Add Zipformer recipe for GigaSpeech (#1254)

Co-authored-by: Yifan Yang <yifanyeung@qq.com> Co-authored-by: yfy62 <yfy62@d3-hpc-sjtu-test-005.cm.cluster>
2025-08-08 09:32:20 +00:00 · 2023-10-21 02:36:59 -05:00 · 2023-10-21 02:36:59 -05:00 · 416852e8a1
commit 416852e8a1
parent eef47adee9
43 changed files with 6036 additions and 2 deletions
--- a/.github/scripts/run-gigaspeech-zipformer-2023-10-17.sh
+++ b/.github/scripts/run-gigaspeech-zipformer-2023-10-17.sh
@ -0,0 +1,94 @@
+#!/usr/bin/env bash
+
+set -e
+
+log() {
+  # This function is from espnet
+  local fname=${BASH_SOURCE[1]##*/}
+  echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
+}
+
+cd egs/gigaspeech/ASR
+
+repo_url=https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
+
+log "Downloading pre-trained model from $repo_url"
+git lfs install
+GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
+repo=$(basename $repo_url)
+
+log "Display test files"
+tree $repo/
+ls -lh $repo/test_wavs/*.wav
+
+pushd $repo/exp
+git lfs pull --include "data/lang_bpe_500/bpe.model"
+git lfs pull --include "data/lang_bpe_500/tokens.txt"
+git lfs pull --include "exp/jit_script.pt"
+git lfs pull --include "exp/pretrained.pt"
+ln -s pretrained.pt epoch-99.pt
+ls -lh *.pt
+popd
+
+log "Export to torchscript model"
+./zipformer/export.py \
+  --exp-dir $repo/exp \
+  --use-averaged-model false \
+  --tokens $repo/data/lang_bpe_500/tokens.txt \
+  --epoch 99 \
+  --avg 1 \
+  --jit 1
+
+ls -lh $repo/exp/*.pt
+
+log "Decode with models exported by torch.jit.script()"
+
+./zipformer/jit_pretrained.py \
+  --tokens $repo/data/lang_bpe_500/tokens.txt \
+  --nn-model-filename $repo/exp/jit_script.pt \
+  $repo/test_wavs/1089-134686-0001.wav \
+  $repo/test_wavs/1221-135766-0001.wav \
+  $repo/test_wavs/1221-135766-0002.wav
+
+for method in greedy_search modified_beam_search fast_beam_search; do
+  log "$method"
+
+  ./zipformer/pretrained.py \
+    --method $method \
+    --beam-size 4 \
+    --checkpoint $repo/exp/pretrained.pt \
+    --tokens $repo/data/lang_bpe_500/tokens.txt \
+    $repo/test_wavs/1089-134686-0001.wav \
+    $repo/test_wavs/1221-135766-0001.wav \
+    $repo/test_wavs/1221-135766-0002.wav
+done
+
+echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
+echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
+if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode"  ]]; then
+  mkdir -p zipformer/exp
+  ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
+  ln -s $PWD/$repo/data/lang_bpe_500 data/
+
+  ls -lh data
+  ls -lh zipformer/exp
+
+  log "Decoding test-clean and test-other"
+
+  # use a small value for decoding with CPU
+  max_duration=100
+
+  for method in greedy_search fast_beam_search modified_beam_search; do
+    log "Decoding with $method"
+
+    ./zipformer/decode.py \
+      --decoding-method $method \
+      --epoch 999 \
+      --avg 1 \
+      --use-averaged-model 0 \
+      --max-duration $max_duration \
+      --exp-dir zipformer/exp
+  done
+
+  rm zipformer/exp/*.pt
+fi
--- a/.github/workflows/run-gigaspeech-zipformer-2023-10-17.yml
+++ b/.github/workflows/run-gigaspeech-zipformer-2023-10-17.yml
@ -0,0 +1,126 @@
+# Copyright      2022  Fangjun Kuang (csukuangfj@gmail.com)
+
+# See ../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+name: run-gigaspeech-zipformer-2023-10-17
+# zipformer
+
+on:
+  push:
+    branches:
+      - master
+  pull_request:
+    types: [labeled]
+
+  schedule:
+    # minute (0-59)
+    # hour (0-23)
+    # day of the month (1-31)
+    # month (1-12)
+    # day of the week (0-6)
+    # nightly build at 15:50 UTC time every day
+    - cron: "50 15 * * *"
+
+concurrency:
+  group: run_gigaspeech_2023_10_17_zipformer-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  run_gigaspeech_2023_10_17_zipformer:
+    if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
+    runs-on: ${{ matrix.os }}
+    strategy:
+      matrix:
+        os: [ubuntu-latest]
+        python-version: [3.8]
+
+      fail-fast: false
+
+    steps:
+      - uses: actions/checkout@v2
+        with:
+          fetch-depth: 0
+
+      - name: Setup Python ${{ matrix.python-version }}
+        uses: actions/setup-python@v2
+        with:
+          python-version: ${{ matrix.python-version }}
+          cache: 'pip'
+          cache-dependency-path: '**/requirements-ci.txt'
+
+      - name: Install Python dependencies
+        run: |
+          grep -v '^#' ./requirements-ci.txt  | xargs -n 1 -L 1 pip install
+          pip uninstall -y protobuf
+          pip install --no-binary protobuf protobuf==3.20.*
+
+      - name: Cache kaldifeat
+        id: my-cache
+        uses: actions/cache@v2
+        with:
+          path: |
+            ~/tmp/kaldifeat
+          key: cache-tmp-${{ matrix.python-version }}-2023-05-22
+
+      - name: Install kaldifeat
+        if: steps.my-cache.outputs.cache-hit != 'true'
+        shell: bash
+        run: |
+          .github/scripts/install-kaldifeat.sh
+
+      - name: Inference with pre-trained model
+        shell: bash
+        env:
+          GITHUB_EVENT_NAME: ${{ github.event_name }}
+          GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
+        run: |
+          mkdir -p egs/gigaspeech/ASR/data
+          ln -sfv ~/tmp/fbank-libri egs/gigaspeech/ASR/data/fbank
+          ls -lh egs/gigaspeech/ASR/data/*
+
+          sudo apt-get -qq install git-lfs tree
+          export PYTHONPATH=$PWD:$PYTHONPATH
+          export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
+          export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
+
+          .github/scripts/run-gigaspeech-zipformer-2023-10-17.sh
+
+      - name: Display decoding results for gigaspeech zipformer
+        if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
+        shell: bash
+        run: |
+          cd egs/gigaspeech/ASR/
+          tree ./zipformer/exp
+
+          cd zipformer
+          echo "results for zipformer"
+          echo "===greedy search==="
+          find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
+          find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
+
+          echo "===fast_beam_search==="
+          find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
+          find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
+
+          echo "===modified beam search==="
+          find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
+          find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
+
+      - name: Upload decoding results for gigaspeech zipformer
+        uses: actions/upload-artifact@v2
+        if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
+        with:
+          name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
+          path: egs/gigaspeech/ASR/zipformer/exp/
--- a/README.md
+++ b/README.md
@ -148,8 +148,11 @@ in the decoding.

 ### GigaSpeech

-We provide two models for this recipe: [Conformer CTC model][GigaSpeech_conformer_ctc]
-and [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][GigaSpeech_pruned_transducer_stateless2].
+We provide three models for this recipe:
+
+- [Conformer CTC model][GigaSpeech_conformer_ctc]
+- [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][GigaSpeech_pruned_transducer_stateless2].
+- [Transducer: Zipformer encoder + Embedding decoder][GigaSpeech_zipformer]

 #### Conformer CTC

@ -165,6 +168,14 @@ and [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned R
 |   fast beam search   | 10.50 | 10.69 |
 | modified beam search | 10.40 | 10.51 |

+#### Transducer: Zipformer encoder + Embedding decoder
+
+|                      |  Dev  | Test  |
+|----------------------|-------|-------|
+|    greedy search     | 10.31 | 10.50 |
+|   fast beam search   | 10.26 | 10.48 |
+| modified beam search | 10.25 | 10.38 |
+

 ### Aishell

@ -378,6 +389,7 @@ Please see: [![Open In Colab](https://colab.research.google.com/assets/colab-bad
 [TED-LIUM3_pruned_transducer_stateless]: egs/tedlium3/ASR/pruned_transducer_stateless
 [GigaSpeech_conformer_ctc]: egs/gigaspeech/ASR/conformer_ctc
 [GigaSpeech_pruned_transducer_stateless2]: egs/gigaspeech/ASR/pruned_transducer_stateless2
+[GigaSpeech_zipformer]: egs/gigaspeech/ASR/zipformer
 [Aidatatang_200zh_pruned_transducer_stateless2]: egs/aidatatang_200zh/ASR/pruned_transducer_stateless2
 [WenetSpeech_pruned_transducer_stateless2]: egs/wenetspeech/ASR/pruned_transducer_stateless2
 [WenetSpeech_pruned_transducer_stateless5]: egs/wenetspeech/ASR/pruned_transducer_stateless5
--- a/egs/gigaspeech/ASR/README.md
+++ b/egs/gigaspeech/ASR/README.md
@ -15,6 +15,7 @@ ln -sfv /path/to/GigaSpeech download/GigaSpeech
 ## Performance Record
 |                                |  Dev  | Test  |
 |--------------------------------|-------|-------|
+|           `zipformer`          | 10.25 | 10.38 |
 |         `conformer_ctc`        | 10.47 | 10.58 |
 | `pruned_transducer_stateless2` | 10.40 | 10.51 |

--- a/egs/gigaspeech/ASR/RESULTS.md
+++ b/egs/gigaspeech/ASR/RESULTS.md
@ -1,4 +1,78 @@
 ## Results
+### zipformer (zipformer + pruned stateless transducer)
+
+See <https://github.com/k2-fsa/icefall/pull/1254> for more details.
+
+[zipformer](./zipformer)
+
+- Non-streaming
+- normal-scaled model, number of model parameters: 65549011, i.e., 65.55 M
+
+You can find a pretrained model, training logs, decoding logs, and decoding results at:
+<https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17>
+
+The tensorboard log for training is available at
+<https://wandb.ai/yifanyeung/icefall-asr-gigaspeech-zipformer-2023-10-20>
+
+You can use <https://github.com/k2-fsa/sherpa> to deploy it.
+
+| decoding method      | test-clean | test-other | comment            |
+|----------------------|------------|------------|--------------------|
+| greedy_search        | 10.31      | 10.50      | --epoch 30 --avg 9 |
+| modified_beam_search | 10.25      | 10.38      | --epoch 30 --avg 9 |
+| fast_beam_search     | 10.26      | 10.48      | --epoch 30 --avg 9 |
+
+The training command is:
+```bash
+export CUDA_VISIBLE_DEVICES="0,1,2,3"
+./zipformer/train.py \
+  --world-size 4 \
+  --num-epochs 30 \
+  --start-epoch 1 \
+  --use-fp16 1 \
+  --exp-dir zipformer/exp \
+  --causal 0 \
+  --subset XL \
+  --max-duration 700 \
+  --use-transducer 1 \
+  --use-ctc 0 \
+  --lr-epochs 1 \
+  --master-port 12345
+```
+
+The decoding command is:
+```bash
+export CUDA_VISIBLE_DEVICES=0
+
+# greedy search
+./zipformer/decode.py \
+  --epoch 30 \
+  --avg 9 \
+  --exp-dir ./zipformer/exp \
+  --max-duration 1000 \
+  --decoding-method greedy_search
+
+# modified beam search
+./zipformer/decode.py \
+  --epoch 30 \
+  --avg 9 \
+  --exp-dir ./zipformer/exp \
+  --max-duration 1000 \
+  --decoding-method modified_beam_search \
+  --beam-size 4
+
+# fast beam search (one best)
+./zipformer/decode.py \
+  --epoch 30 \
+  --avg 9 \
+  --exp-dir ./zipformer/exp \
+  --max-duration 1000 \
+  --decoding-method fast_beam_search \
+  --beam 20.0 \
+  --max-contexts 8 \
+  --max-states 64
+```
+
 ### GigaSpeech BPE training results (Pruned Transducer 2)

 #### 2022-05-12
--- a/egs/gigaspeech/ASR/zipformer/asr_datamodule.py
+++ b/egs/gigaspeech/ASR/zipformer/asr_datamodule.py
@ -0,0 +1,444 @@
+# Copyright      2021  Piotr Żelasko
+# Copyright      2023  Xiaomi Corporation     (Author: Yifan Yang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+import argparse
+import glob
+import inspect
+import logging
+import re
+from functools import lru_cache
+from pathlib import Path
+from typing import Any, Dict, Optional
+
+import lhotse
+import torch
+from lhotse import CutSet, Fbank, FbankConfig, load_manifest, load_manifest_lazy
+from lhotse.dataset import (
+    CutConcatenate,
+    CutMix,
+    DynamicBucketingSampler,
+    K2SpeechRecognitionDataset,
+    PrecomputedFeatures,
+    SimpleCutSampler,
+    SpecAugment,
+)
+from lhotse.dataset.input_strategies import AudioSamples, OnTheFlyFeatures
+from lhotse.utils import fix_random_seed
+from torch.utils.data import DataLoader
+
+from icefall.utils import str2bool
+
+
+class _SeedWorkers:
+    def __init__(self, seed: int):
+        self.seed = seed
+
+    def __call__(self, worker_id: int):
+        fix_random_seed(self.seed + worker_id)
+
+
+class GigaSpeechAsrDataModule:
+    """
+    DataModule for k2 ASR experiments.
+    It assumes there is always one train and valid dataloader,
+    but there can be multiple test dataloaders (e.g. LibriSpeech test-clean
+    and test-other).
+
+    It contains all the common data pipeline modules used in ASR
+    experiments, e.g.:
+    - dynamic batch size,
+    - bucketing samplers,
+    - cut concatenation,
+    - augmentation,
+    - on-the-fly feature extraction
+
+    This class should be derived for specific corpora used in ASR tasks.
