HeavyIQ not enabled anymore?

Comments

7 comments

  • Avatar
    Candido Dessanti

    Hi Gianfranco,

    Are you talking about the 8.0.2 release? The release should be installed on RHEL and derivatives only that have Python 3.11 installed. Could you please rollback to 8.0.1?

    Candido

    0
    Comment actions Permalink
  • Avatar
    Gianfranco Campana

    Yes, I was on the 8.0.2 version.

    And thank you, rolling back to 8.0.1 [Ubuntu 22.04 Docker free GPU version] solved the problem.

    I didn't find this info into the docs.

     

     

    0
    Comment actions Permalink
  • Avatar
    Gianfranco Campana

    Hi Candido,

    today I simply restarted the container in order to free the gpu memory allocated, and soon after the restarting I got: until then HeavyIQ was enabled.

    I don't see errors in /var/lib/heavyai/storage/log/heavyiq_build.log

    After several attempts, I delete all containers and docker images and pulled again the 8.1.0 version, still there is no way to gat HeavyIQ enabled for the 8.1.0 version, that was working fine until a few hours ago with HeavyIQ

     

    0
    Comment actions Permalink
  • Avatar
    Gianfranco Campana


    Just adding what I found in the /logs/iq/console log ( this one in case  I set custom_lmm_type in heavy.conf):

      File "<frozen heavyiq.api>", line 155, in create_app
      File "<frozen heavyiq.config>", line 78, in get_config
      File "<frozen heavyiq.config>", line 60, in validate_config
    ValueError: Custom LLM type is set to 'API_VLLM', but API base URL is not set.
     

    In addition in /logs/all I get:

    time="2024-08-20T08:40:11Z" level=error msg="Could not GET HeavyIQ version"
    time="2024-08-20T08:40:11Z" level=error msg="Could not get HeavyIQ version"
    0
    Comment actions Permalink
  • Avatar
    Todd Mostak

    Gianfranco Campana you have these three items set in your conf file in the [iq] section?

     

    custom_llm_type = "API_VLLM" # API or AZURE

    custom_llm_api_base = "https://api.heavy.ai/v1"

    rag_embed_server_base = "https://embeddings.heavy.ai"
    0
    Comment actions Permalink
  • Avatar
    Gianfranco Campana
     

    Hi Todd,
    custom_llm_api_base was indeed missing, and after adding that to heavy.conf the problem is gone, and HeravyIQ is enabled again.

    In addition, also the autoselect table source problem is solved, I see in the logs that embeddings are correctly generated.

    Thank you Todd.

     
    0
    Comment actions Permalink
  • Avatar
    Todd Mostak

    Awesome, that's great to hear Gianfranco!

    0
    Comment actions Permalink

Please sign in to leave a comment.