Transformers Library issue

Anvil Q&A
1

I am trying to execute the following code in the server side:

from transformers import pipeline
from transformers import AutoModelForSequenceClassification
from transformers import AutoTokenizer, AutoModelForMaskedLM

MODEL = f"cardiffnlp/twitter-roberta-base-sentiment"
tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")
model = AutoModelForSequenceClassification.from_pretrained(MODEL)

@anvil.server.callable
def create_pipeline():
  sentiment_pipeline = pipeline("sentiment-analysis", MODEL)

But, I am getting an error:

anvil.server.ExecutionTerminatedError: Server code exited unexpectedly: e8ccc20b1f

The above got runs smoothly when I use the uplink but not in Anvil’s hosted servers.

Is there a way to overcome this issue?

Thanks a lot in advance.

2

hello, did you ever find a solution to this? currently experiencing the same issue

3

That’s memory, by the look of it. Generally better to run the big stuff via uplink.

4

totally but I’m not able to keep an uplink going indefinitely which is what I need for this app. Is there a way to increase the memory or unload some of it somehow?

5

So, there are a few ways you can approach this: 1) upgrade your plan and get some custom compute.
2) swap out your code for something less intensive.
3) switch out to a third party API and abstract the memory hit
Or
4) set up a lightsail or similar whose job it is purely to run your memory intensive service as an all day uplink.

1 Like
6

It is a memory related issue.

You can solve this quickly by using uplink in your local computer or ask Anvil support to increase you memory, if applicable.

I would suggest you to choose the first option as is the cheapest with the best performance.

1 Like