+    """
+
+    def __init__(self, args: argparse.Namespace):
+        self.args = args
+
+    @classmethod
+    def add_arguments(cls, parser: argparse.ArgumentParser):
+        group = parser.add_argument_group(
+            title="ASR data related options",
+            description="These options are used for the preparation of "
+            "PyTorch DataLoaders from Lhotse CutSet's -- they control the "
+            "effective batch sizes, sampling strategies, applied data "
+            "augmentations, etc.",
+        )
+        group.add_argument(
+            "--manifest-dir",
+            type=Path,
+            default=Path("data/fbank"),
+            help="Path to directory with train/valid/test cuts.",
+        )
+        group.add_argument(
+            "--max-duration",
+            type=int,
+            default=200.0,
+            help="Maximum pooled recordings duration (seconds) in a "
+            "single batch. You can reduce it if it causes CUDA OOM.",
+        )
+        group.add_argument(
+            "--bucketing-sampler",
+            type=str2bool,
+            default=True,
+            help="When enabled, the batches will come from buckets of "
+            "similar duration (saves padding frames).",
+        )
+        group.add_argument(
+            "--num-buckets",
+            type=int,
+            default=30,
+            help="The number of buckets for the DynamicBucketingSampler"
+            "(you might want to increase it for larger datasets).",
+        )
+        group.add_argument(
+            "--concatenate-cuts",
+            type=str2bool,
+            default=False,
+            help="When enabled, utterances (cuts) will be concatenated "
+            "to minimize the amount of padding.",
+        )
+        group.add_argument(
+            "--duration-factor",
+            type=float,
+            default=1.0,
+            help="Determines the maximum duration of a concatenated cut "
+            "relative to the duration of the longest cut in a batch.",
+        )
+        group.add_argument(
+            "--gap",
+            type=float,
+            default=1.0,
+            help="The amount of padding (in seconds) inserted between "
+            "concatenated cuts. This padding is filled with noise when "
+            "noise augmentation is used.",
+        )
+        group.add_argument(
+            "--on-the-fly-feats",
+            type=str2bool,
+            default=False,
+            help="When enabled, use on-the-fly cut mixing and feature "
+            "extraction. Will drop existing precomputed feature manifests "
+            "if available.",
+        )
+        group.add_argument(
+            "--shuffle",
+            type=str2bool,
+            default=True,
+            help="When enabled (=default), the examples will be "
+            "shuffled for each epoch.",
+        )
+        group.add_argument(
+            "--drop-last",
+            type=str2bool,
+            default=True,
+            help="Whether to drop last batch. Used by sampler.",
+        )
+        group.add_argument(
+            "--return-cuts",
+            type=str2bool,
+            default=True,
+            help="When enabled, each batch will have the "
+            "field: batch['supervisions']['cut'] with the cuts that "
+            "were used to construct it.",
+        )
+
+        group.add_argument(
+            "--num-workers",
+            type=int,
+            default=2,
+            help="The number of training dataloader workers that "
+            "collect the batches.",
+        )
+
+        group.add_argument(
+            "--enable-spec-aug",
+            type=str2bool,
+            default=True,
+            help="When enabled, use SpecAugment for training dataset.",
+        )
+
+        group.add_argument(
+            "--spec-aug-time-warp-factor",
+            type=int,
+            default=80,
+            help="Used only when --enable-spec-aug is True. "
+            "It specifies the factor for time warping in SpecAugment. "
+            "Larger values mean more warping. "
+            "A value less than 1 means to disable time warp.",
+        )
+
+        group.add_argument(
+            "--enable-musan",
+            type=str2bool,
+            default=True,
+            help="When enabled, select noise from MUSAN and mix it"
+            "with training dataset. ",
+        )
+
+        group.add_argument(
+            "--input-strategy",
+            type=str,
+            default="PrecomputedFeatures",
+            help="AudioSamples or PrecomputedFeatures",
+        )
+
+        # GigaSpeech specific arguments
+        group.add_argument(
+            "--subset",
+            type=str,
+            default="XL",
+            help="Select the GigaSpeech subset (XS|S|M|L|XL)",
+        )
+        group.add_argument(
+            "--small-dev",
+            type=str2bool,
+            default=False,
+            help="Should we use only 1000 utterances for dev (speeds up training)",
+        )
+
+    def train_dataloaders(
+        self,
+        cuts_train: CutSet,
+        sampler_state_dict: Optional[Dict[str, Any]] = None,
+    ) -> DataLoader:
+        """
+        Args:
+          cuts_train:
+            CutSet for training.
+          sampler_state_dict:
+            The state dict for the training sampler.
+        """
+        transforms = []
+        if self.args.enable_musan:
+            logging.info("Enable MUSAN")
+            logging.info("About to get Musan cuts")
+            cuts_musan = load_manifest(self.args.manifest_dir / "musan_cuts.jsonl.gz")
+            transforms.append(
+                CutMix(cuts=cuts_musan, prob=0.5, snr=(10, 20), preserve_id=True)
+            )
+        else:
+            logging.info("Disable MUSAN")
+
+        if self.args.concatenate_cuts:
+            logging.info(
+                f"Using cut concatenation with duration factor "
+                f"{self.args.duration_factor} and gap {self.args.gap}."
+            )
+            # Cut concatenation should be the first transform in the list,
+            # so that if we e.g. mix noise in, it will fill the gaps between
+            # different utterances.
+            transforms = [
+                CutConcatenate(
+                    duration_factor=self.args.duration_factor, gap=self.args.gap
+                )
+            ] + transforms
+
+        input_transforms = []
+        if self.args.enable_spec_aug:
+            logging.info("Enable SpecAugment")
+            logging.info(f"Time warp factor: {self.args.spec_aug_time_warp_factor}")
+            # Set the value of num_frame_masks according to Lhotse's version.
+            # In different Lhotse's versions, the default of num_frame_masks is
+            # different.
+            num_frame_masks = 10
+            num_frame_masks_parameter = inspect.signature(
+                SpecAugment.__init__
+            ).parameters["num_frame_masks"]
+            if num_frame_masks_parameter.default == 1:
+                num_frame_masks = 2
+            logging.info(f"Num frame mask: {num_frame_masks}")
+            input_transforms.append(
+                SpecAugment(
+                    time_warp_factor=self.args.spec_aug_time_warp_factor,
+                    num_frame_masks=num_frame_masks,
+                    features_mask_size=27,
+                    num_feature_masks=2,
+                    frames_mask_size=100,
+                )
+            )
+        else:
+            logging.info("Disable SpecAugment")
+
+        logging.info("About to create train dataset")
+        train = K2SpeechRecognitionDataset(
+            input_strategy=eval(self.args.input_strategy)(),
+            cut_transforms=transforms,
+            input_transforms=input_transforms,
+            return_cuts=self.args.return_cuts,
+        )
+
+        if self.args.on_the_fly_feats:
+            # NOTE: the PerturbSpeed transform should be added only if we
+            # remove it from data prep stage.
+            # Add on-the-fly speed perturbation; since originally it would
+            # have increased epoch size by 3, we will apply prob 2/3 and use
+            # 3x more epochs.
+            # Speed perturbation probably should come first before
+            # concatenation, but in principle the transforms order doesn't have
+            # to be strict (e.g. could be randomized)
+            # transforms = [PerturbSpeed(factors=[0.9, 1.1], p=2/3)] + transforms   # noqa
+            # Drop feats to be on the safe side.
+            train = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80))),
+                input_transforms=input_transforms,
+                return_cuts=self.args.return_cuts,
+            )
+
+        if self.args.bucketing_sampler:
+            logging.info("Using DynamicBucketingSampler.")
+            train_sampler = DynamicBucketingSampler(
+                cuts_train,
+                max_duration=self.args.max_duration,
+                shuffle=self.args.shuffle,
+                num_buckets=self.args.num_buckets,
+                drop_last=self.args.drop_last,
+            )
+        else:
+            logging.info("Using SimpleCutSampler.")
+            train_sampler = SimpleCutSampler(
+                cuts_train,
+                max_duration=self.args.max_duration,
+                shuffle=self.args.shuffle,
+            )
+        logging.info("About to create train dataloader")
+
+        if sampler_state_dict is not None:
+            logging.info("Loading sampler state dict")
+            train_sampler.load_state_dict(sampler_state_dict)
+
+        # 'seed' is derived from the current random state, which will have
+        # previously been set in the main process.
+        seed = torch.randint(0, 100000, ()).item()
+        worker_init_fn = _SeedWorkers(seed)
+
+        train_dl = DataLoader(
+            train,
+            sampler=train_sampler,
+            batch_size=None,
+            num_workers=self.args.num_workers,
+            persistent_workers=False,
+            worker_init_fn=worker_init_fn,
+        )
+
+        return train_dl
+
+    def valid_dataloaders(self, cuts_valid: CutSet) -> DataLoader:
+        transforms = []
+        if self.args.concatenate_cuts:
+            transforms = [
+                CutConcatenate(
+                    duration_factor=self.args.duration_factor, gap=self.args.gap
+                )
+            ] + transforms
+
+        logging.info("About to create dev dataset")
+        if self.args.on_the_fly_feats:
+            validate = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80))),
+                return_cuts=self.args.return_cuts,
+            )
+        else:
+            validate = K2SpeechRecognitionDataset(
+                cut_transforms=transforms,
+                return_cuts=self.args.return_cuts,
+            )
+        valid_sampler = DynamicBucketingSampler(
+            cuts_valid,
+            max_duration=self.args.max_duration,
+            shuffle=False,
+        )
+        logging.info("About to create dev dataloader")
+        valid_dl = DataLoader(
+            validate,
+            sampler=valid_sampler,
+            batch_size=None,
+            num_workers=2,
+            persistent_workers=False,
+        )
+
+        return valid_dl
+
+    def test_dataloaders(self, cuts: CutSet) -> DataLoader:
+        logging.debug("About to create test dataset")
+        test = K2SpeechRecognitionDataset(
+            input_strategy=OnTheFlyFeatures(Fbank(FbankConfig(num_mel_bins=80)))
+            if self.args.on_the_fly_feats
+            else eval(self.args.input_strategy)(),
+            return_cuts=self.args.return_cuts,
+        )
+        sampler = DynamicBucketingSampler(
+            cuts,
+            max_duration=self.args.max_duration,
+            shuffle=False,
+        )
+        logging.debug("About to create test dataloader")
+        test_dl = DataLoader(
+            test,
+            batch_size=None,
+            sampler=sampler,
+            num_workers=self.args.num_workers,
+        )
+        return test_dl
+
+    @lru_cache()
+    def train_cuts(self) -> CutSet:
+        logging.info(f"About to get train {self.args.subset} cuts")
+        if self.args.subset == "XL":
+            filenames = glob.glob(
+                f"{self.args.manifest_dir}/XL_split/gigaspeech_cuts_XL.*.jsonl.gz"
+            )
+            pattern = re.compile(r"gigaspeech_cuts_XL.([0-9]+).jsonl.gz")
+            idx_filenames = ((int(pattern.search(f).group(1)), f) for f in filenames)
+            idx_filenames = sorted(idx_filenames, key=lambda x: x[0])
+            sorted_filenames = [f[1] for f in idx_filenames]
+            logging.info(
+                f"Loading GigaSpeech {len(sorted_filenames)} splits in lazy mode"
+            )
+            cuts_train = lhotse.combine(
+                lhotse.load_manifest_lazy(p) for p in sorted_filenames
+            )
+        else:
+            path = (
+                self.args.manifest_dir / f"gigaspeech_cuts_{self.args.subset}.jsonl.gz"
+            )
+            cuts_train = CutSet.from_jsonl_lazy(path)
+        return cuts_train
+
+    @lru_cache()
+    def dev_cuts(self) -> CutSet:
+        logging.info("About to get dev cuts")
+        cuts_valid = load_manifest_lazy(
+            self.args.manifest_dir / "gigaspeech_cuts_DEV.jsonl.gz"
+        )
+        if self.args.small_dev:
+            return cuts_valid.subset(first=1000)
+        else:
+            return cuts_valid
+
+    @lru_cache()
+    def test_cuts(self) -> CutSet:
+        logging.info("About to get test cuts")
+        return load_manifest_lazy(
+            self.args.manifest_dir / "gigaspeech_cuts_TEST.jsonl.gz"
+        )
--- a/egs/gigaspeech/ASR/zipformer/beam_search.py
+++ b/egs/gigaspeech/ASR/zipformer/beam_search.py
@ -0,0 +1 @@
+../../../librispeech/ASR/pruned_transducer_stateless2/beam_search.py
--- a/egs/gigaspeech/ASR/zipformer/ctc_decode.py
+++ b/egs/gigaspeech/ASR/zipformer/ctc_decode.py
@ -0,0 +1,847 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021-2022 Xiaomi Corporation (Author: Fangjun Kuang,
+#                                                 Liyong Guo,
+#                                                 Quandong Wang,
+#                                                 Zengwei Yao)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""
+Usage:
+
+(1) ctc-decoding
+./zipformer/ctc_decode.py \
+    --epoch 30 \
+    --avg 15 \
+    --exp-dir ./zipformer/exp \
+    --use-ctc 1 \
+    --max-duration 600 \
+    --decoding-method ctc-decoding
+
+(2) 1best
+./zipformer/ctc_decode.py \
+    --epoch 30 \
+    --avg 15 \
+    --exp-dir ./zipformer/exp \
+    --use-ctc 1 \
+    --max-duration 600 \
+    --hlg-scale 0.6 \
+    --decoding-method 1best
+
+(3) nbest
+./zipformer/ctc_decode.py \
+    --epoch 30 \
+    --avg 15 \
+    --exp-dir ./zipformer/exp \
+    --use-ctc 1 \
+    --max-duration 600 \
+    --hlg-scale 0.6 \
+    --decoding-method nbest
+
+(4) nbest-rescoring
+./zipformer/ctc_decode.py \
+    --epoch 30 \
+    --avg 15 \
+    --exp-dir ./zipformer/exp \
+    --use-ctc 1 \
+    --max-duration 600 \
+    --hlg-scale 0.6 \
+    --nbest-scale 1.0 \
+    --lm-dir data/lm \
+    --decoding-method nbest-rescoring
+
+(5) whole-lattice-rescoring
+./zipformer/ctc_decode.py \
+    --epoch 30 \
+    --avg 15 \
+    --exp-dir ./zipformer/exp \
+    --use-ctc 1 \
+    --max-duration 600 \
+    --hlg-scale 0.6 \
+    --nbest-scale 1.0 \
+    --lm-dir data/lm \
+    --decoding-method whole-lattice-rescoring
+"""
+
+
+import argparse
+import logging
+import math
+from collections import defaultdict
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+import k2
+import sentencepiece as spm
+import torch
+import torch.nn as nn
+from asr_datamodule import GigaSpeechAsrDataModule
+from train import add_model_arguments, get_params, get_model
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.decode import (
+    get_lattice,
+    nbest_decoding,
+    nbest_oracle,
+    one_best_decoding,
+    rescore_with_n_best_list,
+    rescore_with_whole_lattice,
+)
+from icefall.lexicon import Lexicon
+from icefall.utils import (
+    AttributeDict,
+    get_texts,
+    setup_logger,
+    store_transcripts,
+    str2bool,
+    write_error_stats,
+)
+
+LOG_EPS = math.log(1e-10)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=30,
+        help="""It specifies the checkpoint to use for decoding.
