โ02-15-2024 08:13 AM
Hello,
I've been trying to serve registered MLflow models at GPU Model Serving Endpoint, which works except for the models using bitsandbytes library. The library is used to quantise the LLM models into 4-bit/ 8-bit (e.g. Mistral-7B), however, it runs into error while registering at endpoint. This error is shown in the service log:
All libraries needed are registered in the requirements.txt files, it looks like one option to fix the error is to run a bash script to help it locate the right path of the package, but we're not able to do so at serving endpoint.
Has anyone successfully served a quantised LLM model at Databricks model serving using bitsandbytes? If so, how do you get around it? Any help on the topic would be much appreciated.
Thanks
โ02-16-2024 01:29 AM
Hi @phi_alpaca, Serving quantized LLM (Large Language Models) using Databricks Model Serving can be a powerful way to optimize performance and reduce latency.
Letโs explore some options:
Databricks Model Serving with GPU and LLM Optimization:
Quantized LLMs with bitsandbytes:
Troubleshooting the Error:
requirements.txt
file.
โ02-20-2024 01:02 AM
Hi @phi_alpaca , we are facing exactly the same issue trying to serve a bitsandbytes quantized version of Mixtral-8x7B . Did you have any progress resolving this? The answer from @Kaniz_Fatma isn't too helpful and seems to be AI-generated...
As you say, the deployed container is such a black box that we can't take the diagnostic steps listed in the error output.
โ02-20-2024 06:24 AM
Hey @G-M , thanks for sharing your experience as well. Unfortunately I haven't had any luck on my end for resolving this. Would be interested to know if you have any breakthrough down the line. Is it something Databricks might be able to put a small fix in please? @Kaniz_Fatma
Thanks
โ02-22-2024 04:26 AM
Hi, @phi_alpaca have you managed to solve this? We have a similar issue..
โ02-22-2024 05:23 AM
Hey @JAgreenskylake , no luck so far. I have been working around it by not using quantised models, which is not ideal, so really hope it's possible to do that soon.
โ02-26-2024 10:13 AM
We have solved it by providing a conda_env.yaml when we log the model, all we needed was to add cudatoolkit=11.8 to the dependencies.
โ02-27-2024 08:36 AM
Thanks so much for sharing and glad it worked out for you guys!
I will have a go and feed back.
โ03-07-2024 12:51 AM
I seem to have some compatibility issues with cudatoolkit=11.8, would it be possible for you share what versions you use for torch, transformers, accelerate, and bitsandbytes? Thanks!
โ03-07-2024 12:56 AM
These versions are working for us:
torch==1.13.1
transformers==4.35.2
accelerate==0.25.0
bitsandbytes==0.41.3
Join a Regional User Group to connect with local Databricks users. Events will be happening in your city, and you wonโt want to miss the chance to attend and share knowledge.
If there isnโt a group near you, start one and help create a community that brings people together.
Request a New Group