#ollama

1 messages · Page 1 of 1 (latest)

gentle pasture
#

im running openclaw local with ollama local model qwen2.5-coder:3b when doing the ollama run qwen2.5-coder:3b i can ask questions and it responds inmeaditly but when i add it to openclaw and ask something it keeps giving this error: 500 {"error":"model requires more system memory (19.9 GiB) than is available (13.9 GiB)"} and in the terminal it gives this: (check photo)

turbid sigilBOT
#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

gentle pasture
#

nvm cant add photo

#

15:34:00+02:00 [agent/embedded] Profile ollama:default timed out. Trying next account...
15:34:00+02:00 [agent/embedded] embedded run failover decision: runId=fc38064f-3d50-4b0c-9d35-755292f9bf63 stage=assistant decision=surface_error reason=timeout provider=ollama/llama3.1:8b profile=sha256:9c018ec112cf

#

and the weird thing is i dont even have llama3.1:8b installed...

#

@left geode

severe sonnet
#

on your machine running ollama type ollama ps

gentle pasture
severe sonnet
#

how much vram do you have? 32gb?

gentle pasture
#

i just got a computer not a good pc or something

severe sonnet
#

that isn't very helpful lol... go to your task manager (ctrl+alt+del, or right click on your task bar) then performance tab, then pick your gpu

gentle pasture
#

😭

severe sonnet
#

you need a model that is smaller than 4gb then, it's bleeding into your system memory which will be a bad experience

gentle pasture
severe sonnet
#

that's aside from the not having enough, but even if we figure that out, using system memory is going to brick your experience anyway

gentle pasture
#

but not openclaw

severe sonnet
#

no idea tbh, my assumption is you have something running that's taking up too much of your system memory

drifting wind
#

Make an ollama account. Pull a cloud model, reload openclaw using the cloud model.

#

No longer dependent on your system for inference.

#

Also 🖕 my earlier points solved your problem

gentle pasture
#

thats why i went to local

severe sonnet
#

you can find a model that'll fit in your 4gb, it's just not going to be a great experience

drifting wind
#

Nemotron-3-super:cloud has no limits right now

#

Completely free while its in preview

#

Openrouter is a good option for free models and decent rates..

#

But for local runs on low specs you have to choose between slow or stupid... I do stupid with offload to cloud models

gentle pasture
#

i did openrouter as main and gemini as fallback but still getting rate limited fast

drifting wind
#

Try out nemotron-3-super through ollama

gentle pasture
#

like when its not preview anymore

severe sonnet
drifting wind
#

100%

severe sonnet
#

use it while you can though, cuz why not

gentle pasture
#

and then just search for a new one?

drifting wind
#

Yup