+        Note: Epoch counts from 1.
+        You can specify --avg to use more checkpoints for model averaging.""",
+    )
+
+    parser.add_argument(
+        "--iter",
+        type=int,
+        default=0,
+        help="""If positive, --epoch is ignored and it
+        will use the checkpoint exp_dir/checkpoint-iter.pt.
+        You can specify --avg to use more checkpoints for model averaging.
+        """,
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=15,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=True,
+        help="Whether to load averaged model. Currently it only supports "
+        "using --epoch. If True, it would decode with the averaged model "
+        "over the epoch range from `epoch-avg` (excluded) to `epoch`."
+        "Actually only the models with epoch number of `epoch-avg` and "
+        "`epoch` are loaded for averaging. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="zipformer/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--lang-dir",
+        type=Path,
+        default="data/lang_bpe_500",
+        help="The lang dir containing word table and LG graph",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="ctc-decoding",
+        help="""Decoding method.
+        Supported values are:
+        - (1) ctc-decoding. Use CTC decoding. It uses a sentence piece
+          model, i.e., lang_dir/bpe.model, to convert word pieces to words.
+          It needs neither a lexicon nor an n-gram LM.
+        - (2) 1best. Extract the best path from the decoding lattice as the
+          decoding result.
+        - (3) nbest. Extract n paths from the decoding lattice; the path
+          with the highest score is the decoding result.
+        - (4) nbest-rescoring. Extract n paths from the decoding lattice,
+          rescore them with an n-gram LM (e.g., a 4-gram LM), the path with
+          the highest score is the decoding result.
+        - (5) whole-lattice-rescoring. Rescore the decoding lattice with an
+          n-gram LM (e.g., a 4-gram LM), the best path of rescored lattice
+          is the decoding result.
+          you have trained an RNN LM using ./rnn_lm/train.py
+        - (6) nbest-oracle. Its WER is the lower bound of any n-best
+          rescoring method can achieve. Useful for debugging n-best
+          rescoring method.
+        """,
+    )
+
+    parser.add_argument(
+        "--num-paths",
+        type=int,
+        default=100,
+        help="""Number of paths for n-best based decoding method.
+        Used only when "method" is one of the following values:
+        nbest, nbest-rescoring, and nbest-oracle
+        """,
+    )
+
+    parser.add_argument(
+        "--nbest-scale",
+        type=float,
+        default=1.0,
+        help="""The scale to be applied to `lattice.scores`.
+        It's needed if you use any kinds of n-best based rescoring.
+        Used only when "method" is one of the following values:
+        nbest, nbest-rescoring, and nbest-oracle
+        A smaller value results in more unique paths.
+        """,
+    )
+
+    parser.add_argument(
+        "--hlg-scale",
+        type=float,
+        default=0.6,
+        help="""The scale to be applied to `hlg.scores`.
+        """,
+    )
+
+    parser.add_argument(
+        "--lm-dir",
+        type=str,
+        default="data/lm",
+        help="""The n-gram LM dir.
+        It should contain either G_4_gram.pt or G_4_gram.fst.txt
+        """,
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def get_decoding_params() -> AttributeDict:
+    """Parameters for decoding."""
+    params = AttributeDict(
+        {
+            "frame_shift_ms": 10,
+            "search_beam": 20,
+            "output_beam": 8,
+            "min_active_states": 30,
+            "max_active_states": 10000,
+            "use_double_scores": True,
+        }
+    )
+    return params
+
+
+def decode_one_batch(
+    params: AttributeDict,
+    model: nn.Module,
+    HLG: Optional[k2.Fsa],
+    H: Optional[k2.Fsa],
+    bpe_model: Optional[spm.SentencePieceProcessor],
+    batch: dict,
+    word_table: k2.SymbolTable,
+    G: Optional[k2.Fsa] = None,
+) -> Dict[str, List[List[str]]]:
+    """Decode one batch and return the result in a dict. The dict has the
+    following format:
+    - key: It indicates the setting used for decoding. For example,
+           if no rescoring is used, the key is the string `no_rescore`.
+           If LM rescoring is used, the key is the string `lm_scale_xxx`,
+           where `xxx` is the value of `lm_scale`. An example key is
+           `lm_scale_0.7`
+    - value: It contains the decoding result. `len(value)` equals to
+             batch size. `value[i]` is the decoding result for the i-th
+             utterance in the given batch.
+
+    Args:
+      params:
+        It's the return value of :func:`get_params`.
+
+        - params.decoding_method is "1best", it uses 1best decoding without LM rescoring.
+        - params.decoding_method is "nbest", it uses nbest decoding without LM rescoring.
+        - params.decoding_method is "nbest-rescoring", it uses nbest LM rescoring.
+        - params.decoding_method is "whole-lattice-rescoring", it uses whole lattice LM
+          rescoring.
+
+      model:
+        The neural model.
+      HLG:
+        The decoding graph. Used only when params.decoding_method is NOT ctc-decoding.
+      H:
+        The ctc topo. Used only when params.decoding_method is ctc-decoding.
+      bpe_model:
+        The BPE model. Used only when params.decoding_method is ctc-decoding.
+      batch:
+        It is the return value from iterating
+        `lhotse.dataset.K2SpeechRecognitionDataset`. See its documentation
+        for the format of the `batch`.
+      word_table:
+        The word symbol table.
+      G:
+        An LM. It is not None when params.decoding_method is "nbest-rescoring"
+        or "whole-lattice-rescoring". In general, the G in HLG
+        is a 3-gram LM, while this G is a 4-gram LM.
+    Returns:
+      Return the decoding result. See above description for the format of
+      the returned dict. Note: If it decodes to nothing, then return None.
+    """
+    if HLG is not None:
+        device = HLG.device
+    else:
+        device = H.device
+    feature = batch["inputs"]
+    assert feature.ndim == 3
+    feature = feature.to(device)
+    # at entry, feature is (N, T, C)
+
+    supervisions = batch["supervisions"]
+    feature_lens = supervisions["num_frames"].to(device)
+
+    if params.causal:
+        # this seems to cause insertions at the end of the utterance if used with zipformer.
+        pad_len = 30
+        feature_lens += pad_len
+        feature = torch.nn.functional.pad(
+            feature,
+            pad=(0, 0, 0, pad_len),
+            value=LOG_EPS,
+        )
+
+    encoder_out, encoder_out_lens = model.forward_encoder(feature, feature_lens)
+    ctc_output = model.ctc_output(encoder_out)  # (N, T, C)
+
+    supervision_segments = torch.stack(
+        (
+            supervisions["sequence_idx"],
+            torch.div(
+                supervisions["start_frame"],
+                params.subsampling_factor,
+                rounding_mode="floor",
+            ),
+            torch.div(
+                supervisions["num_frames"],
+                params.subsampling_factor,
+                rounding_mode="floor",
+            ),
+        ),
+        1,
+    ).to(torch.int32)
+
+    if H is None:
+        assert HLG is not None
+        decoding_graph = HLG
+    else:
+        assert HLG is None
+        assert bpe_model is not None
+        decoding_graph = H
+
+    lattice = get_lattice(
+        nnet_output=ctc_output,
+        decoding_graph=decoding_graph,
+        supervision_segments=supervision_segments,
+        search_beam=params.search_beam,
+        output_beam=params.output_beam,
+        min_active_states=params.min_active_states,
+        max_active_states=params.max_active_states,
+        subsampling_factor=params.subsampling_factor,
+    )
+
+    if params.decoding_method == "ctc-decoding":
+        best_path = one_best_decoding(
+            lattice=lattice, use_double_scores=params.use_double_scores
+        )
+        # Note: `best_path.aux_labels` contains token IDs, not word IDs
+        # since we are using H, not HLG here.
+        #
+        # token_ids is a lit-of-list of IDs
+        token_ids = get_texts(best_path)
+
+        # hyps is a list of str, e.g., ['xxx yyy zzz', ...]
+        hyps = bpe_model.decode(token_ids)
+
+        # hyps is a list of list of str, e.g., [['xxx', 'yyy', 'zzz'], ... ]
+        hyps = [s.split() for s in hyps]
+        key = "ctc-decoding"
+        return {key: hyps}
+
+    if params.decoding_method == "nbest-oracle":
+        # Note: You can also pass rescored lattices to it.
+        # We choose the HLG decoded lattice for speed reasons
+        # as HLG decoding is faster and the oracle WER
+        # is only slightly worse than that of rescored lattices.
+        best_path = nbest_oracle(
+            lattice=lattice,
+            num_paths=params.num_paths,
+            ref_texts=supervisions["text"],
+            word_table=word_table,
+            nbest_scale=params.nbest_scale,
+            oov="<UNK>",
+        )
+        hyps = get_texts(best_path)
+        hyps = [[word_table[i] for i in ids] for ids in hyps]
+        key = f"oracle_{params.num_paths}_nbest_scale_{params.nbest_scale}"  # noqa
+        return {key: hyps}
+
+    if params.decoding_method in ["1best", "nbest"]:
+        if params.decoding_method == "1best":
+            best_path = one_best_decoding(
+                lattice=lattice, use_double_scores=params.use_double_scores
+            )
+            key = "no_rescore"
+        else:
+            best_path = nbest_decoding(
+                lattice=lattice,
+                num_paths=params.num_paths,
+                use_double_scores=params.use_double_scores,
+                nbest_scale=params.nbest_scale,
+            )
+            key = f"no_rescore-nbest-scale-{params.nbest_scale}-{params.num_paths}"  # noqa
+
+        hyps = get_texts(best_path)
+        hyps = [[word_table[i] for i in ids] for ids in hyps]
+        return {key: hyps}
+
+    assert params.decoding_method in [
+        "nbest-rescoring",
+        "whole-lattice-rescoring",
+    ]
+
+    lm_scale_list = [0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7]
+    lm_scale_list += [0.8, 0.9, 1.0, 1.1, 1.2, 1.3]
+    lm_scale_list += [1.4, 1.5, 1.6, 1.7, 1.8, 1.9, 2.0]
+
+    if params.decoding_method == "nbest-rescoring":
+        best_path_dict = rescore_with_n_best_list(
+            lattice=lattice,
+            G=G,
+            num_paths=params.num_paths,
+            lm_scale_list=lm_scale_list,
+            nbest_scale=params.nbest_scale,
+        )
+    elif params.decoding_method == "whole-lattice-rescoring":
+        best_path_dict = rescore_with_whole_lattice(
+            lattice=lattice,
+            G_with_epsilon_loops=G,
+            lm_scale_list=lm_scale_list,
+        )
+    else:
+        assert False, f"Unsupported decoding method: {params.decoding_method}"
+
+    ans = dict()
+    if best_path_dict is not None:
+        for lm_scale_str, best_path in best_path_dict.items():
+            hyps = get_texts(best_path)
+            hyps = [[word_table[i] for i in ids] for ids in hyps]
+            ans[lm_scale_str] = hyps
+    else:
+        ans = None
+    return ans
+
+
+def decode_dataset(
+    dl: torch.utils.data.DataLoader,
+    params: AttributeDict,
+    model: nn.Module,
+    HLG: Optional[k2.Fsa],
+    H: Optional[k2.Fsa],
+    bpe_model: Optional[spm.SentencePieceProcessor],
+    word_table: k2.SymbolTable,
+    G: Optional[k2.Fsa] = None,
+) -> Dict[str, List[Tuple[str, List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      dl:
+        PyTorch's dataloader containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      HLG:
+        The decoding graph. Used only when params.decoding_method is NOT ctc-decoding.
+      H:
+        The ctc topo. Used only when params.decoding_method is ctc-decoding.
+      bpe_model:
+        The BPE model. Used only when params.decoding_method is ctc-decoding.
+      word_table:
+        It is the word symbol table.
+      G:
+        An LM. It is not None when params.decoding_method is "nbest-rescoring"
+        or "whole-lattice-rescoring". In general, the G in HLG
+        is a 3-gram LM, while this G is a 4-gram LM.
+    Returns:
+      Return a dict, whose key may be "no-rescore" if no LM rescoring
+      is used, or it may be "lm_scale_0.7" if LM rescoring is used.
+      Its value is a list of tuples. Each tuple contains two elements:
+      The first is the reference transcript, and the second is the
+      predicted result.
+    """
+    num_cuts = 0
+
+    try:
+        num_batches = len(dl)
+    except TypeError:
+        num_batches = "?"
