from the training portions of the datasets below.
| Dataset | Train Hours | Source |
|---|---|---|
| LibriSpeech | 960.0 | ESB datasets |
| Earnings-22 | 105.0 | ESB datasets |
| AMI Meeting Corpus | 78.0 | ESB datasets |
| Common Voice Scripted Speech 25.0 - English | ~1,679.0 | Mozilla Data Collective |
| Common Voice Spontaneous Speech 3.0 - English | ~3.6 | Mozilla Data Collective |
| GigaSpeech XL | 10,000.0 | ESB datasets |
| SPGISpeech | 4,900.0 | ESB datasets |
| TED-LIUM Release 3 | 454.0 | ESB datasets |
| VoxPopuli | 523.0 | ESB datasets |
| Total | ~18,700 | ESB datasets and Mozilla Data Collective |
Common Voice Scripted Speech 25.0 - English and Common Voice Spontaneous Speech 3.0 - English are the English scripted and spontaneous Common Voice releases checked on May 12, 2026.
For data normalization, the training transcripts were processed with direct LLM normalization using a self-hosted GLM-4.7 setup, plus custom agentic workflows built in collaboration with Claude Code 4.7 and Codex 5.4.
The model was trained with the k2-fsa/Icefall framework, PyTorch 2.11, and CUDA 13.0 on 8 NVIDIA B300 GPUs using bf16 automatic mixed precision.
From an Icefall checkout, download this model repo locally:
huggingface-cli download soundsgoodai/Zipformer-cr-ctc-transducer-XL-290M \
--local-dir Zipformer-cr-ctc-transducer-XL-290M
Run offline decoding through Icefall. Input audio must already be 16 kHz:
cd icefall/egs/librispeech/ASR
PYTHONPATH=../../.. python zipformer/pretrained.py \
--checkpoint /path/to/Zipformer-cr-ctc-transducer-XL-290M/model.pt \
--tokens /path/to/Zipformer-cr-ctc-transducer-XL-290M/tokens.txt \
--method modified_beam_search \
--beam-size 6 \
--num-encoder-layers "2,2,4,5,4,2" \
--downsampling-factor "1,2,4,8,4,2" \
--feedforward-dim "512,1024,2048,3072,2048,1024" \
--num-heads "4,4,4,8,4,4" \
--encoder-dim "192,384,768,1024,768,384" \
--encoder-unmasked-dim "192,192,320,384,320,192" \
--query-head-dim "32" \
--value-head-dim "12" \
--pos-head-dim "4" \
--pos-dim 48 \
--cnn-module-kernel "31,31,15,15,15,31" \
--decoder-dim 512 \
--joiner-dim 512 \
--context-size 2 \
--causal false \
--chunk-size "16,32,64,-1" \
--left-context-frames "64,128,256,-1" \
--use-transducer true \
--use-ctc true \
--use-attention-decoder false \
--use-cr-ctc true \
/path/to/audio_1.wav \
/path/to/audio_2.wav
For CTC decoding, use Icefall's CTC entrypoint with the same architecture settings:
cd icefall/egs/librispeech/ASR
PYTHONPATH=../../.. python zipformer/pretrained_ctc.py \
--checkpoint /path/to/Zipformer-cr-ctc-transducer-XL-290M/model.pt \
--tokens /path/to/Zipformer-cr-ctc-transducer-XL-290M/tokens.txt \
--method ctc-decoding \
--sample-rate 16000 \
--num-encoder-layers "2,2,4,5,4,2" \
--downsampling-factor "1,2,4,8,4,2" \
--feedforward-dim "512,1024,2048,3072,2048,1024" \
--num-heads "4,4,4,8,4,4" \
--encoder-dim "192,384,768,1024,768,384" \
--encoder-unmasked-dim "192,192,320,384,320,192" \
--query-head-dim "32" \
--value-head-dim "12" \
--pos-head-dim "4" \
--pos-dim 48 \
--cnn-module-kernel "31,31,15,15,15,31" \
--decoder-dim 512 \
--joiner-dim 512 \
--context-size 2 \
--causal false \
--chunk-size "16,32,64,-1" \
--left-context-frames "64,128,256,-1" \
--use-transducer true \
--use-ctc true \
--use-attention-decoder false \
--use-cr-ctc true \
/path/to/audio_1.wav \
/path/to/audio_2.wav
The architecture and decoding values above are also recorded in config.yaml.
The model supports the following Icefall decoding methods:
Transducer:
greedy_searchmodified_beam_searchfast_beam_searchCTC:
ctc-decodingThe reported Open ASR Leaderboard result uses modified_beam_search with beam
size 6. CTC decoding is supported, but the table above does not report CTC WERs.
For this model, use Kaldi-style fbank features. kaldifeat and
kaldi-native-fbank are the recommended feature extraction backends.
Audio should be mono 16 kHz before feature extraction. For sample-rate
conversion, audioop.ratecv is recommended because it matches the resampling
path used for evaluation. On Python 3.13 and newer, use an audioop-compatible
package such as audioop-lts.
The model emits normalized English text with punctuation and capitalization. It does not automatically capitalize the first word of every sentence unless that word is normally capitalized, such as a proper noun, honorific, acronym, or similar named expression.
The model also normalizes common written forms, including numbers, dates, and currency, to digit-based forms.
The open_asr_leaderboard/soundsgoodai runner downloads this model repo,
resamples audio to 16 kHz with audioop, computes fbank features with
kaldi-native-fbank, and decodes with Icefall modified_beam_search.
cd open_asr_leaderboard/soundsgoodai
bash run_zipformer.sh
modified_beam_search with beam size 6.--use-ctc true and --use-cr-ctc true.audioop.ratecv: https://docs.python.org/3.11/library/audioop.html#audioop.ratecv