Find answers from the community

Updated 11 months ago

anyone is hitting the 'rate limit

vvictor

anyone is hitting the 'rate limit exceeded' error while running llamaindex 5 line code examples on Windows?

17 comments

TTeemu

The OpenAI error? The free plans have very low rate limits which might be the issue

vvictor

low enough to trigger that error even for the 5 liner example?

TTeemu

I believe so, how many documents are you working with? Is this the starter tutorial?

LLogan M

(Those rate limits are extremely restrictive in the free tier)

vvictor

its the starter tutorial, so just one small txt file

LLogan M

For the embeddings, it's probably enough. The free tier is basically unusable for embeddings

Try using local embedding maybe

Plain Text

from llama_index import ServiceContext

index = VectorStoreIndex.from_documents(documents, service_context=ServiceContext.from_defaults(embed_model="local"))

vvictor

thanks I will try it. openAI doesn't allow me to upgrade from free tier unless I somehow used up the $5 credit

vvictor

now installing a whole bunch of dependencies

vvictor

like transformers, torchvision

vvictor

I will try after installing them

vvictor

it worked!

vvictor

Thanks for the help!

LLogan M

Yea, some dependencies needed to run locally vs Over an api

Glad it works!

vvictor

another quick question

vvictor

seems like the VectorStoreIndex.from_documents() is not using my GPU. I do have torch cuda installed and verified that it can recognize GPU

vvictor

any parameter i could set to use GPU for local embedding?

LLogan M

Plain Text

import torch
print(torch.cuda.is_available())

This prints true? It should be automatically using the GPU then for the embed model 🤔

Add a reply