Skip to content

C
coderai
Commit d23c2148 authored by Your Name's avatar Your Name
Browse files
Options

Force CPU mode for faster-whisper (CUDA not compatible with Vulkan)

parent 1dafc558
Hide whitespace changes
Inline Side-by-side
Showing with 6 additions and 5 deletions
coderai
View file @ d23c2148
......@@ -2464,7 +2464,7 @@ async def create_transcription(
whisper_model = WhisperModel(
model_to_use,
device="cuda" if torch.cuda.is_available() else "cpu",
device="cpu", # faster-whisper CUDA doesn't work with AMD/Vulkan
compute_type=compute_type
)
# Store in multi_model_manager
......@@ -3796,13 +3796,14 @@ def main():
model_path = download_model(model_to_use, cache_dir)
model_to_use = model_path
# Determine compute type
compute_type = "float16" if torch.cuda.is_available() else "int8"
# Determine compute type - always use CPU on Vulkan backend
# faster-whisper CUDA doesn't work with AMD/Vulkan GPUs
compute_type = "int8"
# Load the model
# Load the model - always use CPU (faster-whisper CUDA doesn't work with AMD/Vulkan)
whisper_model = WhisperModel(
model_to_use,
device="cuda" if torch.cuda.is_available() else "cpu",
device="cpu",
compute_type=compute_type
)
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment