added fp16 support for whisper inference on cuda

README.md

@@ -14,6 +14,7 @@
 
 </p>
 
+
 ### Run your IDE as administrator
 
 you will get following error if administrator permission is not there:
@@ -88,7 +89,7 @@ transcript will also indicate the timeframe in seconds where each speaker speaks
 ```
 from speechlib import Transcriptor
 
-file = "obama1.wav"  # your audio file
+file = "obama_zach.wav"  # your audio file
 voices_folder = "voices" # voices folder containing voice samples for recognition
 language = "en"          # language code
 log_folder = "logs"      # log folder for storing transcripts

examples/transcribe.py

@@ -1,6 +1,6 @@
 from speechlib import Transcriptor
 
-file = "obama1.wav"  # your audio file
+file = "obama_zach.wav"  # your audio file
 voices_folder = "voices" # voices folder containing voice samples for recognition
 language = "en"          # language code
 log_folder = "logs"      # log folder for storing transcripts

library.md

@@ -72,7 +72,7 @@ transcript will also indicate the timeframe in seconds where each speaker speaks
 ```
 from speechlib import Transcriptor
 
-file = "obama1.wav"  # your audio file
+file = "obama_zach.wav"  # your audio file
 voices_folder = "voices" # voices folder containing voice samples for recognition
 language = "en"          # language code
 log_folder = "logs"      # log folder for storing transcripts
@@ -99,13 +99,20 @@ end: ending time of speech in seconds
 text: transcribed text for speech during start and end  
 speaker: speaker of the text 
 
-#### voices_folder structure:  
-
-![voices_folder_structure](voices_folder_structure1.png)
-
-#### Transcription:  
-
-![transcription](transcript.png)
+#### voices folder structure:
+```
+voices_folder    
+|---> person1      
+|        |---> sample1.wav   
+|        |---> sample2.wav     
+|                ...
+|
+|---> person2  
+|        |---> sample1.wav  
+|        |---> sample2.wav   
+|                ...
+|--> ...  
+```
 
 supported language codes:  
 

requirements.txt

@@ -1,8 +1,9 @@
-transformers
-torch
-torchaudio
-pydub
-pyannote.audio
-speechbrain
-accelerate
-faster-whisper
+transformers==4.36.2
+torch==2.1.2
+torchaudio==2.1.2
+pydub==0.25.1
+pyannote.audio==3.1.1
+speechbrain==0.5.16
+accelerate==0.26.1
+faster-whisper==0.10.1
+openai-whisper==20231117

setup.py

@@ -5,7 +5,7 @@
 
 setup(
     name="speechlib",
-    version="1.1.1",  
+    version="1.1.2",  
     description="speechlib is a library that can do speaker diarization, transcription and speaker recognition on an audio file to create transcripts with actual speaker names. This library also contain audio preprocessor functions.",
     packages=find_packages(),
     long_description=long_description,

setup_instruction.md

@@ -9,7 +9,7 @@ for publishing:
     pip install twine
 
 for install locally for testing:
-    pip install dist/speechlib-1.1.0-py3-none-any.whl
+    pip install dist/speechlib-1.1.2-py3-none-any.whl
 
 finally run:
     twine upload dist/*

speechlib/transcribe.py

@@ -32,9 +32,14 @@ def transcribe(file, language, model_size, whisper_type, quantization):
                 Exception("Language code not supported.\nThese are the supported languages:\n", model.supported_languages)
         else:
             try:
-                model = whisper.load_model(model_size)
-                result = model.transcribe(file, language=language)
-                res = result["text"]
+                if torch.cuda.is_available():
+                    model = whisper.load_model(model_size, device="cuda")
+                    result = model.transcribe(file, language=language, fp16=True)
+                    res = result["text"]
+                else:
+                    model = whisper.load_model(model_size, device="cpu")
+                    result = model.transcribe(file, language=language, fp16=False)
+                    res = result["text"]
 
                 return res
             except Exception as err: