Text Generation
Transformers
Safetensors
GGUF
English
stablelm
causal-lm
code
Eval Results
Inference Endpoints
7 papers

How to use FIM?

#2
by ex3ndr - opened

I want to add support to my VS Code plugin (https://marketplace.visualstudio.com/items?itemName=ex3ndr.llama-coder), but it is not clear what tags i need to use for FIM

Stability AI org
edited Jan 17

Hi @ex3ndr . The FIM tokens are

<fim_prefix>", "<fim_middle>","<fim_suffix>","<fim_pad>"

Let me know if this answers the question.

Is there a way to finetune this with FIM objective? I have a code dataset of my own (around 130K samples) and I would like to finetune this model on that datset but with FIM. Any help is appreciated <3

Stability AI org

Hi @sarthak247 , the bigcode project github has a version of megatron that includes FIM: https://github.com/bigcode-project/Megatron-LM

I wish to use it on WebStorm.

Stability AI org

We currently do not have any integrations with any IDE such as WebStorm. There are some VSCode extension that allow utilizing Huggingface's inference engine that might work well with this model.

ncoop57 changed discussion status to closed

Sign up or log in to comment