Text Generation
Transformers
Safetensors
English
llama
conversational
Inference Endpoints
text-generation-inference

Prompt format?

#16
by anuragrawal - opened

Does it follow llama-2 prompt format? Here's the link to llama-2 prompt format: https://huggingface.co/blog/llama2#how-to-prompt-llama-2

It's not clear for me too. Here - https://mychen76.medium.com/tinyllama-colorist-fine-tuned-with-color-dataset-8cd1cf7e5665 - the guy is using this template:
<|im_start|>user
{question}<|im_end|>
<|im_start|>assistant
{answer}<|im_end|>

Llama2 template with, like used in a small dataset mlabonne/guanaco-llama2-1k also seems accepted.
It's also not clear, what to do, if there's a unlabelled text corpus of data.

<|system|>
{system_message}
<|user|>
{prompt}
<|assistant|>
is what the bloke says on the gguf version

It's sick how fast can generation be on cpu :O

The correct prompt format can be found in the Python code sample in the readme:

<|system|>
You are a friendly chatbot who always responds in the style of a pirate.</s>
<|user|>
How many helicopters can a human eat in one sitting?</s>
<|assistant|>

Sign up or log in to comment