Merge branch 'ci-train-2' into ci-train

.github/scripts/docker/Dockerfile

@@ -1,4 +1,5 @@
 ARG PYTHON_VERSION=3.8
+FROM python:${PYTHON_VERSION}
 
 ARG TORCHAUDIO_VERSION="0.13.0"
 ARG TORCH_VERSION="1.13.0"
@@ -8,10 +9,9 @@ ARG KALDIFEAT_VERSION="1.25.3.dev20231221"
 ARG _K2_VERSION="${K2_VERSION}+cpu.torch${TORCH_VERSION}"
 ARG _KALDIFEAT_VERSION="${KALDIFEAT_VERSION}+cpu.torch${TORCH_VERSION}"
 
-FROM python:${PYTHON_VERSION}
-
 RUN apt-get update -y && \
     apt-get install -qq -y \
+    ffmpeg \
     git \
     git-lfs \
     less \

.github/scripts/docker/run.sh

@@ -3,6 +3,9 @@ set -ex
 
 cd /icefall
 export PYTHONPATH=/icefall:$PYTHONPATH
+python3 -c "import torch; print(torch.__file__)"
+python3 -c "import torchaudio; print(torchaudio.__version__)"
+python3 -c "import icefall; print(icefall.__file__)"
 
 cd egs/librispeech/ASR
 

.github/workflows/build-cpu-docker.yml

@@ -8,7 +8,7 @@ concurrency:
 
 jobs:
   build-cpu-docker:
-    name: py${{ matrix.python-version }} torch${{ matrix.torch-version }}
+    name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
@@ -18,6 +18,7 @@ jobs:
         torch-version: ["1.13.0", "2.0.0", "2.0.1", "2.1.0", "2.1.1", "2.1.2"]
         k2-version: ["1.24.4.dev20231220"]
         kaldifeat-version: ["1.25.3.dev20231221"]
+        version: ["1.0"]
 
     steps:
       # refer to https://github.com/actions/checkout
@@ -44,17 +45,22 @@ jobs:
         run: |
           cd .github/scripts/docker
           torch_version=${{ matrix.torch-version }}
-          if [[ $torch_version == "1.13.0" ]]; then
+          if [[ $torch_version == 1.13.0 ]]; then
             torchaudio_version=0.13.0
-          elif [[ $torch_version == "2.0.0" ]]; then
+          elif [[ $torch_version == 2.0.0 ]]; then
             torchaudio_version=2.0.1
-          elif [[ $torch_version == "2.0.1" ]]; then
+          elif [[ $torch_version == 2.0.1 ]]; then
-            torchaudio_version=2.0.2"
+            torchaudio_version=2.0.2
           else
             torchaudio_version=$torch_version
           fi
+          echo "torch_version: $torch_version"
+          echo "torchaudio_version: $torchaudio_version"
 
-          tag=ghcr.io/k2-fsa/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}
+          version=${{ matrix.version }}
+
+          tag=ghcr.io/k2-fsa/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v$version
+          echo "tag: $tag"
 
           docker build \
             -t $tag \

.github/workflows/train-librispeech.yml

@@ -16,13 +16,17 @@ concurrency:
 
 jobs:
   train-librispeech:
-    name: ${{ matrix.python-version }}
+    name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
     runs-on: ${{ matrix.os }}
     strategy:
       fail-fast: false
       matrix:
         os: [ubuntu-latest]
         python-version: ["3.8", "3.9", "3.10"]
+        torch-version: ["1.13.0", "2.0.0", "2.0.1", "2.1.0", "2.1.1", "2.1.2"]
+        k2-version: ["1.24.4.dev20231220"]
+        kaldifeat-version: ["1.25.3.dev20231221"]
+        version: ["1.0"]
 
     steps:
       # refer to https://github.com/actions/checkout
@@ -42,7 +46,7 @@ jobs:
       - name: Run the build process with Docker
         uses: addnab/docker-run-action@v3
         with:
-            image: ghcr.io/k2-fsa/icefall:cpu-py${{ matrix.python-version }}
+            image: ghcr.io/k2-fsa/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
             options: |
               --volume ${{ github.workspace }}/:/icefall
             shell: bash

egs/gigaspeech/ASR/zipformer/my_profile.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/my_profile.py

egs/gigaspeech/ASR/zipformer/profile.py

@@ -1 +0,0 @@
-../../../librispeech/ASR/zipformer/profile.py

egs/librispeech/ASR/pruned_transducer_stateless/my_profile.py

@@ -17,7 +17,7 @@
 # limitations under the License.
 
 """
-Usage: ./pruned_transducer_stateless/profile.py
+Usage: ./pruned_transducer_stateless/my_profile.py
 """
 
 import argparse

egs/librispeech/ASR/pruned_transducer_stateless4/my_profile.py

@@ -17,7 +17,7 @@
 # limitations under the License.
 
 """
-Usage: ./pruned_transducer_stateless4/profile.py
+Usage: ./pruned_transducer_stateless4/my_profile.py
 """
 
 import argparse

egs/librispeech/ASR/pruned_transducer_stateless7/my_profile.py

@@ -17,7 +17,7 @@
 # limitations under the License.
 
 """
-Usage: ./pruned_transducer_stateless7/profile.py
+Usage: ./pruned_transducer_stateless7/my_profile.py
 """
 
 import argparse

egs/librispeech/ASR/zipformer/my_profile.py

@@ -17,7 +17,7 @@
 # limitations under the License.
 
 """
-Usage: ./zipformer/profile.py
+Usage: ./zipformer/my_profile.py
 """
 
 import argparse

egs/tedlium3/ASR/zipformer/my_profile.py

@@ -0,0 +1 @@
+../../../librispeech/ASR/zipformer/my_profile.py

egs/tedlium3/ASR/zipformer/profile.py

@@ -1 +0,0 @@
-../../../librispeech/ASR/zipformer/profile.py