+
+    results = defaultdict(list)
+    for batch_idx, batch in enumerate(dl):
+        texts = batch["supervisions"]["text"]
+        cut_ids = [cut.id for cut in batch["supervisions"]["cut"]]
+
+        hyps_dict = decode_one_batch(
+            params=params,
+            model=model,
+            HLG=HLG,
+            H=H,
+            bpe_model=bpe_model,
+            batch=batch,
+            word_table=word_table,
+            G=G,
+        )
+
+        for name, hyps in hyps_dict.items():
+            this_batch = []
+            assert len(hyps) == len(texts)
+            for cut_id, hyp_words, ref_text in zip(cut_ids, hyps, texts):
+                ref_words = ref_text.split()
+                this_batch.append((cut_id, ref_words, hyp_words))
+
+            results[name].extend(this_batch)
+
+        num_cuts += len(texts)
+
+        if batch_idx % 100 == 0:
+            batch_str = f"{batch_idx}/{num_batches}"
+
+            logging.info(f"batch {batch_str}, cuts processed until now is {num_cuts}")
+    return results
+
+
+def save_results(
+    params: AttributeDict,
+    test_set_name: str,
+    results_dict: Dict[str, List[Tuple[str, List[str], List[str]]]],
+):
+    test_set_wers = dict()
+    for key, results in results_dict.items():
+        recog_path = params.res_dir / f"recogs-{test_set_name}-{params.suffix}.txt"
+        results = sorted(results)
+        store_transcripts(filename=recog_path, texts=results)
+        logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out WERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = params.res_dir / f"errs-{test_set_name}-{params.suffix}.txt"
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(f, f"{test_set_name}-{key}", results)
+            test_set_wers[key] = wer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_wers = sorted(test_set_wers.items(), key=lambda x: x[1])
+    errs_info = params.res_dir / f"wer-summary-{test_set_name}-{params.suffix}.txt"
+    with open(errs_info, "w") as f:
+        print("settings\tWER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, WER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_wers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    GigaSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+    args.lang_dir = Path(args.lang_dir)
+    args.lm_dir = Path(args.lm_dir)
+
+    params = get_params()
+    # add decoding params
+    params.update(get_decoding_params())
+    params.update(vars(args))
+
+    assert params.decoding_method in (
+        "ctc-decoding",
+        "1best",
+        "nbest",
+        "nbest-rescoring",
+        "whole-lattice-rescoring",
+        "nbest-oracle",
+    )
+    params.res_dir = params.exp_dir / params.decoding_method
+
+    if params.iter > 0:
+        params.suffix = f"iter-{params.iter}-avg-{params.avg}"
+    else:
+        params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+
+    if params.causal:
+        assert (
+            "," not in params.chunk_size
+        ), "chunk_size should be one value in decoding."
+        assert (
+            "," not in params.left_context_frames
+        ), "left_context_frames should be one value in decoding."
+        params.suffix += f"-chunk-{params.chunk_size}"
+        params.suffix += f"-left-context-{params.left_context_frames}"
+
+    if params.use_averaged_model:
+        params.suffix += "-use-averaged-model"
+
+    setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
+    logging.info("Decoding started")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"Device: {device}")
+    logging.info(params)
+
+    lexicon = Lexicon(params.lang_dir)
+    max_token_id = max(lexicon.tokens)
+    num_classes = max_token_id + 1  # +1 for the blank
+
+    params.vocab_size = num_classes
+    # <blk> and <unk> are defined in local/train_bpe_model.py
+    params.blank_id = 0
+
+    if params.decoding_method == "ctc-decoding":
+        HLG = None
+        H = k2.ctc_topo(
+            max_token=max_token_id,
+            modified=False,
+            device=device,
+        )
+        bpe_model = spm.SentencePieceProcessor()
+        bpe_model.load(str(params.lang_dir / "bpe.model"))
+    else:
+        H = None
+        bpe_model = None
+        HLG = k2.Fsa.from_dict(
+            torch.load(f"{params.lang_dir}/HLG.pt", map_location=device)
+        )
+        assert HLG.requires_grad is False
+
+        HLG.scores *= params.hlg_scale
+        if not hasattr(HLG, "lm_scores"):
+            HLG.lm_scores = HLG.scores.clone()
+
+    if params.decoding_method in (
+        "nbest-rescoring",
+        "whole-lattice-rescoring",
+    ):
+        if not (params.lm_dir / "G_4_gram.pt").is_file():
+            logging.info("Loading G_4_gram.fst.txt")
+            logging.warning("It may take 8 minutes.")
+            with open(params.lm_dir / "G_4_gram.fst.txt") as f:
+                first_word_disambig_id = lexicon.word_table["#0"]
+
+                G = k2.Fsa.from_openfst(f.read(), acceptor=False)
+                # G.aux_labels is not needed in later computations, so
+                # remove it here.
+                del G.aux_labels
+                # CAUTION: The following line is crucial.
+                # Arcs entering the back-off state have label equal to #0.
+                # We have to change it to 0 here.
+                G.labels[G.labels >= first_word_disambig_id] = 0
+                # See https://github.com/k2-fsa/k2/issues/874
+                # for why we need to set G.properties to None
+                G.__dict__["_properties"] = None
+                G = k2.Fsa.from_fsas([G]).to(device)
+                G = k2.arc_sort(G)
+                # Save a dummy value so that it can be loaded in C++.
+                # See https://github.com/pytorch/pytorch/issues/67902
+                # for why we need to do this.
+                G.dummy = 1
+
+                torch.save(G.as_dict(), params.lm_dir / "G_4_gram.pt")
+        else:
+            logging.info("Loading pre-compiled G_4_gram.pt")
+            d = torch.load(params.lm_dir / "G_4_gram.pt", map_location=device)
+            G = k2.Fsa.from_dict(d)
+
+        if params.decoding_method == "whole-lattice-rescoring":
+            # Add epsilon self-loops to G as we will compose
+            # it with the whole lattice later
+            G = k2.add_epsilon_self_loops(G)
+            G = k2.arc_sort(G)
+            G = G.to(device)
+
+        # G.lm_scores is used to replace HLG.lm_scores during
+        # LM rescoring.
+        G.lm_scores = G.scores.clone()
+    else:
+        G = None
+
+    logging.info("About to create model")
+    model = get_model(params)
+
+    if not params.use_averaged_model:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+        elif params.avg == 1:
+            load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+        else:
+            start = params.epoch - params.avg + 1
+            filenames = []
+            for i in range(start, params.epoch + 1):
+                if i >= 1:
+                    filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+    else:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg + 1
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg + 1:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            filename_start = filenames[-1]
+            filename_end = filenames[0]
+            logging.info(
+                "Calculating the averaged model over iteration checkpoints"
+                f" from {filename_start} (excluded) to {filename_end}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+        else:
+            assert params.avg > 0, params.avg
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+            filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+            logging.info(
+                f"Calculating the averaged model over epoch range from "
+                f"{start} (excluded) to {params.epoch}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+
+    model.to(device)
+    model.eval()
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    # we need cut ids to display recognition results.
+    args.return_cuts = True
+    gigaspeech = GigaSpeechAsrDataModule(args)
+
+    test_clean_cuts = gigaspeech.test_clean_cuts()
+    test_other_cuts = gigaspeech.test_other_cuts()
+
+    test_clean_dl = gigaspeech.test_dataloaders(test_clean_cuts)
+    test_other_dl = gigaspeech.test_dataloaders(test_other_cuts)
+
+    test_sets = ["test-clean", "test-other"]
+    test_dl = [test_clean_dl, test_other_dl]
+
+    for test_set, test_dl in zip(test_sets, test_dl):
+        results_dict = decode_dataset(
+            dl=test_dl,
+            params=params,
+            model=model,
+            HLG=HLG,
+            H=H,
+            bpe_model=bpe_model,
+            word_table=lexicon.word_table,
+            G=G,
+        )
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    main()
--- a/egs/gigaspeech/ASR/zipformer/decode.py
+++ b/egs/gigaspeech/ASR/zipformer/decode.py
--- a/egs/gigaspeech/ASR/zipformer/decode_stream.py
+++ b/egs/gigaspeech/ASR/zipformer/decode_stream.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/decode_stream.py
--- a/egs/gigaspeech/ASR/zipformer/decoder.py
+++ b/egs/gigaspeech/ASR/zipformer/decoder.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/decoder.py
--- a/egs/gigaspeech/ASR/zipformer/encoder_interface.py
+++ b/egs/gigaspeech/ASR/zipformer/encoder_interface.py
@ -0,0 +1 @@
+../../../librispeech/ASR/transducer_stateless/encoder_interface.py
--- a/egs/gigaspeech/ASR/zipformer/export-onnx-ctc.py
+++ b/egs/gigaspeech/ASR/zipformer/export-onnx-ctc.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/export-onnx-ctc.py
--- a/egs/gigaspeech/ASR/zipformer/export-onnx-streaming.py
+++ b/egs/gigaspeech/ASR/zipformer/export-onnx-streaming.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/export-onnx-streaming.py
--- a/egs/gigaspeech/ASR/zipformer/export-onnx.py
+++ b/egs/gigaspeech/ASR/zipformer/export-onnx.py
@ -0,0 +1,620 @@
+#!/usr/bin/env python3
+#
+# Copyright 2023 Xiaomi Corporation (Author: Fangjun Kuang, Wei Kang)
+# Copyright 2023 Danqing Fu (danqing.fu@gmail.com)
+
+"""
+This script exports a transducer model from PyTorch to ONNX.
+
+We use the pre-trained model from
+https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
+as an example to show how to use this file.
+
+1. Download the pre-trained model
+
+cd egs/gigaspeech/ASR
+
+repo_url=https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
+GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
+repo=$(basename $repo_url)
+
+pushd $repo
+git lfs pull --include "exp/pretrained.pt"
+
+cd exp
+ln -s pretrained.pt epoch-99.pt
+popd
+
+2. Export the model to ONNX
+
+./zipformer/export-onnx.py \
+  --tokens $repo/data/lang_bpe_500/tokens.txt \
+  --use-averaged-model 0 \
+  --epoch 99 \
+  --avg 1 \
+  --exp-dir $repo/exp \
+  --num-encoder-layers "2,2,3,4,3,2" \
+  --downsampling-factor "1,2,4,8,4,2" \
+  --feedforward-dim "512,768,1024,1536,1024,768" \
+  --num-heads "4,4,4,8,4,4" \
+  --encoder-dim "192,256,384,512,384,256" \
+  --query-head-dim 32 \
+  --value-head-dim 12 \
+  --pos-head-dim 4 \
+  --pos-dim 48 \
+  --encoder-unmasked-dim "192,192,256,256,256,192" \
+  --cnn-module-kernel "31,31,15,15,15,31" \
+  --decoder-dim 512 \
+  --joiner-dim 512 \
+  --causal False \
+  --chunk-size "16,32,64,-1" \
+  --left-context-frames "64,128,256,-1"
+
+It will generate the following 3 files inside $repo/exp:
+
+  - encoder-epoch-99-avg-1.onnx
+  - decoder-epoch-99-avg-1.onnx
+  - joiner-epoch-99-avg-1.onnx
+
+See ./onnx_pretrained.py and ./onnx_check.py for how to
+use the exported ONNX models.
+"""
+
+import argparse
+import logging
+from pathlib import Path
+from typing import Dict, Tuple
+
+import k2
+import onnx
+import torch
+import torch.nn as nn
+from decoder import Decoder
+from onnxruntime.quantization import QuantType, quantize_dynamic
+from scaling_converter import convert_scaled_to_non_scaled
+from train import add_model_arguments, get_model, get_params
+from zipformer import Zipformer2
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.utils import make_pad_mask, num_tokens, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=28,
+        help="""It specifies the checkpoint to use for averaging.
+        Note: Epoch counts from 0.
+        You can specify --avg to use more checkpoints for model averaging.""",
+    )
+
+    parser.add_argument(
+        "--iter",
+        type=int,
+        default=0,
+        help="""If positive, --epoch is ignored and it
+        will use the checkpoint exp_dir/checkpoint-iter.pt.
+        You can specify --avg to use more checkpoints for model averaging.
+        """,
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=15,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=True,
+        help="Whether to load averaged model. Currently it only supports "
+        "using --epoch. If True, it would decode with the averaged model "
+        "over the epoch range from `epoch-avg` (excluded) to `epoch`."
+        "Actually only the models with epoch number of `epoch-avg` and "
+        "`epoch` are loaded for averaging. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="zipformer/exp",
+        help="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--tokens",
+        type=str,
+        default="data/lang_bpe_500/tokens.txt",
+        help="Path to the tokens.txt",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def add_meta_data(filename: str, meta_data: Dict[str, str]):
+    """Add meta data to an ONNX model. It is changed in-place.
+
+    Args:
+      filename:
+        Filename of the ONNX model to be changed.
+      meta_data:
+        Key-value pairs.
+    """
+    model = onnx.load(filename)
+    for key, value in meta_data.items():
+        meta = model.metadata_props.add()
+        meta.key = key
+        meta.value = value
+
+    onnx.save(model, filename)
+
+
+class OnnxEncoder(nn.Module):
+    """A wrapper for Zipformer and the encoder_proj from the joiner"""
+
+    def __init__(
+        self, encoder: Zipformer2, encoder_embed: nn.Module, encoder_proj: nn.Linear
+    ):
+        """
+        Args:
+          encoder:
+            A Zipformer encoder.
+          encoder_proj:
+            The projection layer for encoder from the joiner.
+        """
+        super().__init__()
+        self.encoder = encoder
+        self.encoder_embed = encoder_embed
+        self.encoder_proj = encoder_proj
+
+    def forward(
+        self,
+        x: torch.Tensor,
+        x_lens: torch.Tensor,
+    ) -> Tuple[torch.Tensor, torch.Tensor]:
+        """Please see the help information of Zipformer.forward
+
+        Args:
+          x:
+            A 3-D tensor of shape (N, T, C)
+          x_lens:
+            A 1-D tensor of shape (N,). Its dtype is torch.int64
+        Returns:
+          Return a tuple containing:
+            - encoder_out, A 3-D tensor of shape (N, T', joiner_dim)
+            - encoder_out_lens, A 1-D tensor of shape (N,)
+        """
+        x, x_lens = self.encoder_embed(x, x_lens)
+        src_key_padding_mask = make_pad_mask(x_lens)
+        x = x.permute(1, 0, 2)
+        encoder_out, encoder_out_lens = self.encoder(x, x_lens, src_key_padding_mask)
+        encoder_out = encoder_out.permute(1, 0, 2)
+        encoder_out = self.encoder_proj(encoder_out)
+        # Now encoder_out is of shape (N, T, joiner_dim)
+
+        return encoder_out, encoder_out_lens
+
+
+class OnnxDecoder(nn.Module):
+    """A wrapper for Decoder and the decoder_proj from the joiner"""
+
+    def __init__(self, decoder: Decoder, decoder_proj: nn.Linear):
+        super().__init__()
+        self.decoder = decoder
+        self.decoder_proj = decoder_proj
+
+    def forward(self, y: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+          y:
+            A 2-D tensor of shape (N, context_size).
+        Returns
+          Return a 2-D tensor of shape (N, joiner_dim)
+        """
+        need_pad = False
+        decoder_output = self.decoder(y, need_pad=need_pad)
+        decoder_output = decoder_output.squeeze(1)
+        output = self.decoder_proj(decoder_output)
+
+        return output
+
+
+class OnnxJoiner(nn.Module):
+    """A wrapper for the joiner"""
+
+    def __init__(self, output_linear: nn.Linear):
+        super().__init__()
+        self.output_linear = output_linear
+
+    def forward(
+        self,
+        encoder_out: torch.Tensor,
+        decoder_out: torch.Tensor,
+    ) -> torch.Tensor:
+        """
+        Args:
+          encoder_out:
+            A 2-D tensor of shape (N, joiner_dim)
+          decoder_out:
+            A 2-D tensor of shape (N, joiner_dim)
+        Returns:
+          Return a 2-D tensor of shape (N, vocab_size)
+        """
+        logit = encoder_out + decoder_out
+        logit = self.output_linear(torch.tanh(logit))
+        return logit
+
+
+def export_encoder_model_onnx(
+    encoder_model: OnnxEncoder,
+    encoder_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the given encoder model to ONNX format.
+    The exported model has two inputs:
+
+        - x, a tensor of shape (N, T, C); dtype is torch.float32
+        - x_lens, a tensor of shape (N,); dtype is torch.int64
+
+    and it has two outputs:
+
+        - encoder_out, a tensor of shape (N, T', joiner_dim)
+        - encoder_out_lens, a tensor of shape (N,)
+
+    Args:
+      encoder_model:
+        The input encoder model
+      encoder_filename:
+        The filename to save the exported ONNX model.
+      opset_version:
+        The opset version to use.
+    """
+    x = torch.zeros(1, 100, 80, dtype=torch.float32)
+    x_lens = torch.tensor([100], dtype=torch.int64)
+
+    encoder_model = torch.jit.trace(encoder_model, (x, x_lens))
+
+    torch.onnx.export(
+        encoder_model,
+        (x, x_lens),
+        encoder_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=["x", "x_lens"],
+        output_names=["encoder_out", "encoder_out_lens"],
+        dynamic_axes={
+            "x": {0: "N", 1: "T"},
+            "x_lens": {0: "N"},
+            "encoder_out": {0: "N", 1: "T"},
+            "encoder_out_lens": {0: "N"},
+        },
+    )
+
+    meta_data = {
+        "model_type": "zipformer2",
+        "version": "1",
+        "model_author": "k2-fsa",
+        "comment": "non-streaming zipformer2",
+    }
+    logging.info(f"meta_data: {meta_data}")
+
+    add_meta_data(filename=encoder_filename, meta_data=meta_data)
+
+
+def export_decoder_model_onnx(
+    decoder_model: OnnxDecoder,
+    decoder_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the decoder model to ONNX format.
+
+    The exported model has one input:
+
+        - y: a torch.int64 tensor of shape (N, decoder_model.context_size)
+
+    and has one output:
+
+        - decoder_out: a torch.float32 tensor of shape (N, joiner_dim)
+
+    Args:
+      decoder_model:
+        The decoder model to be exported.
+      decoder_filename:
+        Filename to save the exported ONNX model.
+      opset_version:
+        The opset version to use.
+    """
+    context_size = decoder_model.decoder.context_size
+    vocab_size = decoder_model.decoder.vocab_size
+
+    y = torch.zeros(10, context_size, dtype=torch.int64)
+    decoder_model = torch.jit.script(decoder_model)
+    torch.onnx.export(
+        decoder_model,
+        y,
+        decoder_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=["y"],
+        output_names=["decoder_out"],
+        dynamic_axes={
+            "y": {0: "N"},
+            "decoder_out": {0: "N"},
+        },
+    )
+
+    meta_data = {
+        "context_size": str(context_size),
+        "vocab_size": str(vocab_size),
+    }
+    add_meta_data(filename=decoder_filename, meta_data=meta_data)
+
+
+def export_joiner_model_onnx(
+    joiner_model: nn.Module,
+    joiner_filename: str,
+    opset_version: int = 11,
+) -> None:
+    """Export the joiner model to ONNX format.
+    The exported joiner model has two inputs:
+
+        - encoder_out: a tensor of shape (N, joiner_dim)
+        - decoder_out: a tensor of shape (N, joiner_dim)
+
+    and produces one output:
+
+        - logit: a tensor of shape (N, vocab_size)
+    """
+    joiner_dim = joiner_model.output_linear.weight.shape[1]
+    logging.info(f"joiner dim: {joiner_dim}")
+
+    projected_encoder_out = torch.rand(11, joiner_dim, dtype=torch.float32)
+    projected_decoder_out = torch.rand(11, joiner_dim, dtype=torch.float32)
+
+    torch.onnx.export(
+        joiner_model,
+        (projected_encoder_out, projected_decoder_out),
+        joiner_filename,
+        verbose=False,
+        opset_version=opset_version,
+        input_names=[
+            "encoder_out",
+            "decoder_out",
+        ],
+        output_names=["logit"],
+        dynamic_axes={
+            "encoder_out": {0: "N"},
+            "decoder_out": {0: "N"},
+            "logit": {0: "N"},
+        },
+    )
+    meta_data = {
+        "joiner_dim": str(joiner_dim),
+    }
+    add_meta_data(filename=joiner_filename, meta_data=meta_data)
+
+
+@torch.no_grad()
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    token_table = k2.SymbolTable.from_file(params.tokens)
+    params.blank_id = token_table["<blk>"]
+    params.vocab_size = num_tokens(token_table) + 1
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_model(params)
+
+    model.to(device)
+
+    if not params.use_averaged_model:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+        elif params.avg == 1:
+            load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+        else:
+            start = params.epoch - params.avg + 1
+            filenames = []
+            for i in range(start, params.epoch + 1):
+                if i >= 1:
+                    filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+    else:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg + 1
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg + 1:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            filename_start = filenames[-1]
+            filename_end = filenames[0]
+            logging.info(
+                "Calculating the averaged model over iteration checkpoints"
+                f" from {filename_start} (excluded) to {filename_end}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+        else:
+            assert params.avg > 0, params.avg
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+            filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+            logging.info(
+                f"Calculating the averaged model over epoch range from "
+                f"{start} (excluded) to {params.epoch}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+
+    model.to("cpu")
+    model.eval()
+
+    convert_scaled_to_non_scaled(model, inplace=True, is_onnx=True)
+
+    encoder = OnnxEncoder(
+        encoder=model.encoder,
+        encoder_embed=model.encoder_embed,
+        encoder_proj=model.joiner.encoder_proj,
+    )
+
+    decoder = OnnxDecoder(
+        decoder=model.decoder,
+        decoder_proj=model.joiner.decoder_proj,
+    )
+
+    joiner = OnnxJoiner(output_linear=model.joiner.output_linear)
+
+    encoder_num_param = sum([p.numel() for p in encoder.parameters()])
+    decoder_num_param = sum([p.numel() for p in decoder.parameters()])
+    joiner_num_param = sum([p.numel() for p in joiner.parameters()])
+    total_num_param = encoder_num_param + decoder_num_param + joiner_num_param
+    logging.info(f"encoder parameters: {encoder_num_param}")
+    logging.info(f"decoder parameters: {decoder_num_param}")
+    logging.info(f"joiner parameters: {joiner_num_param}")
+    logging.info(f"total parameters: {total_num_param}")
+
+    if params.iter > 0:
+        suffix = f"iter-{params.iter}"
+    else:
+        suffix = f"epoch-{params.epoch}"
+
+    suffix += f"-avg-{params.avg}"
+
+    opset_version = 13
+
+    logging.info("Exporting encoder")
+    encoder_filename = params.exp_dir / f"encoder-{suffix}.onnx"
+    export_encoder_model_onnx(
+        encoder,
+        encoder_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported encoder to {encoder_filename}")
+
+    logging.info("Exporting decoder")
+    decoder_filename = params.exp_dir / f"decoder-{suffix}.onnx"
+    export_decoder_model_onnx(
+        decoder,
+        decoder_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported decoder to {decoder_filename}")
+
+    logging.info("Exporting joiner")
+    joiner_filename = params.exp_dir / f"joiner-{suffix}.onnx"
+    export_joiner_model_onnx(
+        joiner,
+        joiner_filename,
+        opset_version=opset_version,
+    )
+    logging.info(f"Exported joiner to {joiner_filename}")
+
+    # Generate int8 quantization models
+    # See https://onnxruntime.ai/docs/performance/model-optimizations/quantization.html#data-type-selection
+
+    logging.info("Generate int8 quantization models")
+
+    encoder_filename_int8 = params.exp_dir / f"encoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=encoder_filename,
+        model_output=encoder_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+    decoder_filename_int8 = params.exp_dir / f"decoder-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=decoder_filename,
+        model_output=decoder_filename_int8,
+        op_types_to_quantize=["MatMul", "Gather"],
+        weight_type=QuantType.QInt8,
+    )
+
+    joiner_filename_int8 = params.exp_dir / f"joiner-{suffix}.int8.onnx"
+    quantize_dynamic(
+        model_input=joiner_filename,
+        model_output=joiner_filename_int8,
+        op_types_to_quantize=["MatMul"],
+        weight_type=QuantType.QInt8,
+    )
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()
--- a/egs/gigaspeech/ASR/zipformer/export.py
+++ b/egs/gigaspeech/ASR/zipformer/export.py
@ -0,0 +1,522 @@
+#!/usr/bin/env python3
+#
+# Copyright 2021-2023 Xiaomi Corporation (Author: Fangjun Kuang,
+#                                                 Zengwei Yao,
+#                                                 Wei Kang)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+# This script converts several saved checkpoints
+# to a single one using model averaging.
+"""
+
+Usage:
+
+Note: This is a example for gigaspeech dataset, if you are using different
+dataset, you should change the argument values according to your dataset.
+
+(1) Export to torchscript model using torch.jit.script()
+
+- For non-streaming model:
+
+./zipformer/export.py \
+  --exp-dir ./zipformer/exp \
+  --tokens data/lang_bpe_500/tokens.txt \
+  --epoch 30 \
+  --avg 9 \
+  --jit 1
+
+It will generate a file `jit_script.pt` in the given `exp_dir`. You can later
+load it by `torch.jit.load("jit_script.pt")`.
+
+Check ./jit_pretrained.py for its usage.
+
+Check https://github.com/k2-fsa/sherpa
+for how to use the exported models outside of icefall.
+
+- For streaming model:
+
+./zipformer/export.py \
+  --exp-dir ./zipformer/exp \
+  --causal 1 \
+  --chunk-size 16 \
+  --left-context-frames 128 \
+  --tokens data/lang_bpe_500/tokens.txt \
+  --epoch 30 \
+  --avg 9 \
+  --jit 1
+
+It will generate a file `jit_script_chunk_16_left_128.pt` in the given `exp_dir`.
+You can later load it by `torch.jit.load("jit_script_chunk_16_left_128.pt")`.
+
+Check ./jit_pretrained_streaming.py for its usage.
+
+Check https://github.com/k2-fsa/sherpa
+for how to use the exported models outside of icefall.
+
+(2) Export `model.state_dict()`
+
+- For non-streaming model:
+
+./zipformer/export.py \
+  --exp-dir ./zipformer/exp \
+  --tokens data/lang_bpe_500/tokens.txt \
+  --epoch 30 \
+  --avg 9
+
+- For streaming model:
+
+./zipformer/export.py \
+  --exp-dir ./zipformer/exp \
+  --causal 1 \
+  --tokens data/lang_bpe_500/tokens.txt \
+  --epoch 30 \
+  --avg 9
+
+It will generate a file `pretrained.pt` in the given `exp_dir`. You can later
+load it by `icefall.checkpoint.load_checkpoint()`.
+
+- For non-streaming model:
+
+To use the generated file with `zipformer/decode.py`,
+you can do:
+
+    cd /path/to/exp_dir
+    ln -s pretrained.pt epoch-9999.pt
+
+    cd /path/to/egs/gigaspeech/ASR
+    ./zipformer/decode.py \
+        --exp-dir ./zipformer/exp \
+        --epoch 9999 \
+        --avg 1 \
+        --max-duration 600 \
+        --decoding-method greedy_search \
+        --bpe-model data/lang_bpe_500/bpe.model
+
+- For streaming model:
+
+To use the generated file with `zipformer/decode.py` and `zipformer/streaming_decode.py`, you can do:
+
+    cd /path/to/exp_dir
+    ln -s pretrained.pt epoch-9999.pt
+
+    cd /path/to/egs/gigaspeech/ASR
+
+    # simulated streaming decoding
+    ./zipformer/decode.py \
+        --exp-dir ./zipformer/exp \
+        --epoch 9999 \
+        --avg 1 \
+        --max-duration 600 \
+        --causal 1 \
+        --chunk-size 16 \
+        --left-context-frames 128 \
+        --decoding-method greedy_search \
+        --bpe-model data/lang_bpe_500/bpe.model
+
+    # chunk-wise streaming decoding
+    ./zipformer/streaming_decode.py \
+        --exp-dir ./zipformer/exp \
+        --epoch 9999 \
+        --avg 1 \
+        --max-duration 600 \
+        --causal 1 \
+        --chunk-size 16 \
+        --left-context-frames 128 \
+        --decoding-method greedy_search \
+        --bpe-model data/lang_bpe_500/bpe.model
+
+Check ./pretrained.py for its usage.
