use cuda or cpu

Dockerfile

@@ -13,6 +13,9 @@ COPY requirements.txt ./
 COPY src/ ./src/
 
 RUN pip3 install -r requirements.txt
+RUN pip3 install torch==2.7.0 --index-url https://download.pytorch.org/whl/cu126
+RUN pip3 install torchaudio==2.7.0 --index-url https://download.pytorch.org/whl/cu126
+RUN pip3 install torchvision==0.22.0 --index-url https://download.pytorch.org/whl/cu126
 
 EXPOSE 8501
 

src/CallCenter.py

@@ -22,7 +22,11 @@ pipelineDiary = Pipeline.from_pretrained(
     "pyannote/speaker-diarization-3.1",
     use_auth_token=hugging_face)
 
-pipelineDiary.to(torch.device("cuda"))
+if torch.cuda.is_available():
+    print("diarize_wav_file Using CUDA")
+    pipelineDiary.to(torch.device("cuda"))
+else:
+    print("diarize_wav_file Using CPU")
 
 def diarize_wav_file(file_name):
     print("DIARIZING " + file_name)

src/transcribe_files.py

@@ -1,10 +1,18 @@
 import os
 
 import whisper
+import torch
 import time
 def transcribe_segments(speakers):
     print(f"Whisper models {whisper.available_models()}")
-    model = whisper.load_model("tiny.en", device="cuda")
+    if torch.cuda.is_available():
+        print("transcribe_segments Using CUDA")
+        device = "cuda"
+    else:
+        device = "cpu"
+        print("transcribe_segments Using CPU")
+
+    model = whisper.load_model("tiny.en", device=device)
     #model = whisper.load_model("medium.en", device="cuda")
 #    model = whisper.load_model("turbo", device="cuda")
     #model = whisper.load_model("large-v3-turbo", device="cuda")

src/transcript_analysis.py

@@ -121,8 +121,8 @@ def transcript_analysis(transcript):
         input += speaker + "\n"
 
     start = time.time()
-    response = use_huggingface2(input)
- #   response = use_openai(input)
+    # response = use_huggingface2(input)
+    response = use_openai(input)
     #response = use_bigbird_pegasus_large_arxiv(input)
     stop = time.time()
     elapsed=stop-start