Open canadiannomad opened 1 month ago
I have the same issue
Same here
I am having the same issue
I am seeing this too. My autocomplete Ex:
const PRIMESTOTWOHUNDRED = {{FILL_HERE}}
</QUERY>
TASK: Fill the {{FILL_HERE}} hole. Answer only with the CORRECT completion, and NOTHING ELSE. Do it now.
<COMPLETION>==========================================================================
==========================================================================
Completion:
<COMPLETION>[2, 3, 5, 7, 11, 13, 17, 19, 23, 29, 31, 37, 41, 43, 47, 53, 59, 61, 67, 71, 73, 79, 83, 89, 97, 101, 103, 107, 109, 113, 127, 131, 137, 139, 149, 151, 157, 163, 167, 173, 179, 181, 191, 193, 197, 199]
We are using an anthropic model bedrock/anthropic.claude-3-sonnet-20240229-v1:0 Our API Proxy has to strip out /n and /r from the response, but then the completion tag is not trimmed out because it is not on a new line.
@mtrevin93 @ilyanoskov @byrondelgado @kuamatzin @canadiannomad Now that there is an open model capable of extremely high autocomplete quality (Codestral) I would recommend using this. It is free until August, and even after will be a LOT cheaper than gpt-4. We have a set up guide here: https://docs.continue.dev/walkthroughs/set-up-codestral
I will still attempt to solve this problem, but there are likely to be many more imperfections due to the fact that gpt-4o is not at all trained for autocomplete.
@sestinj Thank you for your response. Unfortunately, a public API will not work for my use case - I must use an API hosted on my company's infrastructure which has access to all models supported on AWS bedrock.
Before submitting your bug report
Relevant environment info
Description
Every time I try to get a completion the output starts with "...the code that could be inserted"
Here is what I see in the logs:
And in the tab completion, I actually see the "" there.
Presumably that should be stripped.
To reproduce
Log output