+
+Note: If you don't want to train a model from scratch, we have
+provided one for you. You can get it at
+
+- non-streaming model:
+https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
+
+with the following commands:
+
+    sudo apt-get install git-lfs
+    git lfs install
+    git clone https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
+    # You will find the pre-trained models in exp dir
+"""
+
+import argparse
+import logging
+from pathlib import Path
+from typing import List, Tuple
+
+import k2
+import torch
+from scaling_converter import convert_scaled_to_non_scaled
+from torch import Tensor, nn
+from train import add_model_arguments, get_model, get_params
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.utils import make_pad_mask, num_tokens, str2bool
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=30,
+        help="""It specifies the checkpoint to use for decoding.
+        Note: Epoch counts from 1.
+        You can specify --avg to use more checkpoints for model averaging.""",
+    )
+
+    parser.add_argument(
+        "--iter",
+        type=int,
+        default=0,
+        help="""If positive, --epoch is ignored and it
+        will use the checkpoint exp_dir/checkpoint-iter.pt.
+        You can specify --avg to use more checkpoints for model averaging.
+        """,
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=9,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=True,
+        help="Whether to load averaged model. Currently it only supports "
+        "using --epoch. If True, it would decode with the averaged model "
+        "over the epoch range from `epoch-avg` (excluded) to `epoch`."
+        "Actually only the models with epoch number of `epoch-avg` and "
+        "`epoch` are loaded for averaging. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="zipformer/exp",
+        help="""It specifies the directory where all training related
+        files, e.g., checkpoints, log, etc, are saved
+        """,
+    )
+
+    parser.add_argument(
+        "--tokens",
+        type=str,
+        default="data/lang_bpe_500/tokens.txt",
+        help="Path to the tokens.txt",
+    )
+
+    parser.add_argument(
+        "--jit",
+        type=str2bool,
+        default=False,
+        help="""True to save a model after applying torch.jit.script.
+        It will generate a file named jit_script.pt.
+        Check ./jit_pretrained.py for how to use it.
+        """,
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+class EncoderModel(nn.Module):
+    """A wrapper for encoder and encoder_embed"""
+
+    def __init__(self, encoder: nn.Module, encoder_embed: nn.Module) -> None:
+        super().__init__()
+        self.encoder = encoder
+        self.encoder_embed = encoder_embed
+
+    def forward(
+        self, features: Tensor, feature_lengths: Tensor
+    ) -> Tuple[Tensor, Tensor]:
+        """
+        Args:
+            features: (N, T, C)
+            feature_lengths: (N,)
+        """
+        x, x_lens = self.encoder_embed(features, feature_lengths)
+
+        src_key_padding_mask = make_pad_mask(x_lens)
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+
+        encoder_out, encoder_out_lens = self.encoder(x, x_lens, src_key_padding_mask)
+        encoder_out = encoder_out.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+        return encoder_out, encoder_out_lens
+
+
+class StreamingEncoderModel(nn.Module):
+    """A wrapper for encoder and encoder_embed"""
+
+    def __init__(self, encoder: nn.Module, encoder_embed: nn.Module) -> None:
+        super().__init__()
+        assert len(encoder.chunk_size) == 1, encoder.chunk_size
+        assert len(encoder.left_context_frames) == 1, encoder.left_context_frames
+        self.chunk_size = encoder.chunk_size[0]
+        self.left_context_len = encoder.left_context_frames[0]
+
+        # The encoder_embed subsample features (T - 7) // 2
+        # The ConvNeXt module needs (7 - 1) // 2 = 3 frames of right padding after subsampling
+        self.pad_length = 7 + 2 * 3
+
+        self.encoder = encoder
+        self.encoder_embed = encoder_embed
+
+    def forward(
+        self, features: Tensor, feature_lengths: Tensor, states: List[Tensor]
+    ) -> Tuple[Tensor, Tensor, List[Tensor]]:
+        """Streaming forward for encoder_embed and encoder.
+
+        Args:
+            features: (N, T, C)
+            feature_lengths: (N,)
+            states: a list of Tensors
+
+        Returns encoder outputs, output lengths, and updated states.
+        """
+        chunk_size = self.chunk_size
+        left_context_len = self.left_context_len
+
+        cached_embed_left_pad = states[-2]
+        x, x_lens, new_cached_embed_left_pad = self.encoder_embed.streaming_forward(
+            x=features,
+            x_lens=feature_lengths,
+            cached_left_pad=cached_embed_left_pad,
+        )
+        assert x.size(1) == chunk_size, (x.size(1), chunk_size)
+
+        src_key_padding_mask = make_pad_mask(x_lens)
+
+        # processed_mask is used to mask out initial states
+        processed_mask = torch.arange(left_context_len, device=x.device).expand(
+            x.size(0), left_context_len
+        )
+        processed_lens = states[-1]  # (batch,)
+        # (batch, left_context_size)
+        processed_mask = (processed_lens.unsqueeze(1) <= processed_mask).flip(1)
+        # Update processed lengths
+        new_processed_lens = processed_lens + x_lens
+
+        # (batch, left_context_size + chunk_size)
+        src_key_padding_mask = torch.cat([processed_mask, src_key_padding_mask], dim=1)
+
+        x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+        encoder_states = states[:-2]
+
+        (
+            encoder_out,
+            encoder_out_lens,
+            new_encoder_states,
+        ) = self.encoder.streaming_forward(
+            x=x,
+            x_lens=x_lens,
+            states=encoder_states,
+            src_key_padding_mask=src_key_padding_mask,
+        )
+        encoder_out = encoder_out.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+        new_states = new_encoder_states + [
+            new_cached_embed_left_pad,
+            new_processed_lens,
+        ]
+        return encoder_out, encoder_out_lens, new_states
+
+    @torch.jit.export
+    def get_init_states(
+        self,
+        batch_size: int = 1,
+        device: torch.device = torch.device("cpu"),
+    ) -> List[torch.Tensor]:
+        """
+        Returns a list of cached tensors of all encoder layers. For layer-i, states[i*6:(i+1)*6]
+        is (cached_key, cached_nonlin_attn, cached_val1, cached_val2, cached_conv1, cached_conv2).
+        states[-2] is the cached left padding for ConvNeXt module,
+        of shape (batch_size, num_channels, left_pad, num_freqs)
+        states[-1] is processed_lens of shape (batch,), which records the number
+        of processed frames (at 50hz frame rate, after encoder_embed) for each sample in batch.
+        """
+        states = self.encoder.get_init_states(batch_size, device)
+
+        embed_states = self.encoder_embed.get_init_states(batch_size, device)
+        states.append(embed_states)
+
+        processed_lens = torch.zeros(batch_size, dtype=torch.int32, device=device)
+        states.append(processed_lens)
+
+        return states
+
+
+@torch.no_grad()
+def main():
+    args = get_parser().parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    device = torch.device("cpu")
+    # if torch.cuda.is_available():
+    #     device = torch.device("cuda", 0)
+
+    logging.info(f"device: {device}")
+
+    token_table = k2.SymbolTable.from_file(params.tokens)
+    params.blank_id = token_table["<blk>"]
+    params.vocab_size = num_tokens(token_table) + 1
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_model(params)
+
+    if not params.use_averaged_model:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            logging.info(f"averaging {filenames}")
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+        elif params.avg == 1:
+            load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+        else:
+            start = params.epoch - params.avg + 1
+            filenames = []
+            for i in range(start, params.epoch + 1):
+                if i >= 1:
+                    filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+            logging.info(f"averaging {filenames}")
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+    else:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg + 1
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg + 1:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            filename_start = filenames[-1]
+            filename_end = filenames[0]
+            logging.info(
+                "Calculating the averaged model over iteration checkpoints"
+                f" from {filename_start} (excluded) to {filename_end}"
+            )
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+        elif params.avg == 1:
+            load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+        else:
+            assert params.avg > 0, params.avg
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+            filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+            logging.info(
+                f"Calculating the averaged model over epoch range from "
+                f"{start} (excluded) to {params.epoch}"
+            )
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+
+    model.eval()
+
+    if params.jit is True:
+        convert_scaled_to_non_scaled(model, inplace=True)
+        # We won't use the forward() method of the model in C++, so just ignore
+        # it here.
+        # Otherwise, one of its arguments is a ragged tensor and is not
+        # torch scriptabe.
+        model.__class__.forward = torch.jit.ignore(model.__class__.forward)
+
+        # Wrap encoder and encoder_embed as a module
+        if params.causal:
+            model.encoder = StreamingEncoderModel(model.encoder, model.encoder_embed)
+            chunk_size = model.encoder.chunk_size
+            left_context_len = model.encoder.left_context_len
+            filename = f"jit_script_chunk_{chunk_size}_left_{left_context_len}.pt"
+        else:
+            model.encoder = EncoderModel(model.encoder, model.encoder_embed)
+            filename = "jit_script.pt"
+
+        logging.info("Using torch.jit.script")
+        model = torch.jit.script(model)
+        model.save(str(params.exp_dir / filename))
+        logging.info(f"Saved to {filename}")
+    else:
+        logging.info("Not using torchscript. Export model.state_dict()")
+        # Save it using a format so that it can be loaded
+        # by :func:`load_checkpoint`
+        filename = params.exp_dir / "pretrained.pt"
+        torch.save({"model": model.state_dict()}, str(filename))
+        logging.info(f"Saved to {filename}")
+
+
+if __name__ == "__main__":
+    formatter = "%(asctime)s %(levelname)s [%(filename)s:%(lineno)d] %(message)s"
+
+    logging.basicConfig(format=formatter, level=logging.INFO)
+    main()
--- a/egs/gigaspeech/ASR/zipformer/gigaspeech_scoring.py
+++ b/egs/gigaspeech/ASR/zipformer/gigaspeech_scoring.py
@ -0,0 +1 @@
+../conformer_ctc/gigaspeech_scoring.py
--- a/egs/gigaspeech/ASR/zipformer/jit_pretrained.py
+++ b/egs/gigaspeech/ASR/zipformer/jit_pretrained.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/jit_pretrained.py
--- a/egs/gigaspeech/ASR/zipformer/jit_pretrained_ctc.py
+++ b/egs/gigaspeech/ASR/zipformer/jit_pretrained_ctc.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/jit_pretrained_ctc.py
--- a/egs/gigaspeech/ASR/zipformer/jit_pretrained_streaming.py
+++ b/egs/gigaspeech/ASR/zipformer/jit_pretrained_streaming.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/jit_pretrained_streaming.py
--- a/egs/gigaspeech/ASR/zipformer/joiner.py
+++ b/egs/gigaspeech/ASR/zipformer/joiner.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/joiner.py
--- a/egs/gigaspeech/ASR/zipformer/model.py
+++ b/egs/gigaspeech/ASR/zipformer/model.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/model.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_check.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_check.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_check.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_decode.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_decode.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_decode.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_pretrained-streaming.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_pretrained-streaming.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_pretrained-streaming.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_pretrained.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_pretrained.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_pretrained.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_pretrained_ctc.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc_H.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc_H.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_pretrained_ctc_H.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc_HL.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc_HL.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_pretrained_ctc_HL.py
--- a/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py
+++ b/egs/gigaspeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/onnx_pretrained_ctc_HLG.py
--- a/egs/gigaspeech/ASR/zipformer/optim.py
+++ b/egs/gigaspeech/ASR/zipformer/optim.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/optim.py
--- a/egs/gigaspeech/ASR/zipformer/pretrained.py
+++ b/egs/gigaspeech/ASR/zipformer/pretrained.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/pretrained.py
--- a/egs/gigaspeech/ASR/zipformer/pretrained_ctc.py
+++ b/egs/gigaspeech/ASR/zipformer/pretrained_ctc.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/pretrained_ctc.py
--- a/egs/gigaspeech/ASR/zipformer/profile.py
+++ b/egs/gigaspeech/ASR/zipformer/profile.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/profile.py
--- a/egs/gigaspeech/ASR/zipformer/scaling.py
+++ b/egs/gigaspeech/ASR/zipformer/scaling.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/scaling.py
--- a/egs/gigaspeech/ASR/zipformer/scaling_converter.py
+++ b/egs/gigaspeech/ASR/zipformer/scaling_converter.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/scaling_converter.py
--- a/egs/gigaspeech/ASR/zipformer/streaming_beam_search.py
+++ b/egs/gigaspeech/ASR/zipformer/streaming_beam_search.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/streaming_beam_search.py
--- a/egs/gigaspeech/ASR/zipformer/streaming_decode.py
+++ b/egs/gigaspeech/ASR/zipformer/streaming_decode.py
@ -0,0 +1,853 @@
+#!/usr/bin/env python3
+# Copyright 2022-2023 Xiaomi Corporation (Authors: Wei Kang,
+#                                                  Fangjun Kuang,
+#                                                  Zengwei Yao)
+#
+# See ../../../../LICENSE for clarification regarding multiple authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+"""
+Usage:
+./zipformer/streaming_decode.py \
+  --epoch 28 \
+  --avg 15 \
+  --causal 1 \
+  --chunk-size 32 \
+  --left-context-frames 256 \
+  --exp-dir ./zipformer/exp \
+  --decoding-method greedy_search \
+  --num-decode-streams 2000
+"""
+
+import argparse
+import logging
+import math
+from pathlib import Path
+from typing import Dict, List, Optional, Tuple
+
+import k2
+import numpy as np
+import sentencepiece as spm
+import torch
+from asr_datamodule import GigaSpeechAsrDataModule
+from decode_stream import DecodeStream
+from kaldifeat import Fbank, FbankOptions
+from lhotse import CutSet
+from streaming_beam_search import (
+    fast_beam_search_one_best,
+    greedy_search,
+    modified_beam_search,
+)
+from torch import Tensor, nn
+from torch.nn.utils.rnn import pad_sequence
+from train import add_model_arguments, get_params, get_model
+
+from icefall.checkpoint import (
+    average_checkpoints,
+    average_checkpoints_with_averaged_model,
+    find_checkpoints,
+    load_checkpoint,
+)
+from icefall.utils import (
+    AttributeDict,
+    make_pad_mask,
+    setup_logger,
+    store_transcripts,
+    str2bool,
+    write_error_stats,
+)
+
+LOG_EPS = math.log(1e-10)
+
+
+def get_parser():
+    parser = argparse.ArgumentParser(
+        formatter_class=argparse.ArgumentDefaultsHelpFormatter
+    )
+
+    parser.add_argument(
+        "--epoch",
+        type=int,
+        default=28,
+        help="""It specifies the checkpoint to use for decoding.
