fix text_bert incorrect dimension in run_single()

thuiar · Sep 23, 2022 · bda2ace · bda2ace
1 parent 5e5c8ae
commit bda2ace
Show file tree

Hide file tree

Showing 3 changed files with 2 additions and 170 deletions.
diff --git a/.gitignore b/.gitignore
@@ -8,3 +8,4 @@ dist
 __pycache__
 exts
 tmp.csv
+build
diff --git a/src/MSA_FET/dataloader.py b/src/MSA_FET/dataloader.py
diff --git a/src/MSA_FET/single.py b/src/MSA_FET/single.py
@@ -36,8 +36,6 @@ class FeatureExtractionTool(object):
         [ ] Add csv/dataframe output format.
         [ ] Support specifying existing feature files, modify only some of the modalities.
         [ ] Implement resume function for run_dataset().
-        [ ] Forced Alignment & Aligned Feature Extraction.
-        [ ] GPU support in `run_dataset()`. Maybe discard Dataset and Dataloader is a good idea. Just implement multiprocessing pool manually.
         [ ] Clean up tmp folder before run_single.
         [ ] Better error logs, optimize stack traces to avoid duplicate messages.
         [ ] Set gpu_id during init, not in config.
@@ -174,6 +172,7 @@ def _text_extract_single(self, in_file : Path, in_text : str = None) -> np.ndarr
             text = self.text_extractor.load_text_from_file(in_file)
         text_result = self.text_extractor.extract(text)
         text_tokens = self.text_extractor.tokenize(text)
+        text_tokens = text_tokens.transpose(0, 2, 1)
         return text_result, text_tokens
 
     def _aligned_extract_single(
-Original file line number
+Diff line change
@@ Expand Up / @@ -8,3 +8,4 @@ dist @@
     __pycache__
     exts
     tmp.csv
+    build