Text Generation
Transformers
Safetensors
English
llama
text-generation-inference
4-bit precision
gptq

Prompt clarification

#2
by hamishivi - opened

Hi, thanks for your interest in our model!

I saw your note about the prompt, and dug into it a little. What might not be clear (I'll fix this in our readmes shortly) is that we use an additional newline after <|assistant|>. I played around with your model briefly and this seemed to make a big difference, and might explain what you saw when using our model:

Without newline:

$ python script.py
...
*** Generate:
<s> <|user|> Tell me about AI
<|assistant|></s>
*** Pipeline:
<|user|> Tell me about AI
<|assistant|>

With newline:

$ python script.py
...
*** Generate:
<s> <|user|> Tell me about AI
<|assistant|>
AI stands for Artificial Intelligence. It is a branch of computer science that deals with the development of intelligent computer systems. These systems are designed to perform tasks that would normally require human intelligence, such as visual perception, speech recognition, decision-making, and language translation. AI systems use machine learning, a type of artificial intelligence that enables computers to learn from data and improve their performance without being explicitly programmed. AI is used in many industries, including healthcare, finance, and transportation, to automate processes, improve efficiency, and enhance the customer experience.</s>
*** Pipeline:
<|user|> Tell me about AI
<|assistant|>
AI stands for Artificial Intelligence. It is a branch of computer science that deals with the development of intelligent systems and machines that can perform tasks that would normally require human intelligence, such as visual perception, speech recognition, decision-making, language translation, and many others. The ultimate goal of AI research is to create machines or systems that can function autonomously in complex and unpredictable environments, and learn from experience to improve their performance over time.

To be clear, the full prompt should be like: <|user|>\nPROMPT\n<|assistant|>\n

It's very interesting those alternate prompts work, considering we didn't train the model on those!

Hi Hamish

Thanks very much for that clarification! Since making the READMEs I had heard from a couple of people that your prompt was working fine, so I planned to test again but hadn't had a chance to yet.

I will update my READMEs now to make this clear, and remove the other prompts. And yeah it's really interesting that they work despite no training! These models really are smart.

Thanks again for getting in touch, and for for all the great models you guys are uploading.

Sign up or log in to comment