+        Note: Epoch counts from 1.
+        You can specify --avg to use more checkpoints for model averaging.""",
+    )
+
+    parser.add_argument(
+        "--iter",
+        type=int,
+        default=0,
+        help="""If positive, --epoch is ignored and it
+        will use the checkpoint exp_dir/checkpoint-iter.pt.
+        You can specify --avg to use more checkpoints for model averaging.
+        """,
+    )
+
+    parser.add_argument(
+        "--avg",
+        type=int,
+        default=15,
+        help="Number of checkpoints to average. Automatically select "
+        "consecutive checkpoints before the checkpoint specified by "
+        "'--epoch' and '--iter'",
+    )
+
+    parser.add_argument(
+        "--use-averaged-model",
+        type=str2bool,
+        default=True,
+        help="Whether to load averaged model. Currently it only supports "
+        "using --epoch. If True, it would decode with the averaged model "
+        "over the epoch range from `epoch-avg` (excluded) to `epoch`."
+        "Actually only the models with epoch number of `epoch-avg` and "
+        "`epoch` are loaded for averaging. ",
+    )
+
+    parser.add_argument(
+        "--exp-dir",
+        type=str,
+        default="zipformer/exp",
+        help="The experiment dir",
+    )
+
+    parser.add_argument(
+        "--bpe-model",
+        type=str,
+        default="data/lang_bpe_500/bpe.model",
+        help="Path to the BPE model",
+    )
+
+    parser.add_argument(
+        "--decoding-method",
+        type=str,
+        default="greedy_search",
+        help="""Supported decoding methods are:
+        greedy_search
+        modified_beam_search
+        fast_beam_search
+        """,
+    )
+
+    parser.add_argument(
+        "--num_active_paths",
+        type=int,
+        default=4,
+        help="""An interger indicating how many candidates we will keep for each
+        frame. Used only when --decoding-method is modified_beam_search.""",
+    )
+
+    parser.add_argument(
+        "--beam",
+        type=float,
+        default=4,
+        help="""A floating point value to calculate the cutoff score during beam
+        search (i.e., `cutoff = max-score - beam`), which is the same as the
+        `beam` in Kaldi.
+        Used only when --decoding-method is fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--max-contexts",
+        type=int,
+        default=4,
+        help="""Used only when --decoding-method is
+        fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--max-states",
+        type=int,
+        default=32,
+        help="""Used only when --decoding-method is
+        fast_beam_search""",
+    )
+
+    parser.add_argument(
+        "--context-size",
+        type=int,
+        default=2,
+        help="The context size in the decoder. 1 means bigram; 2 means tri-gram",
+    )
+
+    parser.add_argument(
+        "--num-decode-streams",
+        type=int,
+        default=2000,
+        help="The number of streams that can be decoded parallel.",
+    )
+
+    add_model_arguments(parser)
+
+    return parser
+
+
+def get_init_states(
+    model: nn.Module,
+    batch_size: int = 1,
+    device: torch.device = torch.device("cpu"),
+) -> List[torch.Tensor]:
+    """
+    Returns a list of cached tensors of all encoder layers. For layer-i, states[i*6:(i+1)*6]
+    is (cached_key, cached_nonlin_attn, cached_val1, cached_val2, cached_conv1, cached_conv2).
+    states[-2] is the cached left padding for ConvNeXt module,
+    of shape (batch_size, num_channels, left_pad, num_freqs)
+    states[-1] is processed_lens of shape (batch,), which records the number
+    of processed frames (at 50hz frame rate, after encoder_embed) for each sample in batch.
+    """
+    states = model.encoder.get_init_states(batch_size, device)
+
+    embed_states = model.encoder_embed.get_init_states(batch_size, device)
+    states.append(embed_states)
+
+    processed_lens = torch.zeros(batch_size, dtype=torch.int32, device=device)
+    states.append(processed_lens)
+
+    return states
+
+
+def stack_states(state_list: List[List[torch.Tensor]]) -> List[torch.Tensor]:
+    """Stack list of zipformer states that correspond to separate utterances
+    into a single emformer state, so that it can be used as an input for
+    zipformer when those utterances are formed into a batch.
+
+    Args:
+      state_list:
+        Each element in state_list corresponding to the internal state
+        of the zipformer model for a single utterance. For element-n,
+        state_list[n] is a list of cached tensors of all encoder layers. For layer-i,
+        state_list[n][i*6:(i+1)*6] is (cached_key, cached_nonlin_attn, cached_val1,
+        cached_val2, cached_conv1, cached_conv2).
+        state_list[n][-2] is the cached left padding for ConvNeXt module,
+          of shape (batch_size, num_channels, left_pad, num_freqs)
+        state_list[n][-1] is processed_lens of shape (batch,), which records the number
+        of processed frames (at 50hz frame rate, after encoder_embed) for each sample in batch.
+
+    Note:
+      It is the inverse of :func:`unstack_states`.
+    """
+    batch_size = len(state_list)
+    assert (len(state_list[0]) - 2) % 6 == 0, len(state_list[0])
+    tot_num_layers = (len(state_list[0]) - 2) // 6
+
+    batch_states = []
+    for layer in range(tot_num_layers):
+        layer_offset = layer * 6
+        # cached_key: (left_context_len, batch_size, key_dim)
+        cached_key = torch.cat(
+            [state_list[i][layer_offset] for i in range(batch_size)], dim=1
+        )
+        # cached_nonlin_attn: (num_heads, batch_size, left_context_len, head_dim)
+        cached_nonlin_attn = torch.cat(
+            [state_list[i][layer_offset + 1] for i in range(batch_size)], dim=1
+        )
+        # cached_val1: (left_context_len, batch_size, value_dim)
+        cached_val1 = torch.cat(
+            [state_list[i][layer_offset + 2] for i in range(batch_size)], dim=1
+        )
+        # cached_val2: (left_context_len, batch_size, value_dim)
+        cached_val2 = torch.cat(
+            [state_list[i][layer_offset + 3] for i in range(batch_size)], dim=1
+        )
+        # cached_conv1: (#batch, channels, left_pad)
+        cached_conv1 = torch.cat(
+            [state_list[i][layer_offset + 4] for i in range(batch_size)], dim=0
+        )
+        # cached_conv2: (#batch, channels, left_pad)
+        cached_conv2 = torch.cat(
+            [state_list[i][layer_offset + 5] for i in range(batch_size)], dim=0
+        )
+        batch_states += [
+            cached_key,
+            cached_nonlin_attn,
+            cached_val1,
+            cached_val2,
+            cached_conv1,
+            cached_conv2,
+        ]
+
+    cached_embed_left_pad = torch.cat(
+        [state_list[i][-2] for i in range(batch_size)], dim=0
+    )
+    batch_states.append(cached_embed_left_pad)
+
+    processed_lens = torch.cat([state_list[i][-1] for i in range(batch_size)], dim=0)
+    batch_states.append(processed_lens)
+
+    return batch_states
+
+
+def unstack_states(batch_states: List[Tensor]) -> List[List[Tensor]]:
+    """Unstack the zipformer state corresponding to a batch of utterances
+    into a list of states, where the i-th entry is the state from the i-th
+    utterance in the batch.
+
+    Note:
+      It is the inverse of :func:`stack_states`.
+
+    Args:
+        batch_states: A list of cached tensors of all encoder layers. For layer-i,
+          states[i*6:(i+1)*6] is (cached_key, cached_nonlin_attn, cached_val1, cached_val2,
+          cached_conv1, cached_conv2).
+          state_list[-2] is the cached left padding for ConvNeXt module,
+          of shape (batch_size, num_channels, left_pad, num_freqs)
+          states[-1] is processed_lens of shape (batch,), which records the number
+          of processed frames (at 50hz frame rate, after encoder_embed) for each sample in batch.
+
+    Returns:
+        state_list: A list of list. Each element in state_list corresponding to the internal state
+        of the zipformer model for a single utterance.
+    """
+    assert (len(batch_states) - 2) % 6 == 0, len(batch_states)
+    tot_num_layers = (len(batch_states) - 2) // 6
+
+    processed_lens = batch_states[-1]
+    batch_size = processed_lens.shape[0]
+
+    state_list = [[] for _ in range(batch_size)]
+
+    for layer in range(tot_num_layers):
+        layer_offset = layer * 6
+        # cached_key: (left_context_len, batch_size, key_dim)
+        cached_key_list = batch_states[layer_offset].chunk(chunks=batch_size, dim=1)
+        # cached_nonlin_attn: (num_heads, batch_size, left_context_len, head_dim)
+        cached_nonlin_attn_list = batch_states[layer_offset + 1].chunk(
+            chunks=batch_size, dim=1
+        )
+        # cached_val1: (left_context_len, batch_size, value_dim)
+        cached_val1_list = batch_states[layer_offset + 2].chunk(
+            chunks=batch_size, dim=1
+        )
+        # cached_val2: (left_context_len, batch_size, value_dim)
+        cached_val2_list = batch_states[layer_offset + 3].chunk(
+            chunks=batch_size, dim=1
+        )
+        # cached_conv1: (#batch, channels, left_pad)
+        cached_conv1_list = batch_states[layer_offset + 4].chunk(
+            chunks=batch_size, dim=0
+        )
+        # cached_conv2: (#batch, channels, left_pad)
+        cached_conv2_list = batch_states[layer_offset + 5].chunk(
+            chunks=batch_size, dim=0
+        )
+        for i in range(batch_size):
+            state_list[i] += [
+                cached_key_list[i],
+                cached_nonlin_attn_list[i],
+                cached_val1_list[i],
+                cached_val2_list[i],
+                cached_conv1_list[i],
+                cached_conv2_list[i],
+            ]
+
+    cached_embed_left_pad_list = batch_states[-2].chunk(chunks=batch_size, dim=0)
+    for i in range(batch_size):
+        state_list[i].append(cached_embed_left_pad_list[i])
+
+    processed_lens_list = batch_states[-1].chunk(chunks=batch_size, dim=0)
+    for i in range(batch_size):
+        state_list[i].append(processed_lens_list[i])
+
+    return state_list
+
+
+def streaming_forward(
+    features: Tensor,
+    feature_lens: Tensor,
+    model: nn.Module,
+    states: List[Tensor],
+    chunk_size: int,
+    left_context_len: int,
+) -> Tuple[Tensor, Tensor, List[Tensor]]:
+    """
+    Returns encoder outputs, output lengths, and updated states.
+    """
+    cached_embed_left_pad = states[-2]
+    (x, x_lens, new_cached_embed_left_pad,) = model.encoder_embed.streaming_forward(
+        x=features,
+        x_lens=feature_lens,
+        cached_left_pad=cached_embed_left_pad,
+    )
+    assert x.size(1) == chunk_size, (x.size(1), chunk_size)
+
+    src_key_padding_mask = make_pad_mask(x_lens)
+
+    # processed_mask is used to mask out initial states
+    processed_mask = torch.arange(left_context_len, device=x.device).expand(
+        x.size(0), left_context_len
+    )
+    processed_lens = states[-1]  # (batch,)
+    # (batch, left_context_size)
+    processed_mask = (processed_lens.unsqueeze(1) <= processed_mask).flip(1)
+    # Update processed lengths
+    new_processed_lens = processed_lens + x_lens
+
+    # (batch, left_context_size + chunk_size)
+    src_key_padding_mask = torch.cat([processed_mask, src_key_padding_mask], dim=1)
+
+    x = x.permute(1, 0, 2)  # (N, T, C) -> (T, N, C)
+    encoder_states = states[:-2]
+    (
+        encoder_out,
+        encoder_out_lens,
+        new_encoder_states,
+    ) = model.encoder.streaming_forward(
+        x=x,
+        x_lens=x_lens,
+        states=encoder_states,
+        src_key_padding_mask=src_key_padding_mask,
+    )
+    encoder_out = encoder_out.permute(1, 0, 2)  # (T, N, C) ->(N, T, C)
+
+    new_states = new_encoder_states + [
+        new_cached_embed_left_pad,
+        new_processed_lens,
+    ]
+    return encoder_out, encoder_out_lens, new_states
+
+
+def decode_one_chunk(
+    params: AttributeDict,
+    model: nn.Module,
+    decode_streams: List[DecodeStream],
+) -> List[int]:
+    """Decode one chunk frames of features for each decode_streams and
+    return the indexes of finished streams in a List.
+
+    Args:
+      params:
+        It's the return value of :func:`get_params`.
+      model:
+        The neural model.
+      decode_streams:
+        A List of DecodeStream, each belonging to a utterance.
+    Returns:
+      Return a List containing which DecodeStreams are finished.
+    """
+    device = model.device
+    chunk_size = int(params.chunk_size)
+    left_context_len = int(params.left_context_frames)
+
+    features = []
+    feature_lens = []
+    states = []
+    processed_lens = []  # Used in fast-beam-search
+
+    for stream in decode_streams:
+        feat, feat_len = stream.get_feature_frames(chunk_size * 2)
+        features.append(feat)
+        feature_lens.append(feat_len)
+        states.append(stream.states)
+        processed_lens.append(stream.done_frames)
+
+    feature_lens = torch.tensor(feature_lens, device=device)
+    features = pad_sequence(features, batch_first=True, padding_value=LOG_EPS)
+
+    # Make sure the length after encoder_embed is at least 1.
