Closed DIGist closed 1 year ago
Problem Description
I'm trying to get AgentLLM to work with Vicuna and WizardLM through Oobabooga and it's very tough to debug without working examples of the exchanges and it would help enable development of other model support.
Proposed Solution
creation of a debug log button that logs/exports the prompts sent to the model and the models response. a default examples of the model using the various commands would help.
Alternatives Considered
No response
Additional Context
No response
Acknowledgements
- [x] My issue title is concise, descriptive, and in title casing.
- [x] I have searched the existing issues to make sure this feature has not been requested yet.
- [x] I have provided enough information for the maintainers to understand and evaluate this request.
It is set up to work well with OpenAI, but will require some prompt engineering for the smaller models to work well. I had the huggingchat OpenAssistant almost working okay, but 2000 max tokens really hurts for this kind of thing.
As far as a debug log, you can add a print to the AgentLLM.py
file in the run
function.
Right before the instruct
happens, put this there:
print(f"PROMPT: {prompt}")
Problem Description
I'm trying to get AgentLLM to work with Vicuna and WizardLM through Oobabooga and it's very tough to debug without working examples of the exchanges and it would help enable development of other model support.
Proposed Solution
creation of a debug log button that logs/exports the prompts sent to the model and the models response. a default examples of the model using the various commands would help.
Alternatives Considered
No response
Additional Context
No response
Acknowledgements