LocalLLaMA

3 readers

1 users here now

Community to discuss about Llama, the family of large language models created by Meta AI.

founded 1 year ago

MODERATORS

communick@poweruser.forum

Is DeepSeek Coder 1.3b meant to be this bad? (alien.top)

submitted 11 months ago by East-Awareness-249@alien.top to c/localllama@poweruser.forum

7 comments fedilink hide all child comments

I am using kobold.cpp and it couldn't code anything outside of hello world. Am I doing something wrong?

https://preview.redd.it/xdo6q7a25z1c1.png?width=1454&format=png&auto=webp&s=30d0eaed2c6d4d95070f2312a4bc3add0dcc2840

you are viewing a single comment's thread
view the rest of the comments

[–] FullOf_Bad_Ideas@alien.top 1 points 11 months ago (1 children)

Is that a base or some instruct-tuned fine-tune? It wouldn't be too much out of ordinary if it's base, they tend to get crazy. You can try setting repetition penalty to 1, might help a touch.

[–] AfterAte@alien.top 1 points 11 months ago

Also, set the temperature to 0.1 or 0.2. those two things helped me getting it to work nicely.