+    # The encoder_embed subsample features (T - 7) // 2
+    # The ConvNeXt module needs (7 - 1) // 2 = 3 frames of right padding after subsampling
+    tail_length = chunk_size * 2 + 7 + 2 * 3
+    if features.size(1) < tail_length:
+        pad_length = tail_length - features.size(1)
+        feature_lens += pad_length
+        features = torch.nn.functional.pad(
+            features,
+            (0, 0, 0, pad_length),
+            mode="constant",
+            value=LOG_EPS,
+        )
+
+    states = stack_states(states)
+
+    encoder_out, encoder_out_lens, new_states = streaming_forward(
+        features=features,
+        feature_lens=feature_lens,
+        model=model,
+        states=states,
+        chunk_size=chunk_size,
+        left_context_len=left_context_len,
+    )
+
+    encoder_out = model.joiner.encoder_proj(encoder_out)
+
+    if params.decoding_method == "greedy_search":
+        greedy_search(model=model, encoder_out=encoder_out, streams=decode_streams)
+    elif params.decoding_method == "fast_beam_search":
+        processed_lens = torch.tensor(processed_lens, device=device)
+        processed_lens = processed_lens + encoder_out_lens
+        fast_beam_search_one_best(
+            model=model,
+            encoder_out=encoder_out,
+            processed_lens=processed_lens,
+            streams=decode_streams,
+            beam=params.beam,
+            max_states=params.max_states,
+            max_contexts=params.max_contexts,
+        )
+    elif params.decoding_method == "modified_beam_search":
+        modified_beam_search(
+            model=model,
+            streams=decode_streams,
+            encoder_out=encoder_out,
+            num_active_paths=params.num_active_paths,
+        )
+    else:
+        raise ValueError(f"Unsupported decoding method: {params.decoding_method}")
+
+    states = unstack_states(new_states)
+
+    finished_streams = []
+    for i in range(len(decode_streams)):
+        decode_streams[i].states = states[i]
+        decode_streams[i].done_frames += encoder_out_lens[i]
+        if decode_streams[i].done:
+            finished_streams.append(i)
+
+    return finished_streams
+
+
+def decode_dataset(
+    cuts: CutSet,
+    params: AttributeDict,
+    model: nn.Module,
+    sp: spm.SentencePieceProcessor,
+    decoding_graph: Optional[k2.Fsa] = None,
+) -> Dict[str, List[Tuple[List[str], List[str]]]]:
+    """Decode dataset.
+
+    Args:
+      cuts:
+        Lhotse Cutset containing the dataset to decode.
+      params:
+        It is returned by :func:`get_params`.
+      model:
+        The neural model.
+      sp:
+        The BPE model.
+      decoding_graph:
+        The decoding graph. Can be either a `k2.trivial_graph` or HLG, Used
+        only when --decoding_method is fast_beam_search.
+    Returns:
+      Return a dict, whose key may be "greedy_search" if greedy search
+      is used, or it may be "beam_7" if beam size of 7 is used.
+      Its value is a list of tuples. Each tuple contains two elements:
+      The first is the reference transcript, and the second is the
+      predicted result.
+    """
+    device = model.device
+
+    opts = FbankOptions()
+    opts.device = device
+    opts.frame_opts.dither = 0
+    opts.frame_opts.snip_edges = False
+    opts.frame_opts.samp_freq = 16000
+    opts.mel_opts.num_bins = 80
+
+    log_interval = 100
+
+    decode_results = []
+    # Contain decode streams currently running.
+    decode_streams = []
+    for num, cut in enumerate(cuts):
+        # each utterance has a DecodeStream.
+        initial_states = get_init_states(model=model, batch_size=1, device=device)
+        decode_stream = DecodeStream(
+            params=params,
+            cut_id=cut.id,
+            initial_states=initial_states,
+            decoding_graph=decoding_graph,
+            device=device,
+        )
+
+        audio: np.ndarray = cut.load_audio()
+        # audio.shape: (1, num_samples)
+        assert len(audio.shape) == 2
+        assert audio.shape[0] == 1, "Should be single channel"
+        assert audio.dtype == np.float32, audio.dtype
+
+        # The trained model is using normalized samples
+        assert audio.max() <= 1, "Should be normalized to [-1, 1])"
+
+        samples = torch.from_numpy(audio).squeeze(0)
+
+        fbank = Fbank(opts)
+        feature = fbank(samples.to(device))
+        decode_stream.set_features(feature, tail_pad_len=30)
+        decode_stream.ground_truth = cut.supervisions[0].text
+
+        decode_streams.append(decode_stream)
+
+        while len(decode_streams) >= params.num_decode_streams:
+            finished_streams = decode_one_chunk(
+                params=params, model=model, decode_streams=decode_streams
+            )
+            for i in sorted(finished_streams, reverse=True):
+                decode_results.append(
+                    (
+                        decode_streams[i].id,
+                        decode_streams[i].ground_truth.split(),
+                        sp.decode(decode_streams[i].decoding_result()).split(),
+                    )
+                )
+                del decode_streams[i]
+
+        if num % log_interval == 0:
+            logging.info(f"Cuts processed until now is {num}.")
+
+    # decode final chunks of last sequences
+    while len(decode_streams):
+        finished_streams = decode_one_chunk(
+            params=params, model=model, decode_streams=decode_streams
+        )
+        for i in sorted(finished_streams, reverse=True):
+            decode_results.append(
+                (
+                    decode_streams[i].id,
+                    decode_streams[i].ground_truth.split(),
+                    sp.decode(decode_streams[i].decoding_result()).split(),
+                )
+            )
+            del decode_streams[i]
+
+    if params.decoding_method == "greedy_search":
+        key = "greedy_search"
+    elif params.decoding_method == "fast_beam_search":
+        key = (
+            f"beam_{params.beam}_"
+            f"max_contexts_{params.max_contexts}_"
+            f"max_states_{params.max_states}"
+        )
+    elif params.decoding_method == "modified_beam_search":
+        key = f"num_active_paths_{params.num_active_paths}"
+    else:
+        raise ValueError(f"Unsupported decoding method: {params.decoding_method}")
+    return {key: decode_results}
+
+
+def save_results(
+    params: AttributeDict,
+    test_set_name: str,
+    results_dict: Dict[str, List[Tuple[List[str], List[str]]]],
+):
+    test_set_wers = dict()
+    for key, results in results_dict.items():
+        recog_path = (
+            params.res_dir / f"recogs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        results = sorted(results)
+        store_transcripts(filename=recog_path, texts=results)
+        logging.info(f"The transcripts are stored in {recog_path}")
+
+        # The following prints out WERs, per-word error statistics and aligned
+        # ref/hyp pairs.
+        errs_filename = (
+            params.res_dir / f"errs-{test_set_name}-{key}-{params.suffix}.txt"
+        )
+        with open(errs_filename, "w") as f:
+            wer = write_error_stats(
+                f, f"{test_set_name}-{key}", results, enable_log=True
+            )
+            test_set_wers[key] = wer
+
+        logging.info("Wrote detailed error stats to {}".format(errs_filename))
+
+    test_set_wers = sorted(test_set_wers.items(), key=lambda x: x[1])
+    errs_info = (
+        params.res_dir / f"wer-summary-{test_set_name}-{key}-{params.suffix}.txt"
+    )
+    with open(errs_info, "w") as f:
+        print("settings\tWER", file=f)
+        for key, val in test_set_wers:
+            print("{}\t{}".format(key, val), file=f)
+
+    s = "\nFor {}, WER of different settings are:\n".format(test_set_name)
+    note = "\tbest for {}".format(test_set_name)
+    for key, val in test_set_wers:
+        s += "{}\t{}{}\n".format(key, val, note)
+        note = ""
+    logging.info(s)
+
+
+@torch.no_grad()
+def main():
+    parser = get_parser()
+    GigaSpeechAsrDataModule.add_arguments(parser)
+    args = parser.parse_args()
+    args.exp_dir = Path(args.exp_dir)
+
+    params = get_params()
+    params.update(vars(args))
+
+    params.res_dir = params.exp_dir / "streaming" / params.decoding_method
+
+    if params.iter > 0:
+        params.suffix = f"iter-{params.iter}-avg-{params.avg}"
+    else:
+        params.suffix = f"epoch-{params.epoch}-avg-{params.avg}"
+
+    assert params.causal, params.causal
+    assert "," not in params.chunk_size, "chunk_size should be one value in decoding."
+    assert (
+        "," not in params.left_context_frames
+    ), "left_context_frames should be one value in decoding."
+    params.suffix += f"-chunk-{params.chunk_size}"
+    params.suffix += f"-left-context-{params.left_context_frames}"
+
+    # for fast_beam_search
+    if params.decoding_method == "fast_beam_search":
+        params.suffix += f"-beam-{params.beam}"
+        params.suffix += f"-max-contexts-{params.max_contexts}"
+        params.suffix += f"-max-states-{params.max_states}"
+
+    if params.use_averaged_model:
+        params.suffix += "-use-averaged-model"
+
+    setup_logger(f"{params.res_dir}/log-decode-{params.suffix}")
+    logging.info("Decoding started")
+
+    device = torch.device("cpu")
+    if torch.cuda.is_available():
+        device = torch.device("cuda", 0)
+
+    logging.info(f"Device: {device}")
+
+    sp = spm.SentencePieceProcessor()
+    sp.load(params.bpe_model)
+
+    # <blk> and <unk> is defined in local/train_bpe_model.py
+    params.blank_id = sp.piece_to_id("<blk>")
+    params.unk_id = sp.piece_to_id("<unk>")
+    params.vocab_size = sp.get_piece_size()
+
+    logging.info(params)
+
+    logging.info("About to create model")
+    model = get_model(params)
+
+    if not params.use_averaged_model:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+        elif params.avg == 1:
+            load_checkpoint(f"{params.exp_dir}/epoch-{params.epoch}.pt", model)
+        else:
+            start = params.epoch - params.avg + 1
+            filenames = []
+            for i in range(start, params.epoch + 1):
+                if start >= 0:
+                    filenames.append(f"{params.exp_dir}/epoch-{i}.pt")
+            logging.info(f"averaging {filenames}")
+            model.to(device)
+            model.load_state_dict(average_checkpoints(filenames, device=device))
+    else:
+        if params.iter > 0:
+            filenames = find_checkpoints(params.exp_dir, iteration=-params.iter)[
+                : params.avg + 1
+            ]
+            if len(filenames) == 0:
+                raise ValueError(
+                    f"No checkpoints found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            elif len(filenames) < params.avg + 1:
+                raise ValueError(
+                    f"Not enough checkpoints ({len(filenames)}) found for"
+                    f" --iter {params.iter}, --avg {params.avg}"
+                )
+            filename_start = filenames[-1]
+            filename_end = filenames[0]
+            logging.info(
+                "Calculating the averaged model over iteration checkpoints"
+                f" from {filename_start} (excluded) to {filename_end}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+        else:
+            assert params.avg > 0, params.avg
+            start = params.epoch - params.avg
+            assert start >= 1, start
+            filename_start = f"{params.exp_dir}/epoch-{start}.pt"
+            filename_end = f"{params.exp_dir}/epoch-{params.epoch}.pt"
+            logging.info(
+                f"Calculating the averaged model over epoch range from "
+                f"{start} (excluded) to {params.epoch}"
+            )
+            model.to(device)
+            model.load_state_dict(
+                average_checkpoints_with_averaged_model(
+                    filename_start=filename_start,
+                    filename_end=filename_end,
+                    device=device,
+                )
+            )
+
+    model.to(device)
+    model.eval()
+    model.device = device
+
+    decoding_graph = None
+    if params.decoding_method == "fast_beam_search":
+        decoding_graph = k2.trivial_graph(params.vocab_size - 1, device=device)
+
+    num_param = sum([p.numel() for p in model.parameters()])
+    logging.info(f"Number of model parameters: {num_param}")
+
+    gigaspeech = GigaSpeechAsrDataModule(args)
+
+    dev_cuts = gigaspeech.dev_cuts()
+    test_cuts = gigaspeech.test_cuts()
+
+    test_sets = ["dev", "test"]
+    test_cuts = [dev_cuts, test_cuts]
+
+    for test_set, test_cut in zip(test_sets, test_cuts):
+        results_dict = decode_dataset(
+            cuts=test_cut,
+            params=params,
+            model=model,
+            sp=sp,
+            decoding_graph=decoding_graph,
+        )
+
+        save_results(
+            params=params,
+            test_set_name=test_set,
+            results_dict=results_dict,
+        )
+
+    logging.info("Done!")
+
+
+if __name__ == "__main__":
+    main()
--- a/egs/gigaspeech/ASR/zipformer/subsampling.py
+++ b/egs/gigaspeech/ASR/zipformer/subsampling.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/subsampling.py
--- a/egs/gigaspeech/ASR/zipformer/test_scaling.py
+++ b/egs/gigaspeech/ASR/zipformer/test_scaling.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/test_scaling.py
--- a/egs/gigaspeech/ASR/zipformer/test_subsampling.py
+++ b/egs/gigaspeech/ASR/zipformer/test_subsampling.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/test_subsampling.py
--- a/egs/gigaspeech/ASR/zipformer/train.py
+++ b/egs/gigaspeech/ASR/zipformer/train.py
--- a/egs/gigaspeech/ASR/zipformer/zipformer.py
+++ b/egs/gigaspeech/ASR/zipformer/zipformer.py
@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/zipformer.py
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/pruned_transducer_stateless2/beam_search.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/decode_stream.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/transducer_stateless/encoder_interface.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/export-onnx-ctc.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/jit_pretrained.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/joiner.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/onnx_check.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/pretrained.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/profile.py`
				`@ -0,0 +1 @@`
				`../../../librispeech/ASR/zipformer/scaling.py`