joshuachou2018 / autoba Goto Github PK

Amazing work, tks for providing such useful tool!
Yet I don't quite understand how to deploy the tool locally. Especially when I download the codellam model. There are few problems:

1. Any difference between these model?
2. Is this tool compatible with latest codellam2 models ?
3. What is the "URL from email"? I'm not really proficient at LLM models, can we get some examples or can you provide some related info about how to get this URL?
4. Possibility to train or fine-tune to get a novel model in specific domain? For example, a metagenomics analysis machine for my own purpose by fine-tune the model?
   Wish to get some answers. Again, tks for develop such amazing tool !

Hello there,

Thanks for contributing such a great work to open community.

I tried case1.1 with openai gpt 3.5.

I only get response in the first step which is the "plan phase", the agent successfully created a plan.

But I failed in second step which is "code phase", the response message is empty.

Any response will be appreciated!

Best,

ZD

I can run autoBA as instructed in the repo README if using gpt. however when I downloaded llama weights (llama-7b-instruct) I encounter the issue where after running app.py with --model codellama-7bi I get the following:

`

start loading model
initializing model parallel with size 1
initializing ddp with size 1
initializing pipeline with size 1
Killed

`

I am using a fresh AWS-EC2 instance.