Hi Tyler, great videos you put out every time. Thank you. A question, is it possible to have a different RAG for each agent in Autogen separately? Thank you again and keep it up!
Hey, thanks for watching, appreciate it! So, a different rag for each agent. I think this would be possible because I would imagine you can just instantiate multiple Rag Agents, they are call RetrieveProxyAgent I believe.
I could not get any model to run under the Multi-Model mode of LMStudio, but I am running a Linux Beta version on an older laptop, so I am not surprised I am having issues. It looks like they have taken a leap forward. In the future I will keep trying and maybe try it on a larger cloud server.
so I have 8GB Ram, and I run the phi-2 model, but that's one of the lower end models as far as needing hardware and that's okay. But as soon as I get to a 7B parameter model like Mistral, it's really slow. This is ollama's recommendations for RAM: Note: You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models. But, I would add 8GB of RAM to each of those in my opinion for it to be smooth and quick.
1:20 LM studio says playground supports only full GPU offload, but in your case the top indicator shows that you use RAM (that has maximum size of 8 GB?) instead of VRAM. Is this error on their side?
Hmm, good question! I will have to look in their discord to see if this has come up with other people. I didn't really notice this tbh, I'll try to get back to you
I saw the update today and didn’t realize what the start server button was doing on this tab. Thank you for explaining! :)
Awesome, glad it was helpful! Yeah I was a little confused as well
the AI red car in the background looks pretty real. A new feature called TylerVision by TylerGen hahaha.
Hey I do my best around here hahaha :D I like the sound of TylerVision...can have some interesting ideas
@@TylerReedAI joking aside, it sounds cool. "TylerVision by TylerLabs" 🦾
Hi Tyler, great videos you put out every time. Thank you. A question, is it possible to have a different RAG for each agent in Autogen separately? Thank you again and keep it up!
Hey, thanks for watching, appreciate it! So, a different rag for each agent. I think this would be possible because I would imagine you can just instantiate multiple Rag Agents, they are call RetrieveProxyAgent I believe.
Glad they update it
Yup agreed, good update
A.I. 🔥 so happy to be alive to see it🥹
Need to try this on my CrewAI frameworks.
Let me know how this goes when you try it out!
@@TylerReedAI Absolutely! Likewise if you have time friend 🙂
I could not get any model to run under the Multi-Model mode of LMStudio, but I am running a Linux Beta version on an older laptop, so I am not surprised I am having issues. It looks like they have taken a leap forward. In the future I will keep trying and maybe try it on a larger cloud server.
Let me know! Yeah that is the beta version, and I just checked in their discord and there are a few issues with the beta version right now
What would you recommend I run this on? I presume that some of the heftier models require about 64GB of ram yes?
so I have 8GB Ram, and I run the phi-2 model, but that's one of the lower end models as far as needing hardware and that's okay. But as soon as I get to a 7B parameter model like Mistral, it's really slow. This is ollama's recommendations for RAM:
Note: You should have at least 8 GB of RAM available to run the 7B models, 16 GB to run the 13B models, and 32 GB to run the 33B models.
But, I would add 8GB of RAM to each of those in my opinion for it to be smooth and quick.
1:20 LM studio says playground supports only full GPU offload, but in your case the top indicator shows that you use RAM (that has maximum size of 8 GB?) instead of VRAM. Is this error on their side?
Hmm, good question! I will have to look in their discord to see if this has come up with other people. I didn't really notice this tbh, I'll try to get back to you
@@TylerReedAI ok, I'll be waiting for the responce. Thanks in advance!
is there RAG now? else the update is nothing more than bloatware.
Not with this update, but it's more of allowing multiple models to run locally off of 1 server. Which is pretty helpful