paolorechia
commented
1 year ago Hi, which model are you using?
WizardLM 7b unquant yields the best results for me.
Sadly the most complex instruction I got it correctly in a consistent way was the simplified Chuck Norris Joke and also the Cat jokes
Anything much more complex results in failures.
I was playing with some new prompts flows, I wanted to get an AutoGPT like tooling for coding - but nothing too promising yet
Really like your initiative, I’d love to take a look there too - sadly, I’m also somewhat low on time this week.
TestUser @.***> schrieb am Mi. 3. Mai 2023 um 15:17:
So I appear to have the basic integration working with oobabooga, but i've been struggling a little bit with some of your agent examples. I took a look at the langchain autogpt example here to see what some of the differences were: https://python.langchain.com/en/latest/use_cases/autonomous_agents/autogpt.html
It does a few things differently than in your files, and it seems to not require the big prompt templates you are using. I have yet to successfully get an of the example agents in this repo to do more complex things than a basic single instruct. However, these examples are still tied to the openAI embeddings and openAI apis.
I was trying to figure out how to take this example and convert it off of openai embeddings to use the sentence transformers and chroma for the memory. These might be really really great examples to re-implement in your setup and show people how to use it with local models instead of all the openai stuff.
I'm kind of close to getting it to work? I've tried porting your Chroma example over to replace the FAISS version, but I've mostly failed.
— Reply to this email directly, view it on GitHub https://github.com/paolorechia/learn-langchain/issues/17, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABJDFZZHY55WDXEXXV2R3L3XEJLFVANCNFSM6AAAAAAXUMZENU . You are receiving this because you are subscribed to this thread.Message ID: @.***>
unoriginalscreenname
So I appear to have the basic integration working with oobabooga, but i've been struggling a little bit with some of your agent examples. I took a look at the langchain autogpt example here to see what some of the differences were: https://python.langchain.com/en/latest/use_cases/autonomous_agents/autogpt.html
It does a few things differently than in your files, and it seems to not require the big prompt templates you are using. I have yet to successfully get an of the example agents in this repo to do more complex things than a basic single instruct. However, these examples are still tied to the openAI embeddings and openAI apis.
I was trying to figure out how to take this example and convert it off of openai embeddings to use the sentence transformers and chroma for the memory. These might be really really great examples to re-implement in your setup and show people how to use it with local models instead of all the openai stuff.
I'm kind of close to getting it to work? I've tried porting your Chroma example over to replace the FAISS version, but I've mostly failed.