"save state" so that if we crash during the pipeline we might have some sort of output

Author: Noah-Jaffe

main.py

@@ -1,5 +1,4 @@
 import os
-import re
 from time import sleep, time
 import tkinter as tk
 from tkinter import BOTH, CENTER, E, LEFT, RIGHT, SOLID, TOP, W, X, Button, IntVar, Label, Spinbox, StringVar, Tk, Toplevel, filedialog, Frame, messagebox, font
@@ -19,6 +18,8 @@
 import json
 from huggingface_hub.hf_api import repo_exists as is_valid_model_id
 from PIL import Image, ImageTk
+import psutil
+from torch.cuda import is_available as is_cuda_available, mem_get_info as get_cuda_mem_info
 
 # import logging
 
@@ -246,6 +247,24 @@ def start_transcribe(self):
                 print(f"Converting {item.get_file()} to mp3 type so that it can be transcribed!")
                 item.filepath = convert_file_to_type(item.get_file(), ntype)
                 print(f"Convertion completed! Audio file can be found {item.get_file()}")
+            
+            # priority_levels = [
+            #     psutil.NORMAL_PRIORITY_CLASS, # normal,
+            #     psutil.ABOVE_NORMAL_PRIORITY_CLASS, # above normal
+            #     psutil.ABOVE_NORMAL_PRIORITY_CLASS, # above normal
+            #     psutil.HIGH_PRIORITY_CLASS, # high priority
+            # ]
+            # priority_points = 0
+            # curr_state = psutil.virtual_memory()
+            # if (curr_state.total/(2**30) > 16):
+            #     # 16gb+ ram
+            #     priority_points += 1
+            # if (is_cuda_available()):
+            #     # has cuda
+            #     priority_points += 1
+            #     if (get_cuda_mem_info()[1]/(2**30) > 10):
+            #         # has big cuda
+            #         priority_points += 1
             proc = subprocess.Popen(
                 args=[
                     sys.executable,
@@ -262,6 +281,7 @@ def start_transcribe(self):
                     cwd=os.getcwd(),
                     start_new_session=True
                 )
+            # psutil.Process(proc.pid).nice(priority_levels[priority_points])
             self.root.title("Transcriber - PLEASE DONT KILL ME - I AM WORKING! I PROMISE!")
             while proc.poll() == None:
                 try:
@@ -274,6 +294,7 @@ def start_transcribe(self):
         except:
             pass
         self.root.title("Transcriber")
+        # spawn_popup_activity("Transcriber", "Completed transcribing the files!")
 
     def show_error(self, *args):
         """Display the error to the user as a popup window"""

transcribe_proc.py

@@ -7,6 +7,7 @@
 from huggingface_hub.hf_api import repo_exists as is_valid_model_id
 import pycountry
 from CustomAiEngine import CustomAiEngine
+import traceback
 
 
 def transcribe_file(input_file, model_name=None, num_speakers=2, lang="eng"):
@@ -45,27 +46,50 @@ def transcribe_file(input_file, model_name=None, num_speakers=2, lang="eng"):
         retrace,  # uncertain how this benifits us
         # morphosyntax,
         utr,
-        fa
+        # fa
     ] if action]
 
-    # create a pipeline
-    nlp = ba.BatchalignPipeline(*pipeline_activity)
-    doc = ba.Document.new(media_path=input_file, lang=lang)
-    doc = nlp(doc)
-    chat = ba.CHATFile(doc=doc)
     n = 0
     output_file = f"{input_file}{'_'+str(n) if n > 0 else ''}.cha"
     while 1:
         output_file = f"{input_file}{'_'+str(n) if n > 0 else ''}.cha"
         if not os.path.exists(output_file):
             break
         n += 1
-    chat.write(output_file, write_wor=False)
+    doc = ba.Document.new(media_path=input_file, lang=lang)
+    for idx, activity in enumerate(pipeline_activity, start=1):
+        nlp = ba.BatchalignPipeline(activity)
+        try:
+            doc = nlp(doc)
+            chat = ba.CHATFile(doc=doc)
+            chat.write(output_file, write_wor=False)
+            with open(output_file,'a',encoding='utf-8') as f:
+                f.write(f"@DEBUG Completed step {idx}/{len(pipeline_activity)} - {(type(activity).__name__).replace('Engine','')}\n")
+        except Exception as e:
+            print(traceback.format_exc(e))
+            print(f"{output_file} made it to step: {idx-1}/{len(pipeline_activity)}")
+            with open(output_file,'a',encoding='utf-8') as f:
+                f.write(f"@DEBUG error during step {idx}/{len(pipeline_activity)} - {(type(activity).__name__).replace('Engine','')}\n")
+        
+        
+    # create a pipeline
+    # nlp = ba.BatchalignPipeline(*pipeline_activity)
+    # doc = ba.Document.new(media_path=input_file, lang=lang)
+    # doc = nlp(doc)
+    # chat = ba.CHATFile(doc=doc)
+    # n = 0
+    # output_file = f"{input_file}{'_'+str(n) if n > 0 else ''}.cha"
+    # while 1:
+    #     output_file = f"{input_file}{'_'+str(n) if n > 0 else ''}.cha"
+    #     if not os.path.exists(output_file):
+    #         break
+    #     n += 1
+    # chat.write(output_file, write_wor=False)
     print(f"Wrote to {output_file}", flush=True)
-    try:
-        os.startfile(output_file)
-    except:
-        pass
+    # try:
+    #     os.startfile(output_file)
+    # except:
+    #     pass
     # this is process blocking so we dont do it for now
     # return spawn_popup_activity(title="COMPLETED!",message=f"Completed transcription of\n{input_file}\nOutput file can be found here:\n{output_file}\nOpen file now?", yes=lambda: os.startfile(output_file))
 
@@ -77,6 +101,7 @@ def spawn_popup_activity(title, message, yes=None, no=None):
         return no()
 
 if __name__ == "__main__":
+    sys.argv.append('{"input_file":"D:\\\\Parent-Child Interactions\\\\Parent-Child Interactions\\\\P002_PCI.mp3","num_speakers":3,"lang":"","model_name":"openai/whisper-base.en"}')
     print("Attempting to transcribe for:", sys.argv[1:], flush=True)
     print(sys.argv, flush=True)
     for data in sys.argv[1:]: