fix python path for ctm script

CoEDL · Oct 28, 2021 · f92e2d4 · f92e2d4
1 parent af7177c
commit f92e2d4
Show file tree

Hide file tree

Showing 4 changed files with 8 additions and 3 deletions.
diff --git a/elpis/engines/kaldi/inference/gmm-decode-conf/gmm-decode-conf.sh b/elpis/engines/kaldi/inference/gmm-decode-conf/gmm-decode-conf.sh
@@ -58,6 +58,7 @@
 # Make the split dir with scp, utt2spk etc files
 . $PWD/make_split.sh
 
+echo "==== Using GMM Decode Conf ===="
 
 # AUDIO --> FEATURE VECTORS
 echo "==== Extracting Feature Vectors ===="
@@ -115,7 +116,7 @@ echo >> data/infer/one-best-hypothesis.txt
 cat data/infer/one-best-hypothesis.txt
 
 echo "==== Build the Elan file ===="
-"${POETRY_PATH}/bin/python" /elpis/elpis/engines/common/output/ctm_to_elan.py \
+python3 /elpis/elpis/engines/common/output/ctm_to_elan.py \
     --ctm data/infer/ctm_with_conf.ctm \
     --wav data/infer/split1/1/wav.scp \
     --seg data/infer/split1/1/segments \

diff --git a/elpis/engines/kaldi/inference/gmm-decode-online-conf/0_feature_vec.sh b/elpis/engines/kaldi/inference/gmm-decode-online-conf/0_feature_vec.sh
@@ -7,6 +7,8 @@
 
 export PATH=$PATH:/kaldi/src/online2bin
 
+echo "==== Using GMM Decode Online Conf ===="
+
 # Extract feature vectors for online training
 echo "==== Extracting Feature Vectors ===="
 steps/make_mfcc.sh --nj 1 data/infer exp/make_mfcc/infer mfcc
diff --git a/elpis/engines/kaldi/inference/gmm-decode-online-conf/4_ctm_output.sh b/elpis/engines/kaldi/inference/gmm-decode-online-conf/4_ctm_output.sh
@@ -23,7 +23,7 @@ echo >> data/infer/one-best-hypothesis.txt
 cat data/infer/one-best-hypothesis.txt
 
 echo "==== Build the Elan file ===="
-"${POETRY_PATH}/bin/python" /elpis/elpis/engines/common/output/ctm_to_elan.py \
+python3 /elpis/elpis/engines/common/output/ctm_to_elan.py \
     --ctm data/infer/ctm_with_conf.ctm \
     --wav data/infer/split1/1/wav.scp \
     --seg data/infer/split1/1/segments \

diff --git a/elpis/engines/kaldi/objects/transcription.py b/elpis/engines/kaldi/objects/transcription.py
@@ -102,9 +102,10 @@ def transcribe(self, on_complete: Callable = None):
             exp_path.mkdir(parents=True, exist_ok=True)
 
         print("========= reset templates dir")
-        # TODO use gmm-decode for short audio and gmm-decode-online for long audio (gmm-decode is quicker)
+        # Use gmm-decode for short audio and gmm-decode-online for long audio (gmm-decode is quicker for short audio)
         # Stage names (rh side) are used in the GUI for i18n
         if (self.audio_duration > 10):
+            print("==== Using gmm-decode-online-conf")
             template_dir_path = 'gmm-decode-online-conf'
             stage_names = {
                 "0_feature_vec.sh": "featureExtraction",
@@ -114,6 +115,7 @@ def transcribe(self, on_complete: Callable = None):
                 "4_ctm_output.sh": "ctmOutput"
             }
         else:
+            print("==== Using gmm-decode-conf")
             template_dir_path = 'gmm-decode-conf'
             stage_names = {
                 "gmm-decode-conf.sh": "transcribing"