batchalign 0.7.6a14__tar.gz → 0.7.6a16__tar.gz

{batchalign-0.7.6a14/batchalign.egg-info → batchalign-0.7.6a16}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: batchalign
-Version: 0.7.6a14
+Version: 0.7.6a16
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu

{batchalign-0.7.6a14 → batchalign-0.7.6a16}/batchalign/cli/dispatch.py

@@ -65,7 +65,7 @@ def _dispatch(command, lang, num_speakers,
     for basedir, _, fs in os.walk(in_dir):
         for f in fs:
             path = Path(os.path.join(basedir, f))
-            ext = path.suffix.strip(".").strip()
+            ext = path.suffix.strip(".").strip().lower()
 
             # calculate input path, convert if needed
             inp_path = str(path)

{batchalign-0.7.6a14 → batchalign-0.7.6a16}/batchalign/models/whisper/infer_asr.py

@@ -67,85 +67,16 @@ class WhisperASRModel(object):
         self.__config = GenerationConfig.from_pretrained(base)
         self.__config.no_repeat_ngram_size = 4
 
-        if language == "Cantonese":
-            self.pipe = pipeline(
-                "automatic-speech-recognition",
-                model=model,
-                # tokenizer=WhisperTokenizer.from_pretrained(base),
-                chunk_length_s=30,
-                # stride_length_s=3,
-                device=DEVICE,
-                # torch_dtype=torch.float32,
-                return_timestamps="word",
-            )
-            self.__config = GenerationConfig.from_model_config(self.pipe.model.config)
-            self.__config.no_repeat_ngram_size = 4
-            self.__config.use_cache = False
-
-            forced_decoder_ids = self.pipe.tokenizer.get_decoder_prompt_ids(language="yue", task="transcribe")
-            
-            suppress_tokens = []
-
-            # Define other parameters
-            return_attention_mask = False
-            pad_token_id = 50257
-            bos_token_id = 50257
-            eos_token_id = 50257
-            decoder_start_token_id = 50258
-            begin_suppress_tokens = [
-                220,
-                50257
-            ],
-            alignment_heads = [
-                [5, 3],
-                [5, 9],
-                [8, 0],
-                [8, 4],
-                [8, 8],
-                [9, 0],
-                [9, 7],
-                [9, 9],
-                [10, 5]
-            ]
-            lang_to_id = {"<|yue|>": 50325}
-            task_to_id = {"transcribe": 50359}
-            is_multilingual = True
-            max_initial_timestamp_index = 50
-            no_timestamps_token_id = 50363
-            prev_sot_token_id = 50361
-            max_length = 448
-
-            # Assign values to generation config
-            self.__config.forced_decoder_ids = forced_decoder_ids
-            self.__config.suppress_tokens = suppress_tokens
-            self.__config.pad_token_id = pad_token_id
-            self.__config.bos_token_id = bos_token_id
-            self.__config.eos_token_id = eos_token_id
-            self.__config.decoder_start_token_id = decoder_start_token_id
-            self.__config.lang_to_id = lang_to_id
-            self.__config.task_to_id = task_to_id
-            self.__config.alignment_heads = alignment_heads
-            self.__config.alignment_heads = alignment_heads
-            self.__config.begin_suppress_tokens = begin_suppress_tokens
-            self.__config.is_multilingual = is_multilingual
-            self.__config.max_initial_timestamp_index = max_initial_timestamp_index
-            self.__config.no_timestamps_token_id = no_timestamps_token_id
-            self.__config.prev_sot_token_id = prev_sot_token_id
-            self.__config.max_length =max_length
-
-            self.pipe.model.generation_config = self.__config
-
-        else:
-            self.pipe = pipeline(
-                "automatic-speech-recognition",
-                model=model,
-                tokenizer=WhisperTokenizer.from_pretrained(base),
-                chunk_length_s=25,
-                stride_length_s=3,
-                device=DEVICE,
-                torch_dtype=torch.float32,
-                return_timestamps="word",
-            )
+        self.pipe = pipeline(
+            "automatic-speech-recognition",
+            model=model,
+            tokenizer=WhisperTokenizer.from_pretrained(base),
+            chunk_length_s=25,
+            stride_length_s=3,
+            device=DEVICE,
+            torch_dtype=torch.float32,
+            return_timestamps="word",
+        )
         L.debug("Done, initalizing processor and config...")
         processor = WhisperProcessor.from_pretrained(base)
         L.debug("Whisper initialization done.")

{batchalign-0.7.6a14 → batchalign-0.7.6a16}/batchalign/pipelines/morphosyntax/ud.py

@@ -213,11 +213,15 @@ def handler__NOUN(word, lang=None):
     if word.deprel == "obj" and case.strip() == "":
         case = "Acc"
 
+    ger = ""
+    if word.text.endswith("ing") and lang == "en":
+        ger += "-Ger"
+
     # clear defaults
     if gender_str == "-Com,Neut" or gender_str == "-Com" or gender_str == "-ComNeut": gender_str=""
     if number_str == "-Sing": number_str=""
 
-    return handler(word, lang)+gender_str+number_str+stringify_feats(case, type)
+    return handler(word, lang)+gender_str+number_str+stringify_feats(case, type)+ger
 
 def handler__PROPN(word, lang=None):
     # code as noun

batchalign-0.7.6a16/batchalign/version

@@ -0,0 +1,3 @@
+0.7.6-alpha.16
+October 16, 2024
+fixing asr for file names

{batchalign-0.7.6a14 → batchalign-0.7.6a16/batchalign.egg-info}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: batchalign
-Version: 0.7.6a14
+Version: 0.7.6a16
 Summary: Python Speech Language Sample Analysis
 Author: Brian MacWhinney, Houjun Liu
 Author-email: macw@cmu.edu, houjun@cmu.edu

batchalign-0.7.6a14/batchalign/version

@@ -1,3 +0,0 @@
-0.7.6-alpha.14
-October 14, 2024
-creaky