KoljaB
Docker Compose failing
Command used:
docker-compose up -d
rtstt container fails on startup with following error log:
2025-05-11 16:14:49
2025-05-11 16:14:49 ==========
2025-05-11 16:14:49 == CUDA ==
2025-05-11 16:14:49 ==========
2025-05-11 16:14:49
2025-05-11 16:14:49 CUDA Version 12.4.1
2025-05-11 16:14:49
2025-05-11 16:14:49 Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2025-05-11 16:14:49
2025-05-11 16:14:49 This container image and its contents are governed by the NVIDIA Deep Learning Container License.
2025-05-11 16:14:49 By pulling and using the container, you accept the terms and conditions of this license:
2025-05-11 16:14:49 https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license
2025-05-11 16:14:49
2025-05-11 16:14:49 A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.
2025-05-11 16:14:49
2025-05-11 16:14:51 Starting server, please wait...
2025-05-11 16:14:51 Traceback (most recent call last):
2025-05-11 16:14:51 File "/app/example_browserclient/server.py", line 3, in <module>
2025-05-11 16:14:51 from RealtimeSTT import AudioToTextRecorder
2025-05-11 16:14:51 File "/app/RealtimeSTT/__init__.py", line 1, in <module>
2025-05-11 16:14:51 from .audio_recorder import AudioToTextRecorder
2025-05-11 16:14:51 File "/app/RealtimeSTT/audio_recorder.py", line 38, in <module>
2025-05-11 16:14:51 import soundfile as sf
2025-05-11 16:14:51 ModuleNotFoundError: No module named 'soundfile'
Adding soundfile leads to the following error & shutdown
2025-05-11 16:24:01
2025-05-11 16:24:01 ==========
2025-05-11 16:24:01 == CUDA ==
2025-05-11 16:24:01 ==========
2025-05-11 16:24:01
2025-05-11 16:24:01 CUDA Version 12.4.1
2025-05-11 16:24:01
2025-05-11 16:24:01 Container image Copyright (c) 2016-2023, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
2025-05-11 16:24:01
2025-05-11 16:24:01 This container image and its contents are governed by the NVIDIA Deep Learning Container License.
2025-05-11 16:24:01 By pulling and using the container, you accept the terms and conditions of this license:
2025-05-11 16:24:01 https://developer.nvidia.com/ngc/nvidia-deep-learning-container-license
2025-05-11 16:24:01
2025-05-11 16:24:01 A copy of this license is made available in this container at /NGC-DL-CONTAINER-LICENSE for your convenience.
2025-05-11 16:24:01
2025-05-11 16:24:02 Starting server, please wait...
2025-05-11 16:24:02 Initializing RealtimeSTT...
2025-05-11 16:24:02 2025-05-11 20:24:02,569 - root - INFO - Initializing faster_whisper main transcription model large-v2
2025-05-11 16:24:03 Unable to load any of {libcudnn_ops.so.9.1.0, libcudnn_ops.so.9.1, libcudnn_ops.so.9, libcudnn_ops.so}
2025-05-11 16:24:03 Invalid handle. Cannot load symbol cudnnCreateTensorDescriptor
Last one can be solved with pip install "CTranslate2<4.5.0"
I'll update the repo soon, no clue why it works on some systems.
After fixing all of the above it still for some reason cannot open the websocket.
Firefox can’t establish a connection to the server at ws://localhost:8001/.
> curl -I localhost:8001
curl: (52) Empty reply from server
> curl -I ws://localhost:8001
curl: (52) Empty reply from server
No errors from either application
First, Thanks Kolja for all the amazing work done.
As he mentioned, CTransalte2 isn't in the requirements-gpu. But also missing this:
soundfile==0.13.1 CTranslate2<4.5.0 tqdm==4.66.2
After installing this deps the Dockerfile for GPU works (I commented CPU part as both are in same file)
After fixing all of the above it still for some reason cannot open the websocket.
Firefox can’t establish a connection to the server at ws://localhost:8001/.> curl -I localhost:8001 curl: (52) Empty reply from server > curl -I ws://localhost:8001 curl: (52) Empty reply from serverNo errors from either application
For anyone experiencing this issue: The server needs to be bound to 0.0.0.0 instead of localhost
async with websockets.serve(echo, "0.0.0.0", 8001):
