[BUG] Can't transcribe anything, including simple YouTube videos when using the GPU

[moda20]

Description

I can't transcribe any video from youtube or file that i input, the error message on the UI is "Failed: Could not transcribe"

To Reproduce

Steps to reproduce the behavior:

1. open the new transcribe dialog
2. enter youtube video ( example : https://www.youtube.com/watch?v=cy9b2v8Rhvs)
3. choose lang as "en"
4. chose devie as GPU
5. wait for some time during the "waiting for transcription"
6. see the error appear

Expected behavior

transcribing the video

Environment

• OS: ubuntu 24.04
• Browser:
• Version:
• Hosting: local using docker

Logs and Configuration

Docker Compose Logs

Run the following command in the project folder, force the error, and paste the logs below: docker compose logs -f --tail 50

whisper logs

2024-06-12T20:12:53.343697119Z 2024-06-12 20:12:53,343 CRIT Supervisor is running as root.  Privileges were not dropped because no user is specified in the config file.  If you intend to run as root, you can set user=root in the config file to avoid this message.
2024-06-12T20:12:53.344460371Z 2024-06-12 20:12:53,344 INFO supervisord started with pid 1
2024-06-12T20:12:54.347376631Z 2024-06-12 20:12:54,347 INFO spawned: 'backend' with pid 7
2024-06-12T20:12:54.348359851Z 2024-06-12 20:12:54,348 INFO spawned: 'frontend' with pid 8
2024-06-12T20:12:54.349436037Z 2024-06-12 20:12:54,349 INFO spawned: 'nginx' with pid 9
2024-06-12T20:12:54.350488018Z 2024-06-12 20:12:54,350 INFO spawned: 'transcription' with pid 10
2024-06-12T20:12:55.399517055Z 2024-06-12 20:12:55,399 INFO success: backend entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2024-06-12T20:12:55.399533316Z 2024-06-12 20:12:55,399 INFO success: frontend entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2024-06-12T20:12:55.399555508Z 2024-06-12 20:12:55,399 INFO success: nginx entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)
2024-06-12T20:12:55.399573833Z 2024-06-12 20:12:55,399 INFO success: transcription entered RUNNING state, process has stayed up for > than 1 seconds (startsecs)

mongodb logs

2024-06-12T19:58:24.898217789Z {"t":{"$date":"2024-06-12T19:58:24.898+00:00"},"s":"I",  "c":"STORAGE",  "id":4795901, "ctx":"SignalHandler","msg":"WiredTiger closed","attr":{"durationMillis":17}}
2024-06-12T19:58:24.898225233Z {"t":{"$date":"2024-06-12T19:58:24.898+00:00"},"s":"I",  "c":"STORAGE",  "id":22279,   "ctx":"SignalHandler","msg":"shutdown: removing fs lock..."}
2024-06-12T19:58:24.898257535Z {"t":{"$date":"2024-06-12T19:58:24.898+00:00"},"s":"I",  "c":"-",        "id":4784931, "ctx":"SignalHandler","msg":"Dropping the scope cache for shutdown"}
2024-06-12T19:58:24.898262364Z {"t":{"$date":"2024-06-12T19:58:24.898+00:00"},"s":"I",  "c":"FTDC",     "id":20626,   "ctx":"SignalHandler","msg":"Shutting down full-time diagnostic data capture"}
2024-06-12T19:58:24.899898606Z {"t":{"$date":"2024-06-12T19:58:24.899+00:00"},"s":"I",  "c":"CONTROL",  "id":20565,   "ctx":"SignalHandler","msg":"Now exiting"}
2024-06-12T19:58:24.899984089Z {"t":{"$date":"2024-06-12T19:58:24.899+00:00"},"s":"I",  "c":"CONTROL",  "id":8423404, "ctx":"SignalHandler","msg":"mongod shutdown complete","attr":{"Summary of time elapsed":{"Statistics":{"Enter terminal shutdown":"0 ms","Step down the replication coordinator for shutdown":"1 ms","Time spent in quiesce mode":"0 ms","Shut down FLE Crud subsystem":"0 ms","Shut down MirrorMaestro":"0 ms","Shut down WaitForMajorityService":"0 ms","Shut down the logical session cache":"1 ms","Shut down the transport layer":"0 ms","Shut down the global connection pool":"0 ms","Shut down the flow control ticket holder":"0 ms","Kill all operations for shutdown":"0 ms","Shut down all tenant migration access blockers on global shutdown":"0 ms","Shut down all open transactions":"0 ms","Acquire the RSTL for shutdown":"0 ms","Shut down the IndexBuildsCoordinator and wait for index builds to finish":"0 ms","Shut down the replica set monitor":"0 ms","Shut down the migration util executor":"0 ms","Shut down the health log":"0 ms","Shut down the TTL monitor":"0 ms","Shut down expired pre-images and documents removers":"0 ms","Shut down the storage engine":"18 ms","Wait for the oplog cap maintainer thread to stop":"0 ms","Shut down full-time data capture":"0 ms","shutdownTask total elapsed time":"21 ms"}}}}
2024-06-12T19:58:24.900005018Z {"t":{"$date":"2024-06-12T19:58:24.899+00:00"},"s":"I",  "c":"CONTROL",  "id":23138,   "ctx":"SignalHandler","msg":"Shutting down","attr":{"exitCode":0}}
2024-06-12T19:58:25.882379893Z
2024-06-12T19:58:25.882397878Z MongoDB init process complete; ready for start up.
2024-06-12T19:58:25.882400823Z

Docker Compose File (the default)

version: "3.9"

services:
  mongo:
    image: mongo
    env_file:
      - .env
    restart: unless-stopped
    volumes:
      - ./whishper_data/db_data:/data/db
      - ./whishper_data/db_data/logs/:/var/log/mongodb/
    environment:
      MONGO_INITDB_ROOT_USERNAME: ${DB_USER:-whishper}
      MONGO_INITDB_ROOT_PASSWORD: ${DB_PASS:-whishper}
    expose:
      - 27017
    command: ['--logpath', '/var/log/mongodb/mongod.log']

  translate:
    container_name: whisper-libretranslate
    image: libretranslate/libretranslate:latest-cuda
    restart: unless-stopped
    volumes:
      - ./whishper_data/libretranslate/data:/home/libretranslate/.local/share
      - ./whishper_data/libretranslate/cache:/home/libretranslate/.local/cache
    env_file:
      - .env
    user: root
    tty: true
    environment:
      LT_DISABLE_WEB_UI: True
      LT_LOAD_ONLY: ${LT_LOAD_ONLY:-en,fr,es}
      LT_UPDATE_MODELS: True
    expose:
      - 5000
    networks:
      default:
        aliases:
          - translate
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            count: all
            capabilities: [gpu]

  whishper:
    pull_policy: always
    image: pluja/whishper:${WHISHPER_VERSION:-latest-gpu}
    env_file:
      - .env
    volumes:
      - ./whishper_data/uploads:/app/uploads
      - ./whishper_data/logs:/var/log/whishper
    container_name: whishper
    restart: unless-stopped
    networks:
      default:
        aliases:
          - whishper
    ports:
      - 8082:80
    depends_on:
      - mongo
      - translate
    environment:
      PUBLIC_INTERNAL_API_HOST: "http://127.0.0.1:80"
      PUBLIC_TRANSLATION_API_HOST: ""
      PUBLIC_API_HOST: ${WHISHPER_HOST:-}
      PUBLIC_WHISHPER_PROFILE: gpu
      WHISPER_MODELS_DIR: /app/models
      UPLOAD_DIR: /app/uploads
    deploy:
      resources:
        reservations:
          devices:
          - driver: nvidia
            count: all
            capabilities: [gpu]

[thewriteway]

did you try this? #100 (comment)

[moda20]

@thewriteway yes, that was an error i got at first but i changed the IP and now i can STT fine when using the CPU, the issue is when using the GPU

[thewriteway]

are you passing through the cuda environment path correctly? @moda20

for example you should see something like this when you run this command:

echo $PATH | tr ':' '\n' | grep CUDA or
echo $PATH | tr ':' '\n' | grep cuda

/mnt/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.5/bin
/mnt/c/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v12.5/libnvvp

in my example its a windows WSL but you should see the equivalent for linux too,

Note it needs CUDA 12.4 or above on the host machine.

[moda20]

@thewriteway i am using CUDA 12.2, seems like the driver that supports higher versions of CUDA is not available on ubuntu24.04. i setup this machine a some small-time ago and the latest available version was driver 535 and CUDA 12.2 versions